rumdl_lib/rules/

md044_proper_names.rs

use crate::utils::fast_hash;
use crate::utils::regex_cache::{escape_regex, get_cached_regex};

use crate::lint_context::SourcePositions;
use crate::rule::{Fix, LintError, LintResult, LintWarning, Rule, RuleCategory, Severity};
use crate::utils::frontmatter_values;
use std::collections::{HashMap, HashSet};
use std::sync::{Arc, Mutex};

mod md044_config;
pub(super) use md044_config::MD044Config;

type WarningPosition = (usize, usize, String, usize); // (line, column, found_name, proper_name_index)

/// Bound repeated delimiter scans in one line. Ordinary links need no index.
/// Queries can start after an opening byte that was escaped in the full line,
/// so the index also records those starts at their surrounding nesting depth.
#[derive(Default)]
struct FallbackDelimiters {
    scanned_bytes: usize,
    pairs: Option<HashMap<usize, usize>>,
}

impl FallbackDelimiters {
    fn find(&mut self, bytes: &[u8], from: usize, open: u8, close: u8) -> Option<usize> {
        debug_assert!(from > 0 && bytes.get(from - 1) == Some(&open));
        if self.pairs.is_none() && self.scanned_bytes >= bytes.len().saturating_mul(2) {
            let mut pairs = HashMap::new();
            let mut pending = Vec::new();
            let mut depth = 0usize;
            let mut pos = 0;
            while pos < bytes.len() {
                match bytes[pos] {
                    b'\\' => {
                        if bytes.get(pos + 1) == Some(&open) {
                            pending.push((depth, pos + 2));
                        }
                        pos += 1;
                    }
                    byte if byte == open => {
                        depth += 1;
                        pending.push((depth, pos + 1));
                    }
                    byte if byte == close => {
                        while pending.last().is_some_and(|&(level, _)| level == depth) {
                            let (_, start) = pending.pop().unwrap();
                            pairs.insert(start, pos + 1);
                        }
                        depth = depth.saturating_sub(1);
                    }
                    _ => {}
                }
                pos += 1;
            }
            self.pairs = Some(pairs);
        }
        if let Some(pairs) = &self.pairs {
            return pairs.get(&from).copied();
        }
        let end = Self::scan(bytes, from, open, close);
        self.scanned_bytes = self.scanned_bytes.saturating_add(end.unwrap_or(bytes.len()) - from);
        end
    }

    fn scan(bytes: &[u8], from: usize, open: u8, close: u8) -> Option<usize> {
        let mut depth = 1usize;
        let mut pos = from;
        while pos < bytes.len() {
            match bytes[pos] {
                b'\\' => pos += 1,
                byte if byte == open => depth += 1,
                byte if byte == close => {
                    depth -= 1;
                    if depth == 0 {
                        return Some(pos + 1);
                    }
                }
                _ => {}
            }
            pos += 1;
        }
        None
    }
}

/// Rule MD044: Proper names should be capitalized
///
/// See [docs/md044.md](../../docs/md044.md) for full documentation, configuration, and examples.
///
/// This rule is triggered when proper names are not capitalized correctly in the document.
/// For example, if you have defined "JavaScript" as a proper name, the rule will flag any
/// occurrences of "javascript" or "Javascript" as violations.
///
/// ## Purpose
///
/// Ensuring consistent capitalization of proper names improves document quality and
/// professionalism. This is especially important for technical documentation where
/// product names, programming languages, and technologies often have specific
/// capitalization conventions.
///
/// ## Configuration Options
///
/// The rule supports the following configuration options:
///
/// ```yaml
/// MD044:
///   names: []                # List of proper names to check for correct capitalization
///   code-blocks: false       # Whether to check code blocks (default: false)
///   whole-words: false       # Treat hyphen/snake-case compounds as whole tokens
/// ```
///
/// Example configuration:
///
/// ```yaml
/// MD044:
///   names: ["JavaScript", "Node.js", "TypeScript"]
///   code-blocks: true
/// ```
///
/// ## Performance Optimizations
///
/// This rule implements several performance optimizations:
///
/// 1. **Regex Caching**: Pre-compiles and caches regex patterns for each proper name
/// 2. **Content Caching**: Caches results based on content hashing for repeated checks
/// 3. **Efficient Text Processing**: Uses optimized algorithms to avoid redundant text processing
/// 4. **Smart Code Block Detection**: Efficiently identifies and optionally excludes code blocks
///
/// ## Edge Cases Handled
///
/// - **Word Boundaries**: Only matches complete words, not substrings within other words
/// - **Case Sensitivity**: Properly handles case-specific matching
/// - **Code Blocks**: Optionally checks code blocks (controlled by code-blocks setting)
/// - **Markdown Formatting**: Handles proper names within Markdown formatting elements
///
/// ## Fix Behavior
///
/// When fixing issues, this rule replaces incorrect capitalization with the correct form
/// as defined in the configuration.
///
#[derive(Clone)]
pub struct MD044ProperNames {
    config: MD044Config,
    // Cache the combined regex pattern string
    combined_pattern: Option<String>,
    // Lowercase variants map to the first configured spelling that accepts them.
    name_variants: HashMap<String, usize>,
    /// Lowercased `ignore_frontmatter_fields`, for case-insensitive lookup.
    ignore_fields: HashSet<String>,
    // Memoizes name violations keyed by content hash. Deliberately behind an
    // `Arc<Mutex<..>>` so it is SHARED across clones: rule instances are cloned
    // per config group and recreated for inline-config overrides, and the same
    // file's content is frequently re-checked (check then fix), so a shared
    // cache avoids recomputing. `check()` stays observationally pure (same ctx
    // in, same warnings out); the cache only affects how fast that answer is
    // produced. The lock is held only for the map get/insert, never across the
    // regex scan.
    content_cache: Arc<Mutex<HashMap<u64, Vec<WarningPosition>>>>,
}

impl MD044ProperNames {
    pub fn new(names: Vec<String>, code_blocks: bool) -> Self {
        let config = MD044Config {
            names,
            code_blocks,
            ..Default::default()
        };
        Self::from_config_struct(config)
    }

    // Helper function for consistent ASCII normalization
    fn ascii_normalize(s: &str) -> String {
        s.replace(['é', 'è', 'ê', 'ë'], "e")
            .replace(['à', 'á', 'â', 'ä', 'ã', 'å'], "a")
            .replace(['ï', 'î', 'í', 'ì'], "i")
            .replace(['ü', 'ú', 'ù', 'û'], "u")
            .replace(['ö', 'ó', 'ò', 'ô', 'õ'], "o")
            .replace('ñ', "n")
            .replace('ç', "c")
    }

    pub fn from_config_struct(config: MD044Config) -> Self {
        let combined_pattern = Self::create_combined_pattern(&config);
        let name_variants = Self::build_name_variants(&config);
        let ignore_fields = config
            .ignore_frontmatter_fields
            .iter()
            .flatten()
            .map(|f| f.to_lowercase())
            .collect();
        Self {
            config,
            combined_pattern,
            name_variants,
            ignore_fields,
            content_cache: Arc::new(Mutex::new(HashMap::new())),
        }
    }

    // Create a combined regex pattern for all proper names
    fn create_combined_pattern(config: &MD044Config) -> Option<String> {
        if config.names.is_empty() {
            return None;
        }

        // Create patterns for all names and their variations
        let mut patterns: Vec<String> = config
            .names
            .iter()
            .flat_map(|name| {
                let mut variations = vec![];
                let lower_name = name.to_lowercase();

                // Add the lowercase version
                variations.push(escape_regex(&lower_name));

                // Add version without dots
                let lower_name_no_dots = lower_name.replace('.', "");
                if lower_name != lower_name_no_dots {
                    variations.push(escape_regex(&lower_name_no_dots));
                }

                // Add ASCII-normalized versions for common accented characters
                let ascii_normalized = Self::ascii_normalize(&lower_name);

                if ascii_normalized != lower_name {
                    variations.push(escape_regex(&ascii_normalized));

                    // Also add version without dots
                    let ascii_no_dots = ascii_normalized.replace('.', "");
                    if ascii_normalized != ascii_no_dots {
                        variations.push(escape_regex(&ascii_no_dots));
                    }
                }

                variations
            })
            .collect();

        // Sort patterns by length (longest first) to avoid shorter patterns matching within longer ones
        patterns.sort_by_key(|b| std::cmp::Reverse(b.len()));

        // Combine all patterns into a single regex with capture groups
        // Don't use \b as it doesn't work with Unicode - we'll check boundaries manually
        Some(format!(r"(?i)({})", patterns.join("|")))
    }

    fn build_name_variants(config: &MD044Config) -> HashMap<String, usize> {
        let mut variants = HashMap::new();
        for (index, name) in config.names.iter().enumerate() {
            let lower_name = name.to_lowercase();
            variants.entry(lower_name.clone()).or_insert(index);

            let lower_no_dots = lower_name.replace('.', "");
            if lower_name != lower_no_dots {
                variants.entry(lower_no_dots).or_insert(index);
            }

            let ascii_normalized = Self::ascii_normalize(&lower_name);
            if ascii_normalized != lower_name {
                variants.entry(ascii_normalized.clone()).or_insert(index);

                let ascii_no_dots = ascii_normalized.replace('.', "");
                if ascii_normalized != ascii_no_dots {
                    variants.entry(ascii_no_dots).or_insert(index);
                }
            }
        }

        variants
    }

    // The caller has already checked for configured names present in the content.
    fn find_name_violations(&self, ctx: &crate::lint_context::LintContext) -> Vec<WarningPosition> {
        let content = ctx.content;
        // Check if we have cached results
        let hash = fast_hash(content);
        {
            // Use a separate scope for borrowing to minimize lock time
            if let Ok(cache) = self.content_cache.lock()
                && let Some(cached) = cache.get(&hash)
            {
                return cached.clone();
            }
        }

        let mut violations = Vec::new();

        // Get the regex from global cache
        let combined_regex = match &self.combined_pattern {
            Some(pattern) => match get_cached_regex(pattern) {
                Ok(regex) => regex,
                Err(_) => return Vec::new(),
            },
            None => return Vec::new(),
        };

        // Attribution is only needed when fields are actually excluded.
        let field_map = if self.ignore_fields.is_empty() {
            Vec::new()
        } else {
            frontmatter_values::field_map(ctx)
        };

        // Matches inside a linter directive comment (rumdl, markdownlint, Vale,
        // remark-lint), whatever lines it spans, are skipped by position so a fix
        // never rewrites a directive. Prose around the comment is still checked,
        // and a comment shown in code is an example, not a directive.
        let directive_comments =
            crate::inline_config::config_comment_ranges(content, |open| ctx.is_in_code_block_or_span(open));

        let mut wikilink_urls = None;

        // Use ctx.lines for better performance
        for (line_idx, line_info) in ctx.lines.iter().enumerate() {
            let line_num = line_idx + 1;
            let line = line_info.content(ctx.content);

            // Skip code fence lines (```language or ~~~language)
            let trimmed = line.trim_start();
            if trimmed.starts_with("```") || trimmed.starts_with("~~~") {
                continue;
            }

            // Skip if in code block (when code_blocks = false)
            if !self.config.code_blocks && line_info.in_code_block {
                continue;
            }

            // Skip if in HTML block (when html_elements = false)
            if !self.config.html_elements && line_info.in_html_block {
                continue;
            }

            // Skip HTML comments using pre-computed line flag
            if !self.config.html_comments && line_info.in_html_comment {
                continue;
            }

            // Skip JSX expressions and MDX comments (MDX flavor)
            if line_info.in_jsx_expression || line_info.in_mdx_comment {
                continue;
            }

            // Skip Obsidian comments (Obsidian flavor)
            if line_info.in_obsidian_comment {
                continue;
            }

            // For frontmatter lines, determine offset where checkable value content starts.
            // YAML keys should not be checked against proper names - only values.
            let fm_value_offset = if line_info.in_front_matter {
                frontmatter_values::value_offset(line)
            } else {
                0
            };
            if fm_value_offset == usize::MAX {
                continue;
            }
            if line_info.in_front_matter
                && let Some(Some(field)) = field_map.get(line_idx)
                && self.ignore_fields.contains(field)
            {
                continue;
            }
            let fm_value_span = if line_info.in_front_matter {
                frontmatter_values::value_span(line)
            } else {
                None
            };

            // Early return: skip lines that don't contain any potential matches
            let line_lower = line.to_lowercase();
            let has_line_matches = self.name_variants.keys().any(|name| line_lower.contains(name));

            if !has_line_matches {
                continue;
            }

            // Reuse fallback ranges across all matches on this line.
            let mut angle_urls = None;
            let mut fallback_code = None;
            let mut fallback_urls = None;

            // Use the combined regex to find all matches in one pass
            for cap in combined_regex.find_iter(line) {
                let found_name = &line[cap.start()..cap.end()];

                // Check word boundaries manually for Unicode support
                let start_pos = cap.start();
                let end_pos = cap.end();

                // Skip matches in the key portion of frontmatter lines
                if start_pos < fm_value_offset {
                    continue;
                }

                // Skip matches inside HTML tag attributes (handles multi-line tags)
                let byte_pos = line_info.byte_offset + start_pos;
                if crate::inline_config::offset_in_ranges(&directive_comments, byte_pos) {
                    continue;
                }
                if ctx.is_in_html_tag(byte_pos) {
                    continue;
                }

                // Skip matches inside a template shortcode tag. A shortcode is
                // resolved by name against a template, so `{{% nodejs %}}` names
                // the shortcode to invoke rather than mentioning a product, and
                // rewriting it points the invocation at a template that does not
                // exist. The range covers the tag only, so the Markdown body
                // between paired tags is still checked.
                if ctx.is_in_shortcode(byte_pos) {
                    continue;
                }

                if !Self::is_at_word_boundary(line, start_pos, true) || !Self::is_at_word_boundary(line, end_pos, false)
                {
                    continue; // Not at word boundary
                }

                if Self::is_file_extension_suffix(line, start_pos, found_name) {
                    continue;
                }

                if self.config.whole_words
                    && ((Self::is_joined_identifier_boundary(line, start_pos, true)
                        && !Self::is_underscore_emphasis_boundary(ctx, byte_pos, true))
                        || (Self::is_joined_identifier_boundary(line, end_pos, false)
                            && !Self::is_underscore_emphasis_boundary(ctx, byte_pos + found_name.len(), false)))
                {
                    continue;
                }

                // Skip if in inline code when code_blocks is false
                if !self.config.code_blocks {
                    if ctx.is_in_code_block_or_span(byte_pos) {
                        continue;
                    }
                    // pulldown-cmark doesn't parse markdown syntax inside HTML
                    // comments, HTML blocks, or frontmatter, so backtick-wrapped
                    // text isn't detected by is_in_code_block_or_span. Check directly.
                    if (line_info.in_html_comment || line_info.in_html_block || line_info.in_front_matter)
                        && crate::inline_config::offset_in_ranges(
                            fallback_code.get_or_insert_with(|| Self::backtick_code_ranges(line)),
                            start_pos,
                        )
                    {
                        continue;
                    }
                }

                // Skip if in link URL or reference definition
                if Self::is_in_link(ctx, byte_pos) {
                    continue;
                }

                // Skip if inside an angle-bracket URL (e.g., <https://...>)
                // The link parser skips autolinks inside HTML comments,
                // so we detect them directly in the line text.
                if crate::inline_config::offset_in_ranges(
                    angle_urls.get_or_insert_with(|| Self::angle_bracket_url_ranges(line)),
                    start_pos,
                ) {
                    continue;
                }

                // Skip if inside a Markdown inline link URL in contexts where
                // pulldown-cmark doesn't parse Markdown syntax (HTML comments,
                // HTML blocks, frontmatter).
                if (line_info.in_html_comment || line_info.in_html_block || line_info.in_front_matter)
                    && crate::inline_config::offset_in_ranges(
                        fallback_urls.get_or_insert_with(|| Self::markdown_link_url_ranges(line)),
                        start_pos,
                    )
                {
                    continue;
                }

                // Skip if inside the URL portion of a WikiLink followed by a
                // parenthesised destination — [[text]](url). pulldown-cmark
                // registers [[text]] as a WikiLink in ctx.links() but leaves the
                // (url) as plain text, so is_in_link() misses those bytes.
                if crate::inline_config::offset_in_ranges(
                    wikilink_urls.get_or_insert_with(|| Self::wikilink_url_ranges(ctx)),
                    byte_pos,
                ) {
                    continue;
                }

                // Skip if inside a bare URL (https://foo.com in plain prose).
                // Bare URLs are not in ctx.links() (flagging them is MD034's
                // domain), but a URL is still a URL: domains match
                // case-insensitively but paths are case-sensitive, so a
                // proper-name "fix" inside one can break the link.
                if ctx.is_in_bare_url(byte_pos) {
                    continue;
                }

                // Skip if inside a file path within a frontmatter value. Domains
                // match case-insensitively but paths are case-sensitive, so
                // rewriting a name inside one breaks the reference it points at.
                // Body prose is deliberately never consulted here; see
                // `is_in_path_like_token` for why the exemption stops at the
                // frontmatter boundary.
                if let Some(fm_value) = fm_value_span
                    && Self::is_in_path_like_token(line, start_pos, fm_value)
                {
                    continue;
                }

                // Find which proper name this matches
                if let Some(&proper_name_index) = self.name_variants.get(&found_name.to_lowercase()) {
                    // Only flag if it's not already correct
                    if found_name != self.config.names[proper_name_index] {
                        violations.push((line_num, cap.start() + 1, found_name.to_string(), proper_name_index));
                    }
                }
            }
        }

        // Store in cache (ignore if mutex is poisoned)
        if let Ok(mut cache) = self.content_cache.lock() {
            cache.insert(hash, violations.clone());
        }
        violations
    }

    /// Check if a byte position is within a link URL (not link text)
    ///
    /// Link text should be checked for proper names, but URLs should be skipped.
    /// For `[text](url)` - check text, skip url
    /// For `[text][ref]` - check text, skip reference portion
    /// For `[[page]]` (WikiLinks) - check the page name, skip brackets
    /// For `[[page|text]]` - check text, skip the page name it resolves to
    fn is_in_link(ctx: &crate::lint_context::LintContext, byte_pos: usize) -> bool {
        use pulldown_cmark::LinkType;

        if let Some(link) = ctx.link_containing(byte_pos) {
            // WikiLinks [[text]] start with '[[', regular links [text] start with '['
            let (text_start, text_end) = if matches!(link.link_type, LinkType::WikiLink { .. }) {
                // The displayed part of a wikilink runs to the closing `]]`, and
                // starts after the pipe when there is one: in `[[target|display]]`
                // the target is a reference like a URL, and renaming it changes
                // what the link resolves to without changing what a reader sees.
                // Measuring from the start instead would put the window over the
                // target, which is what used to happen.
                let span = &ctx.content[link.byte_offset..link.byte_end];
                let start = match span.find('|') {
                    Some(pipe) => link.byte_offset + pipe + 1,
                    None => link.byte_offset + 2,
                };
                (start, link.byte_end.saturating_sub(2))
            } else {
                let start = link.byte_offset + 1;
                (start, start + link.text.len())
            };

            // If position is within the text portion, skip only if text is a URL.
            // WikiLinks use the page name as both text and url; never treat them
            // as bare-domain URLs regardless of whether the name contains dots.
            if byte_pos >= text_start && byte_pos < text_end {
                let is_wikilink = matches!(link.link_type, LinkType::WikiLink { .. });
                if Self::link_text_is_url(&link.text)
                    || (!is_wikilink && Self::link_text_matches_link_url(&link.text, &link.url))
                {
                    return true;
                }
                // Displayed text can hold an image, and that image's destination is
                // a URL like any other: renaming inside `[[Target|see
                // ![alt](image.png)]]` breaks the image. Nested *links* need no such
                // handling, since they are in `ctx.links()` too and the search above
                // already picks the innermost one.
                return Self::image_verdict(ctx, byte_pos).unwrap_or(false);
            }
            // Position is in the URL/reference portion, skip it
            return true;
        }

        if let Some(verdict) = Self::image_verdict(ctx, byte_pos) {
            return verdict;
        }

        // Check pre-computed reference definitions
        ctx.is_in_reference_def(byte_pos)
    }

    /// Whether a byte position inside an image should be skipped.
    ///
    /// `Some(true)` for the URL or reference portion, `Some(false)` for the alt
    /// text, and `None` when the position is in no image at all.
    fn image_verdict(ctx: &crate::lint_context::LintContext, byte_pos: usize) -> Option<bool> {
        let image = ctx.image_containing(byte_pos)?;

        // Image starts with '![' so alt text starts at byte_offset + 2
        let alt_start = image.byte_offset + 2;
        let alt_end = alt_start + image.alt_text.len();

        // Inside the alt text, don't skip; otherwise this is the URL or reference
        Some(!(byte_pos >= alt_start && byte_pos < alt_end))
    }

    /// Check if link text is a URL that should not have proper name corrections.
    fn link_text_is_url(text: &str) -> bool {
        let lower = text.trim().to_ascii_lowercase();
        lower.starts_with("http://")
            || lower.starts_with("https://")
            || lower.starts_with("www.")
            || lower.starts_with("//")
    }

    /// Check if link text is the bare hostname/path of its destination URL.
    ///
    /// When the display text is the URL with the scheme stripped (e.g.,
    /// `[example.github.io](https://example.github.io)`), the text is a domain
    /// label, not a prose reference to a product, and should not be corrected.
    ///
    /// Requires the text to contain a dot, which distinguishes domain-like display
    /// text from single-word WikiLink targets (e.g. `[[javascript]]`) where
    /// `url == text` but neither is a domain name. Dotted WikiLink targets are
    /// excluded separately via the `!is_wikilink` guard in `is_in_link`. Comparison
    /// is case-insensitive because URL schemes and hostnames are case-insensitive.
    fn link_text_matches_link_url(text: &str, url: &str) -> bool {
        let text = text.trim();
        // Only domain-like text (containing a dot) can be a bare hostname.
        if !text.contains('.') {
            return false;
        }
        let url_lower = url.to_ascii_lowercase();
        let url_without_scheme = url_lower
            .strip_prefix("https://")
            .or_else(|| url_lower.strip_prefix("http://"))
            .or_else(|| url_lower.strip_prefix("//"))
            .unwrap_or(&url_lower);
        let text_lower = text.to_ascii_lowercase();
        // Exact match: text equals the URL with the scheme removed.
        if url_without_scheme == text_lower.as_str() {
            return true;
        }
        // Prefix match: text is the hostname portion and the URL has a path/query/fragment.
        url_without_scheme.len() > text_lower.len()
            && url_without_scheme.starts_with(text_lower.as_str())
            && matches!(
                url_without_scheme.as_bytes().get(text_lower.len()),
                Some(b'/') | Some(b'?') | Some(b'#')
            )
    }

    /// Collect disjoint, source-ordered angle-bracket URL ranges (`<scheme://...>`).
    ///
    /// The link parser skips autolinks inside HTML comments, so `ctx.links()` won't
    /// contain them. This function detects angle-bracket URLs directly in the line
    /// text, covering both HTML comments and regular text as a safety net.
    fn angle_bracket_url_ranges(line: &str) -> Vec<std::ops::Range<usize>> {
        let mut ranges = Vec::new();
        let bytes = line.as_bytes();
        let len = bytes.len();
        let mut i = 0;
        while i < len {
            if bytes[i] == b'<' {
                let after_open = i + 1;
                // Check for a valid URI scheme per CommonMark autolink spec:
                // scheme = [a-zA-Z][a-zA-Z0-9+.-]{0,31}
                // followed by ':'
                if after_open < len && bytes[after_open].is_ascii_alphabetic() {
                    let mut s = after_open + 1;
                    let scheme_max = (after_open + 32).min(len);
                    while s < scheme_max
                        && (bytes[s].is_ascii_alphanumeric()
                            || bytes[s] == b'+'
                            || bytes[s] == b'-'
                            || bytes[s] == b'.')
                    {
                        s += 1;
                    }
                    if s < len && bytes[s] == b':' {
                        // Valid scheme found; scan for closing '>' with no spaces or '<'
                        let mut j = s + 1;
                        let mut found_close = false;
                        while j < len {
                            match bytes[j] {
                                b'>' => {
                                    found_close = true;
                                    break;
                                }
                                b' ' | b'<' => break,
                                _ => j += 1,
                            }
                        }
                        if found_close {
                            ranges.push(i..j + 1);
                            i = j + 1;
                            continue;
                        }
                    }
                }
            }
            i += 1;
        }
        ranges
    }

    #[cfg(test)]
    fn is_in_angle_bracket_url(line: &str, pos: usize) -> bool {
        crate::inline_config::offset_in_ranges(&Self::angle_bracket_url_ranges(line), pos)
    }

    /// Destinations immediately following parsed WikiLinks. The parser leaves
    /// their parentheses as prose, so ordinary link ranges do not cover them.
    /// Scan once for all requested openers, preserving the fallback's raw-byte
    /// nesting, backslash escapes and rejection of unescaped whitespace.
    fn wikilink_url_ranges(ctx: &crate::lint_context::LintContext) -> Vec<std::ops::Range<usize>> {
        use pulldown_cmark::LinkType;
        let content = ctx.content.as_bytes();
        let starts: HashSet<_> = ctx
            .links()
            .iter()
            .filter(|link| matches!(link.link_type, LinkType::WikiLink { .. }))
            .map(|link| link.byte_end)
            .filter(|&end| content.get(end) == Some(&b'('))
            .collect();
        let mut ranges: Vec<std::ops::Range<usize>> = Vec::new();
        let Some(mut pos) = starts.iter().min().copied() else {
            return ranges;
        };
        let last_start = *starts.iter().max().unwrap();
        // Only requested destinations need stack entries; other parentheses
        // contribute to depth without allocating per nesting level.
        let mut stack = Vec::new();
        let mut depth = 0usize;
        while pos < content.len() {
            match content[pos] {
                b'\\' => pos += 1,
                b'(' => {
                    depth += 1;
                    if starts.contains(&pos) {
                        stack.push((pos, depth));
                    }
                }
                b')' => {
                    if stack.last().is_some_and(|&(_, level)| level == depth) {
                        let (open, _) = stack.pop().unwrap();
                        // Balanced destinations are nested or disjoint. An outer
                        // destination subsumes the inner ranges at the end of the
                        // vector, leaving source-ordered disjoint ranges for lookup.
                        while ranges.last().is_some_and(|range| range.start > open) {
                            ranges.pop();
                        }
                        ranges.push(open + 1..pos + 1);
                    }
                    depth = depth.saturating_sub(1);
                }
                // Every still-open destination crosses this whitespace and is invalid.
                b' ' | b'\t' | b'\n' | b'\r' => {
                    stack.clear();
                    depth = 0;
                }
                _ => {}
            }
            if pos >= last_start && stack.is_empty() {
                break;
            }
            pos += 1;
        }
        ranges
    }

    /// Collect a line's Markdown link destinations and reference labels.
    ///
    /// Used as a text-level fallback for HTML comments, HTML blocks, and
    /// frontmatter where pulldown-cmark skips link parsing entirely. Operates on
    /// raw line bytes and therefore cannot distinguish real links from text that
    /// merely resembles link syntax; do not call on regular markdown lines.
    /// - `[text](url)` — bytes after `(` through the closing `)`.
    /// - `[text][ref]` — bytes after the second `[` through its closing `]`.
    fn markdown_link_url_ranges(line: &str) -> Vec<std::ops::Range<usize>> {
        let mut ranges = Vec::new();
        let bytes = line.as_bytes();
        let mut brackets = FallbackDelimiters::default();
        let mut parentheses = FallbackDelimiters::default();
        let mut i = 0;
        while i < bytes.len() {
            // Preserve the fallback's existing treatment of double escapes.
            if bytes[i] == b'['
                && (i == 0 || bytes[i - 1] != b'\\' || (i >= 2 && bytes[i - 2] == b'\\'))
                && let Some(j) = brackets.find(bytes, i + 1, b'[', b']')
            {
                let end = match bytes.get(j) {
                    Some(b'(') => parentheses.find(bytes, j + 1, b'(', b')'),
                    Some(b'[') => brackets.find(bytes, j + 1, b'[', b']'),
                    _ => None,
                };
                if let Some(end) = end {
                    ranges.push(j + 1..end);
                    i = end;
                    continue;
                }
            }
            i += 1;
        }
        ranges
    }

    /// Collect content ranges of backtick-delimited code on a line.
    ///
    /// pulldown-cmark does not parse markdown syntax inside HTML comments, so
    /// `ctx.is_in_code_block_or_span` returns false for backtick-wrapped text
    /// within comments. This function detects backtick code spans directly in
    /// the line text, looking for matching runs of exactly the same length.
    fn backtick_code_ranges(line: &str) -> Vec<std::ops::Range<usize>> {
        let mut ranges = Vec::new();
        let bytes = line.as_bytes();
        let len = bytes.len();
        let mut i = 0;
        while i < len {
            if bytes[i] == b'`' {
                // Count the opening backtick sequence length
                let open_start = i;
                while i < len && bytes[i] == b'`' {
                    i += 1;
                }
                let tick_len = i - open_start;

                // Scan forward for a closing sequence of exactly tick_len backticks
                while i < len {
                    if bytes[i] == b'`' {
                        let close_start = i;
                        while i < len && bytes[i] == b'`' {
                            i += 1;
                        }
                        if i - close_start == tick_len {
                            // Matched pair found; the code span content is between
                            // the end of the opening backticks and the start of the
                            // closing backticks (exclusive of the backticks themselves).
                            let content_start = open_start + tick_len;
                            let content_end = close_start;
                            ranges.push(content_start..content_end);
                            // Continue scanning after this pair
                            break;
                        }
                        // Not the right length; keep scanning
                    } else {
                        i += 1;
                    }
                }
            } else {
                i += 1;
            }
        }
        ranges
    }

    #[cfg(test)]
    fn is_in_backtick_code_in_line(line: &str, pos: usize) -> bool {
        crate::inline_config::offset_in_ranges(&Self::backtick_code_ranges(line), pos)
    }

    // Check if a character is a word boundary (handles Unicode)
    fn is_word_boundary_char(c: char) -> bool {
        !c.is_alphanumeric()
    }

    // Check if position is at a word boundary using byte-level lookups.
    fn is_at_word_boundary(content: &str, pos: usize, is_start: bool) -> bool {
        if is_start {
            if pos == 0 {
                return true;
            }
            match content[..pos].chars().next_back() {
                None => true,
                Some(c) => Self::is_word_boundary_char(c),
            }
        } else {
            if pos >= content.len() {
                return true;
            }
            match content[pos..].chars().next() {
                None => true,
                Some(c) => Self::is_word_boundary_char(c),
            }
        }
    }

    /// Preserve a name immediately after a single dot, as in `favorite.toml`
    /// or `.toml`. Check only the match and its immediate prefix: scanning for
    /// a whole filename would collide with Markdown punctuation. The stem
    /// stays checkable, and `node.js` still matches the whole name `Node.js`.
    fn is_file_extension_suffix(line: &str, start: usize, name: &str) -> bool {
        let Some(prefix) = line[..start].strip_suffix('.') else {
            return false;
        };

        // An ellipsis can introduce prose without a space. Multiword names
        // and names with other punctuation (such as `.NET`) are not suffixes.
        !prefix.ends_with('.')
            && name.starts_with(char::is_alphanumeric)
            && name.ends_with(char::is_alphanumeric)
            && name
                .chars()
                .all(|character| character.is_alphanumeric() || matches!(character, '.' | '-' | '_'))
    }

    /// Whether `pos` is separated from another Unicode-alphanumeric
    /// identifier component by a run of `-` and/or `_`.
    ///
    /// Markdown emphasis is handled separately using the parser's emphasis
    /// spans; raw separator scanning alone cannot distinguish `foo__name__bar`
    /// (literal intraword underscores) from `foo-__name__-bar` (strong
    /// emphasis between identifier-looking punctuation).
    fn is_joined_identifier_boundary(content: &str, pos: usize, is_start: bool) -> bool {
        if is_start {
            let mut chars = content[..pos].char_indices().rev();
            let mut saw_separator = false;
            for (_, character) in &mut chars {
                if character == '-' || character == '_' {
                    saw_separator = true;
                } else {
                    return saw_separator && character.is_alphanumeric();
                }
            }
            false
        } else {
            let mut saw_separator = false;
            for character in content[pos..].chars() {
                if character == '-' || character == '_' {
                    saw_separator = true;
                } else {
                    return saw_separator && character.is_alphanumeric();
                }
            }
            false
        }
    }

    /// Whether this exact boundary touches an underscore emphasis delimiter.
    /// Being anywhere inside emphasis is not enough: `_foo_abc_bar_` still
    /// contains an identifier, whereas `foo-__abc__-bar` emphasizes just `abc`.
    /// Check the two boundaries independently so `_abc_def_` remains protected.
    fn is_underscore_emphasis_boundary(ctx: &crate::lint_context::LintContext, pos: usize, is_start: bool) -> bool {
        ctx.source_emphasis_spans().iter().any(|span| {
            if span.marker != '_' {
                return false;
            }
            let width = if span.is_strong { 2 } else { 1 };
            if is_start {
                span.byte_offset + width == pos
            } else {
                span.byte_end.checked_sub(width) == Some(pos)
            }
        })
    }

    /// Whether the match at `match_start` sits inside a file path, which must
    /// not be rewritten. `fm_value` is the semantic value span of the
    /// frontmatter line the match was found on.
    ///
    /// This exemption is deliberately scoped to frontmatter values only and
    /// is never applied to body prose. In frontmatter the value span is known
    /// exactly (`frontmatter_values::value_span`), so token bounds can be clamped to
    /// it with no risk of crossing into unrelated syntax. Body prose has no
    /// such known span: a token there must be delimited by scanning the raw
    /// line for whitespace and Markdown punctuation, and that scan
    /// unavoidably collides with Markdown link/image/wikilink syntax (see the
    /// module-level history of defects from trying this). Reusing this
    /// function for body text is a structural mismatch, not a missing edge
    /// case, so `fm_value` is required rather than optional: a caller cannot
    /// accidentally invoke this for a line that has no known value span.
    ///
    /// A slash is mandatory for exempting a whole path token: a blanket
    /// dotted-token exemption would swallow proper names such as `Node.js`.
    /// Extension suffixes are protected separately without exempting the stem.
    ///
    /// The 3+ segment sole-value signal only fires for a genuinely
    /// single-token value. A quoted value that collapsed from several
    /// whitespace-separated words (`"myapp/gitlab github/bitbucket"`) must
    /// instead satisfy the path-prefix or file-extension signal; otherwise
    /// two unrelated slash-pairs joined by a space would vacuously look like
    /// a 3-segment path. This means an extensionless path containing a
    /// literal space (`docs/My App/myapp`) is no longer exempt, a deliberate
    /// narrowing rather than an oversight.
    fn is_in_path_like_token(line: &str, match_start: usize, fm_value: (usize, usize)) -> bool {
        let (value_start, value_end) = fm_value;
        if match_start < value_start || match_start >= value_end {
            return false;
        }

        // A quoted scalar is one token even with spaces in it, but only when
        // every whitespace-separated word in it carries a slash (e.g. a path
        // containing a space, `docs/My App/myapp`). A quoted sentence with
        // ordinary prose words (`"We support github/gitlab/bitbucket now"`)
        // falls back to per-word tokenization instead, otherwise quoting
        // alone would make the token span the whole value and vacuously
        // satisfy the sole-value check below.
        let quoted_words: Vec<&str> = if frontmatter_values::value_is_quoted(line, value_start) {
            line[value_start..value_end].split_whitespace().collect()
        } else {
            Vec::new()
        };
        let is_single_quoted_path = !quoted_words.is_empty() && quoted_words.iter().all(|word| word.contains('/'));
        // Collapsing several whitespace-separated words into one token is
        // only safe evidence for signals (a) and (b): a shared slash prefix
        // or a real extension on the last segment. It is not evidence for
        // the segment-count signal below, which assumes a single word split
        // into path segments by '/'; two unrelated slash-pairs joined by a
        // space (`"myapp/gitlab github/bitbucket"`) would otherwise satisfy
        // that count vacuously.
        let is_multi_word_collapse = is_single_quoted_path && quoted_words.len() > 1;

        let (raw_start, raw_end) = if is_single_quoted_path {
            (value_start, value_end)
        } else {
            frontmatter_values::token_bounds(line, match_start, value_start, value_end)
        };

        let (start, end) = frontmatter_values::trim_token_bounds(line, raw_start, raw_end);
        if match_start < start || match_start >= end {
            return false;
        }

        let token = &line[start..end];
        if !token.contains('/') {
            return false;
        }
        if token.starts_with('/') || token.starts_with("./") || token.starts_with("../") || token.starts_with("~/") {
            return true;
        }
        if token.rsplit('/').next().is_some_and(|seg| seg.contains('.')) {
            return true;
        }

        if is_multi_word_collapse {
            return false;
        }

        // Three or more segments is only a path signal when the token is the
        // entire frontmatter value. In prose, `github/gitlab/bitbucket` is
        // shorthand, not a path.
        let sole_value = {
            let (ts, te) = frontmatter_values::trim_token_bounds(line, value_start, value_end);
            ts == start && te == end
        };
        sole_value && token.split('/').filter(|s| !s.is_empty()).count() >= 3
    }
}

impl Rule for MD044ProperNames {
    fn native_format_policy(&self) -> Option<crate::format_proof::NativeFormatPolicy> {
        (self.config == Default::default()).then_some(crate::format_proof::NativeFormatPolicy::for_rule::<Self>(44))
    }

    fn name(&self) -> &'static str {
        "MD044"
    }

    fn description(&self) -> &'static str {
        "Proper names should have the correct capitalization"
    }

    fn category(&self) -> RuleCategory {
        RuleCategory::Other
    }

    fn should_skip(&self, ctx: &crate::lint_context::LintContext) -> bool {
        if self.config.names.is_empty() {
            return true;
        }
        // Quick check if any configured name variants exist (case-insensitive)
        let content_lower = if ctx.content.is_ascii() {
            ctx.content.to_ascii_lowercase()
        } else {
            ctx.content.to_lowercase()
        };
        !self.name_variants.keys().any(|name| content_lower.contains(name))
    }

    fn check(&self, ctx: &crate::lint_context::LintContext) -> LintResult {
        let content = ctx.content;
        if content.is_empty() || self.config.names.is_empty() || self.combined_pattern.is_none() {
            return Ok(Vec::new());
        }

        // Compute lowercase content once and reuse across all checks
        let content_lower = if content.is_ascii() {
            content.to_ascii_lowercase()
        } else {
            content.to_lowercase()
        };

        // Early return: use pre-computed name_variants for the quick check
        let has_potential_matches = self.name_variants.keys().any(|name| content_lower.contains(name));

        if !has_potential_matches {
            return Ok(Vec::new());
        }
        let violations = self.find_name_violations(ctx);

        let mut positions = SourcePositions::new(content, &ctx.lines);
        let warnings = violations
            .into_iter()
            .map(|(line, column, found_name, proper_name_index)| {
                let proper_name = &self.config.names[proper_name_index];
                // `column` is a 1-indexed byte offset into the line (from regex .start() + 1).
                // Build the Fix range directly in bytes to avoid the character-based
                // line_col_to_byte_range_with_length function, which would misinterpret
                // the byte offset as a character count on lines with multi-byte content.
                let line_start = ctx.line_start_byte(line).unwrap_or(0);
                let byte_start = line_start + (column - 1);
                let byte_end = byte_start + found_name.len();
                // The displayed columns are character offsets; convert from the byte
                // offset within the line so they are correct on multi-byte lines.
                let char_col = positions.column_on_line(line - 1, byte_start) + 1;
                LintWarning {
                    rule_name: Some(self.name().to_string()),
                    line,
                    column: char_col,
                    end_line: line,
                    end_column: char_col + found_name.chars().count(),
                    message: format!("Proper name '{found_name}' should be '{proper_name}'"),
                    severity: Severity::Warning,
                    fix: Some(Fix::new(byte_start..byte_end, proper_name.clone())),
                }
            })
            .collect();

        Ok(warnings)
    }

    fn fixes_from_diagnostics(&self) -> bool {
        true
    }

    fn fix(&self, ctx: &crate::lint_context::LintContext) -> Result<String, LintError> {
        if self.should_skip(ctx) {
            return Ok(ctx.content.to_string());
        }
        let warnings = self.check(ctx)?;
        if warnings.is_empty() {
            return Ok(ctx.content.to_string());
        }
        let warnings =
            crate::utils::fix_utils::filter_warnings_by_inline_config(warnings, ctx.inline_config(), self.name());
        crate::utils::fix_utils::apply_warning_fixes(ctx.content, &warnings)
            .map_err(crate::rule::LintError::InvalidInput)
    }

    fn as_any(&self) -> &dyn std::any::Any {
        self
    }

    crate::impl_rule_config_methods!(MD044Config);
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::lint_context::LintContext;

    fn create_context(content: &str) -> LintContext<'_> {
        LintContext::new(content, crate::config::MarkdownFlavor::Standard, None)
    }

    fn field_map_for(content: &str) -> Vec<Option<String>> {
        let ctx = create_context(content);
        frontmatter_values::field_map(&ctx)
    }

    fn whole_word_rule(names: &[&str]) -> MD044ProperNames {
        MD044ProperNames::from_config_struct(MD044Config {
            names: names.iter().map(|name| (*name).to_string()).collect(),
            whole_words: true,
            ..Default::default()
        })
    }

    #[test]
    fn file_extension_suffixes_preserve_case_but_prose_is_checked() {
        let content = "I like toml in my favorite.toml file.\n\
                       Extensions: .toml (.toml) [.toml] \".toml\" '.toml' “.toml” ‘.toml’.\n\
                       Files: config.toml.example v2.toml café.toml my_conf.toml foo-.toml file(1).toml.\n\
                       Patterns: *.toml **.toml** _.toml_ /.toml \\.toml.\n\
                       Next.js uses js, and toml is a format.\n";
        let expected = content
            .replacen("I like toml", "I like TOML", 1)
            .replace("uses js, and toml", "uses JS, and TOML");

        for whole_words in [false, true] {
            let rule = MD044ProperNames::from_config_struct(MD044Config {
                names: vec!["TOML".to_string(), "JS".to_string()],
                whole_words,
                ..Default::default()
            });
            let ctx = create_context(content);
            assert_eq!(rule.check(&ctx).unwrap().len(), 3);
            assert_eq!(rule.fix(&ctx).unwrap(), expected);
            assert!(rule.check(&create_context(&expected)).unwrap().is_empty());
            assert_eq!(rule.fix(&create_context(&expected)).unwrap(), expected);
        }
    }

    #[test]
    fn file_extensions_keep_dotted_names_stems_and_function_calls_checkable() {
        let rule = MD044ProperNames::new(
            ["TOML", "Oklch", "Node.js", "MyApp", ".NET", "GitHub"]
                .map(str::to_string)
                .to_vec(),
            false,
        );
        let content = "Use node.js and .net for myapp.\n\
                       See docs/myapp.md and docs/myapp(1).md.\n\
                       Oklch syntax is oklch(.5 .5 255); github(tm) is a reference.\n\
                       I like TOML in my favorite.toml file.\n\
                       Use `favorite.toml` and `oklch(.5 .5 255)`.\n";
        let expected = "Use Node.js and .NET for MyApp.\n\
                        See docs/MyApp.md and docs/MyApp(1).md.\n\
                        Oklch syntax is Oklch(.5 .5 255); GitHub(tm) is a reference.\n\
                        I like TOML in my favorite.toml file.\n\
                        Use `favorite.toml` and `oklch(.5 .5 255)`.\n";
        let ctx = create_context(content);
        assert_eq!(rule.check(&ctx).unwrap().len(), 7);
        assert_eq!(rule.fix(&ctx).unwrap(), expected);
        assert!(rule.check(&create_context(expected)).unwrap().is_empty());
    }

    #[test]
    fn file_extensions_do_not_exempt_ellipsis_or_multiword_prose() {
        let rule = MD044ProperNames::new(vec!["TOML".to_string(), "My App".to_string()], false);
        let content = "Wait...toml or …toml; e.g. toml and file. toml.\n\
                       Sentence.my app is still prose.\n";
        let expected = "Wait...TOML or …TOML; e.g. TOML and file. TOML.\n\
                        Sentence.My App is still prose.\n";
        let ctx = create_context(content);
        assert_eq!(rule.check(&ctx).unwrap().len(), 5);
        assert_eq!(rule.fix(&ctx).unwrap(), expected);
    }

    #[test]
    fn file_extension_syntax_applies_to_dotted_tokens_and_html_prose() {
        let rule = MD044ProperNames::new(["TOML", "Node.js", "GitHub"].map(str::to_string).to_vec(), false);
        // A single dot is a syntax cue, not proof of a filename. In particular,
        // missing a space after a sentence period can hide a capitalization error.
        let content = ".github api.node.js api.github.com i.e.toml\n\
                       node.js github and toml\n\
                       <!-- favorite.toml toml -->\n\
                       <div>favorite.toml toml</div>\n";
        let expected = ".github api.node.js api.github.com i.e.toml\n\
                        Node.js GitHub and TOML\n\
                        <!-- favorite.toml TOML -->\n\
                        <div>favorite.toml TOML</div>\n";
        let ctx = create_context(content);
        assert_eq!(rule.check(&ctx).unwrap().len(), 5);
        assert_eq!(rule.fix(&ctx).unwrap(), expected);
        assert!(rule.check(&create_context(expected)).unwrap().is_empty());
    }

    #[test]
    fn file_extensions_in_markdown_preserve_suffixes_and_check_prose() {
        let content = "**favorite.toml** *favorite.toml* _favorite.toml_\n\
                       [favorite.toml](other.md) [toml](favorite.toml)\n\
                       ![favorite.toml](img.png) ![toml](img.toml)\n\
                       [[favorite.toml]] [[Page|favorite.toml]] [[Page|toml]]\n\
                       `favorite.toml toml`\n\
                       ```text\n\
                       favorite.toml toml\n\
                       ```\n";
        for code_blocks in [false, true] {
            let rule = MD044ProperNames::new(vec!["TOML".to_string()], code_blocks);
            let mut expected = content.replace("[toml]", "[TOML]").replace("|toml]]", "|TOML]]");
            if code_blocks {
                expected = expected.replace("favorite.toml toml", "favorite.toml TOML");
            }
            for flavor in [
                crate::config::MarkdownFlavor::Standard,
                crate::config::MarkdownFlavor::Obsidian,
            ] {
                let ctx = LintContext::new(content, flavor, None);
                assert_eq!(rule.check(&ctx).unwrap().len(), if code_blocks { 5 } else { 3 });
                assert_eq!(rule.fix(&ctx).unwrap(), expected);
            }
        }
    }

    #[test]
    fn file_extensions_in_frontmatter_preserve_suffixes_and_check_prose() {
        let rule = MD044ProperNames::new(["TOML", "Node.js", "MyApp"].map(str::to_string).to_vec(), false);
        for content in [
            "---\ntoml: favorite.toml\ntitle: node.js guide to toml\nimage: myapp.png\npath: docs/myapp.md\n---\n",
            "+++\ntoml = \"favorite.toml\"\ntitle = \"node.js guide to toml\"\nimage = \"myapp.png\"\npath = \"docs/myapp.md\"\n+++\n",
        ] {
            let expected = content
                .replace("node.js guide to toml", "Node.js guide to TOML")
                .replace("myapp.png", "MyApp.png");
            let ctx = create_context(content);
            assert_eq!(rule.check(&ctx).unwrap().len(), 3);
            assert_eq!(rule.fix(&ctx).unwrap(), expected);
        }
    }

    #[test]
    fn whole_words_preserves_compound_identifiers_but_checks_standalone_names() {
        let rule = whole_word_rule(&["ABC"]);
        let content = "123-abc-def\nfoo_abc_bar\nabc\n(abc) [abc]\n";
        let ctx = create_context(content);

        let warnings = rule.check(&ctx).unwrap();
        assert_eq!(warnings.len(), 3);
        assert_eq!(
            warnings.iter().map(|warning| warning.line).collect::<Vec<_>>(),
            [3, 4, 4]
        );
        assert_eq!(rule.fix(&ctx).unwrap(), "123-abc-def\nfoo_abc_bar\nABC\n(ABC) [ABC]\n");
    }

    #[test]
    fn whole_words_preserves_identifiers_inside_emphasis() {
        let rule = whole_word_rule(&["ABC"]);
        let content = "_123-abc-def_ _foo_abc_bar_ __foo__abc__bar__\n\
                       _abc_def_ _foo_abc_\n\
                       _abc_ __abc__ ___abc___ foo-__abc__-bar\n";
        let ctx = create_context(content);
        let warnings = rule.check(&ctx).unwrap();
        assert_eq!(warnings.len(), 4);
        assert!(warnings.iter().all(|warning| warning.line == 3));
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(
            fixed,
            "_123-abc-def_ _foo_abc_bar_ __foo__abc__bar__\n\
             _abc_def_ _foo_abc_\n\
             _ABC_ __ABC__ ___ABC___ foo-__ABC__-bar\n"
        );
        assert!(rule.check(&create_context(&fixed)).unwrap().is_empty());
        assert_eq!(rule.fix(&create_context(&fixed)).unwrap(), fixed);
    }

    #[test]
    fn whole_words_distinguishes_unicode_compounds_from_markdown_emphasis() {
        let rule = whole_word_rule(&["ABC"]);
        let content = "café_abc_menu 東京-abc-駅 foo__abc__bar _abc_ __abc__ foo-__abc__-bar\n";
        let ctx = create_context(content);

        assert_eq!(rule.check(&ctx).unwrap().len(), 3);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(
            fixed,
            "café_abc_menu 東京-abc-駅 foo__abc__bar _ABC_ __ABC__ foo-__ABC__-bar\n"
        );
        assert_eq!(rule.fix(&create_context(&fixed)).unwrap(), fixed);
    }

    #[test]
    fn repeated_intraword_underscores_keep_legacy_behavior_without_whole_words() {
        let rule = MD044ProperNames::new(vec!["ABC".to_string()], false);
        let content = "foo__abc__bar\n";
        let ctx = create_context(content);

        assert_eq!(rule.check(&ctx).unwrap().len(), 1);
        assert_eq!(rule.fix(&ctx).unwrap(), "foo__ABC__bar\n");
    }

    #[test]
    fn whole_words_respects_code_and_link_exclusions_and_multiword_names() {
        let rule = whole_word_rule(&["ABC", "My App"]);
        let content = "`abc` [label](https://example.test/abc) my app x-my app-y\n";
        let ctx = create_context(content);

        let warnings = rule.check(&ctx).unwrap();
        assert_eq!(warnings.len(), 1);
        assert_eq!(warnings[0].message, "Proper name 'my app' should be 'My App'");
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "`abc` [label](https://example.test/abc) My App x-my app-y\n");
        assert_eq!(rule.fix(&create_context(&fixed)).unwrap(), fixed);
    }

    #[test]
    fn first_configured_spelling_wins_for_colliding_variants() {
        for (names, content, expected) in [
            (["Node.js", "Nodejs"], "nodejs", "Node.js"),
            (["Nodejs", "Node.js"], "nodejs", "Nodejs"),
            (["Café", "Cafe"], "cafe", "Café"),
            (["Cafe", "Café"], "cafe", "Cafe"),
            (["Node.js", "NODE.JS"], "node.js", "Node.js"),
        ] {
            let rule = MD044ProperNames::new(names.map(str::to_string).to_vec(), false);
            let ctx = create_context(content);
            let warnings = rule.check(&ctx).unwrap();
            assert_eq!(warnings.len(), 1);
            assert_eq!(warnings[0].fix.as_ref().unwrap().replacement, expected);
            assert_eq!(rule.fix(&ctx).unwrap(), expected);
            assert_eq!(rule.clone().check(&ctx).unwrap(), warnings);
        }
    }

    #[test]
    fn test_field_map_nested_lines_inherit_top_level_key() {
        let map = field_map_for("---\nseo:\n  canonical: docs/a.md\n  keywords:\n    - myapp\ntitle: x\n---\n");
        assert_eq!(map[2].as_deref(), Some("seo"));
        assert_eq!(map[4].as_deref(), Some("seo"));
        assert_eq!(map[5].as_deref(), Some("title"));
    }

    #[test]
    fn test_field_map_block_scalar_bracket_does_not_swallow_next_key() {
        let map = field_map_for("---\ndescription: |\n  [myapp\ntitle: myapp\n---\n");
        assert_eq!(map[2].as_deref(), Some("description"));
        assert_eq!(
            map[3].as_deref(),
            Some("title"),
            "an indent-0 key always starts a new key"
        );
    }

    #[test]
    fn test_field_map_quoted_key_with_colon() {
        let map = field_map_for("---\n\"og:title\": myapp\n---\n");
        assert_eq!(map[1].as_deref(), Some("og:title"));
    }

    #[test]
    fn test_field_map_top_level_sequence_clears_attribution() {
        let map = field_map_for("---\n- myapp\n---\n");
        assert_eq!(map[1], None);
    }

    #[test]
    fn test_field_map_toml_table_body_belongs_to_table_root() {
        let map = field_map_for("+++\n[seo]\ncanonical = \"docs/a.md\"\n\n[[authors]]\nname = \"myapp\"\n+++\n");
        assert_eq!(map[2].as_deref(), Some("seo"));
        assert_eq!(map[5].as_deref(), Some("authors"));
    }

    #[test]
    fn test_field_map_toml_dotted_assignment_uses_root() {
        let map = field_map_for("+++\nseo.canonical = \"docs/a.md\"\n+++\n");
        assert_eq!(map[1].as_deref(), Some("seo"));
    }

    #[test]
    fn test_field_map_toml_array_continuation_inherits() {
        let map = field_map_for("+++\nseo = [\n\"docs/guide/myapp\"\n]\n+++\n");
        assert_eq!(map[2].as_deref(), Some("seo"));
    }

    #[test]
    fn test_field_map_indent_zero_flow_continuation_is_not_attributed_to_parent() {
        // Documented limitation: this is the safe direction. The value keeps
        // being checked, exactly as it is today.
        let map = field_map_for("---\nseo: [\n{name: myapp}\n]\n---\n");
        assert_eq!(map[2].as_deref(), Some("{name"));
    }

    #[test]
    fn test_field_map_toml_nested_array_inherits_and_title_not_corrupted() {
        let map = field_map_for("+++\nmatrix = [\n  [1, 2],\n  [3, 4],\n]\ntitle = \"x\"\n+++\n");
        assert_eq!(map[2].as_deref(), Some("matrix"), "nested array line inherits matrix");
        assert_eq!(map[3].as_deref(), Some("matrix"), "nested array line inherits matrix");
        assert_eq!(
            map[5].as_deref(),
            Some("title"),
            "title must not inherit stale attribution from a closed nested array"
        );
    }

    #[test]
    fn test_field_map_toml_nested_array_last_element_without_trailing_comma() {
        let map = field_map_for("+++\nmatrix = [\n  [1, 2],\n  [2]\n]\ntitle = \"x\"\n+++\n");
        assert_eq!(
            map[3].as_deref(),
            Some("matrix"),
            "last element without a trailing comma still inherits matrix"
        );
        assert_eq!(
            map[5].as_deref(),
            Some("title"),
            "title must not inherit stale attribution from a closed nested array"
        );
    }

    #[test]
    fn test_field_map_toml_nested_array_then_real_table_header_non_regression_guard() {
        // Not a bug reproduction: `toml_table_header` runs unconditionally
        // on every non-continuation line, so a real `[table]` header always
        // resyncs `current`/`in_toml_table` regardless of what came before.
        // This passed before the array-depth fix and always will; it guards
        // against a future change accidentally gating the header check
        // itself behind array-depth state.
        let map = field_map_for("+++\nmatrix = [\n  [1, 2],\n  [3, 4],\n]\n\n[seo]\ncanonical = \"docs/a.md\"\n+++\n");
        assert_eq!(map[6].as_deref(), Some("seo"), "real table header after a closed array");
        assert_eq!(
            map[7].as_deref(),
            Some("seo"),
            "table body still attributes to the table"
        );
    }

    #[test]
    fn test_field_map_toml_unclosed_array_resyncs_on_next_assignment() {
        // A forgotten closing bracket is a plausible authoring typo. Without
        // recovery, the stuck array depth would attribute `title` to
        // `matrix` forever, and excluding `matrix` would silently suppress
        // a real violation on `title`. The indent-0 assignment must resync
        // regardless of the unclosed depth.
        let map = field_map_for("+++\nmatrix = [\n  [1, 2],\ntitle = \"x\"\n+++\n");
        assert_eq!(
            map[3].as_deref(),
            Some("title"),
            "title must resync even though the array was never closed"
        );
    }

    #[test]
    fn test_field_map_toml_column_zero_array_elements_inherit_and_title_not_corrupted() {
        // TOML does not require indentation inside a multi-line array, so
        // `[1, 2],` at column 0 is a valid array element, not a table
        // header. `toml_table_header` must reject it: it starts with `[`
        // and contains a later `]`, but does not END with `]` (there is a
        // trailing `,`) and its inner content has an unquoted comma.
        let map = field_map_for("+++\nmatrix = [\n[1, 2],\n[3, 4],\n]\ntitle = \"x\"\n+++\n");
        assert_eq!(
            map[2].as_deref(),
            Some("matrix"),
            "column-0 array element inherits matrix"
        );
        assert_eq!(
            map[3].as_deref(),
            Some("matrix"),
            "column-0 array element inherits matrix"
        );
        assert_eq!(
            map[5].as_deref(),
            Some("title"),
            "title must not inherit stale attribution from a misread array element"
        );
    }

    #[test]
    fn test_field_map_toml_column_zero_array_last_element_without_trailing_comma() {
        let map = field_map_for("+++\nmatrix = [\n[1, 2],\n[2]\n]\ntitle = \"x\"\n+++\n");
        assert_eq!(
            map[3].as_deref(),
            Some("matrix"),
            "column-0 last element without a trailing comma still inherits matrix"
        );
        assert_eq!(
            map[5].as_deref(),
            Some("title"),
            "title must not inherit stale attribution from a misread array element"
        );
    }

    #[test]
    fn test_correctly_capitalized_names() {
        let rule = MD044ProperNames::new(
            vec![
                "JavaScript".to_string(),
                "TypeScript".to_string(),
                "Node.js".to_string(),
            ],
            true,
        );

        let content = "This document uses JavaScript, TypeScript, and Node.js correctly.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(result.is_empty(), "Should not flag correctly capitalized names");
    }

    #[test]
    fn test_incorrectly_capitalized_names() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string(), "TypeScript".to_string()], true);

        let content = "This document uses javascript and typescript incorrectly.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should flag two incorrect capitalizations");
        assert_eq!(result[0].message, "Proper name 'javascript' should be 'JavaScript'");
        assert_eq!(result[0].line, 1);
        assert_eq!(result[0].column, 20);
        assert_eq!(result[1].message, "Proper name 'typescript' should be 'TypeScript'");
        assert_eq!(result[1].line, 1);
        assert_eq!(result[1].column, 35);
    }

    #[test]
    fn test_names_at_beginning_of_sentences() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string(), "Python".to_string()], true);

        let content = "javascript is a great language. python is also popular.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should flag names at beginning of sentences");
        assert_eq!(result[0].line, 1);
        assert_eq!(result[0].column, 1);
        assert_eq!(result[1].line, 1);
        assert_eq!(result[1].column, 33);
    }

    #[test]
    fn test_names_in_code_blocks_checked_by_default() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        let content = r#"Here is some text with JavaScript.

```javascript
// This javascript should be checked
const lang = "javascript";
```

But this javascript should be flagged."#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 3, "Should flag javascript inside and outside code blocks");
        assert_eq!(result[0].line, 4);
        assert_eq!(result[1].line, 5);
        assert_eq!(result[2].line, 8);
    }

    #[test]
    fn test_names_in_code_blocks_ignored_when_disabled() {
        let rule = MD044ProperNames::new(
            vec!["JavaScript".to_string()],
            false, // code_blocks = false means skip code blocks
        );

        let content = r#"```
javascript in code block
```"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            0,
            "Should not flag javascript in code blocks when code_blocks is false"
        );
    }

    #[test]
    fn test_names_in_inline_code_checked_by_default() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        let content = "This is `javascript` in inline code and javascript outside.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // When code_blocks=true, inline code should be checked
        assert_eq!(result.len(), 2, "Should flag javascript inside and outside inline code");
        assert_eq!(result[0].column, 10); // javascript in inline code
        assert_eq!(result[1].column, 41); // javascript outside
    }

    #[test]
    fn test_multiple_names_in_same_line() {
        let rule = MD044ProperNames::new(
            vec!["JavaScript".to_string(), "TypeScript".to_string(), "React".to_string()],
            true,
        );

        let content = "I use javascript, typescript, and react in my projects.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 3, "Should flag all three incorrect names");
        assert_eq!(result[0].message, "Proper name 'javascript' should be 'JavaScript'");
        assert_eq!(result[1].message, "Proper name 'typescript' should be 'TypeScript'");
        assert_eq!(result[2].message, "Proper name 'react' should be 'React'");
    }

    #[test]
    fn test_case_sensitivity() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        let content = "JAVASCRIPT, Javascript, javascript, and JavaScript variations.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 3, "Should flag all incorrect case variations");
        // JavaScript (correct) should not be flagged
        assert!(result.iter().all(|w| w.message.contains("should be 'JavaScript'")));
    }

    #[test]
    fn test_configuration_with_custom_name_list() {
        let config = MD044Config {
            names: vec!["GitHub".to_string(), "GitLab".to_string(), "DevOps".to_string()],
            code_blocks: true,
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "We use github, gitlab, and devops for our workflow.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 3, "Should flag all custom names");
        assert_eq!(result[0].message, "Proper name 'github' should be 'GitHub'");
        assert_eq!(result[1].message, "Proper name 'gitlab' should be 'GitLab'");
        assert_eq!(result[2].message, "Proper name 'devops' should be 'DevOps'");
    }

    #[test]
    fn test_empty_configuration() {
        let rule = MD044ProperNames::new(vec![], true);

        let content = "This has javascript and typescript but no configured names.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(result.is_empty(), "Should not flag anything with empty configuration");
    }

    #[test]
    fn test_names_with_special_characters() {
        let rule = MD044ProperNames::new(
            vec!["Node.js".to_string(), "ASP.NET".to_string(), "C++".to_string()],
            true,
        );

        let content = "We use nodejs, asp.net, ASP.NET, and c++ in our stack.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // nodejs should match Node.js (dotless variation)
        // asp.net should be flagged (wrong case)
        // ASP.NET should not be flagged (correct)
        // c++ should be flagged
        assert_eq!(result.len(), 3, "Should handle special characters correctly");

        let messages: Vec<&str> = result.iter().map(|w| w.message.as_str()).collect();
        assert!(messages.contains(&"Proper name 'nodejs' should be 'Node.js'"));
        assert!(messages.contains(&"Proper name 'asp.net' should be 'ASP.NET'"));
        assert!(messages.contains(&"Proper name 'c++' should be 'C++'"));
    }

    #[test]
    fn test_word_boundaries() {
        let rule = MD044ProperNames::new(vec!["Java".to_string(), "Script".to_string()], true);

        let content = "JavaScript is not java or script, but Java and Script are separate.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Should only flag lowercase "java" and "script" as separate words
        assert_eq!(result.len(), 2, "Should respect word boundaries");
        assert!(result.iter().any(|w| w.column == 19)); // "java" position
        assert!(result.iter().any(|w| w.column == 27)); // "script" position
    }

    #[test]
    fn test_fix_method() {
        let rule = MD044ProperNames::new(
            vec![
                "JavaScript".to_string(),
                "TypeScript".to_string(),
                "Node.js".to_string(),
            ],
            true,
        );

        let content = "I love javascript, typescript, and nodejs!";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(fixed, "I love JavaScript, TypeScript, and Node.js!");
    }

    #[test]
    fn test_fix_multiple_occurrences() {
        let rule = MD044ProperNames::new(vec!["Python".to_string()], true);

        let content = "python is great. I use python daily. PYTHON is powerful.";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(fixed, "Python is great. I use Python daily. Python is powerful.");
    }

    #[test]
    fn test_fix_checks_code_blocks_by_default() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        let content = r#"I love javascript.

```
const lang = "javascript";
```

More javascript here."#;

        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        let expected = r#"I love JavaScript.

```
const lang = "JavaScript";
```

More JavaScript here."#;

        assert_eq!(fixed, expected);
    }

    #[test]
    fn test_multiline_content() {
        let rule = MD044ProperNames::new(vec!["Rust".to_string(), "Python".to_string()], true);

        let content = r#"First line with rust.
Second line with python.
Third line with RUST and PYTHON."#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 4, "Should flag all incorrect occurrences");
        assert_eq!(result[0].line, 1);
        assert_eq!(result[1].line, 2);
        assert_eq!(result[2].line, 3);
        assert_eq!(result[3].line, 3);
    }

    #[test]
    fn test_default_config() {
        let config = MD044Config::default();
        assert!(config.names.is_empty());
        assert!(!config.code_blocks);
        assert!(config.html_elements);
        assert!(config.html_comments);
    }

    #[test]
    fn test_default_config_checks_html_comments() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "# Guide\n\n<!-- javascript mentioned here -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Default config should check HTML comments");
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_default_config_skips_code_blocks() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "# Guide\n\n```\njavascript in code\n```\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 0, "Default config should skip code blocks");
    }

    #[test]
    fn test_standalone_html_comment_checked() {
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "# Heading\n\n<!-- this is a test example -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should flag proper name in standalone HTML comment");
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_inline_config_comments_not_flagged() {
        let config = MD044Config {
            names: vec!["RUMDL".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        // Lines 1, 3, 4, 6 are inline config comments — should not be flagged.
        // Lines 2, 5 contain "rumdl" in regular text — flagged by rule.check(),
        // but would be suppressed by the linting engine's inline config filtering.
        let content = "<!-- rumdl-disable MD044 -->\nSome rumdl text here.\n<!-- rumdl-enable MD044 -->\n<!-- markdownlint-disable -->\nMore rumdl text.\n<!-- markdownlint-enable -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should only flag body lines, not config comments");
        assert_eq!(result[0].line, 2);
        assert_eq!(result[1].line, 5);
    }

    #[test]
    fn test_compact_inline_config_comments_not_flagged() {
        let config = MD044Config {
            names: vec!["RUMDL".to_string(), "Markdownlint".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        // Compact directive lines are skipped like the spaced ones; the two body
        // lines are flagged, and so is a comment that is not a directive.
        let content = "<!--rumdl-disable MD044-->\nSome rumdl text.\n<!--markdownlint-enable MD044 -->\n<!--\trumdl-capture-->\nMore rumdl text.\n<!-- note markdownlint -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        let lines: Vec<usize> = result.iter().map(|w| w.line).collect();
        assert_eq!(lines, vec![2, 5, 6], "{result:?}");
    }

    #[test]
    fn test_html_comment_skipped_when_disabled() {
        let config = MD044Config {
            names: vec!["Test".to_string()],
            code_blocks: true,
            html_comments: false,
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "# Heading\n\n<!-- this is a test example -->\n\nRegular test here.\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            1,
            "Should only flag 'test' outside HTML comment when html_comments=false"
        );
        assert_eq!(result[0].line, 5);
    }

    #[test]
    fn test_fix_corrects_html_comment_content() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "# Guide\n\n<!-- javascript mentioned here -->\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(fixed, "# Guide\n\n<!-- JavaScript mentioned here -->\n");
    }

    #[test]
    fn test_fix_does_not_modify_inline_config_comments() {
        let config = MD044Config {
            names: vec!["RUMDL".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- rumdl-disable -->\nSome rumdl text.\n<!-- rumdl-enable -->\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // Config comments should be untouched
        assert!(fixed.contains("<!-- rumdl-disable -->"));
        assert!(fixed.contains("<!-- rumdl-enable -->"));
        // Body text inside disable block should NOT be fixed (rule is disabled)
        assert!(
            fixed.contains("Some rumdl text."),
            "Line inside rumdl-disable block should not be modified by fix()"
        );
    }

    #[test]
    fn test_fix_respects_inline_disable_partial() {
        let config = MD044Config {
            names: vec!["RUMDL".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content =
            "<!-- rumdl-disable MD044 -->\nSome rumdl text.\n<!-- rumdl-enable MD044 -->\n\nSome rumdl text outside.\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // Line inside disable block should be preserved
        assert!(
            fixed.contains("Some rumdl text.\n<!-- rumdl-enable"),
            "Line inside disable block should not be modified"
        );
        // Line outside disable block should be fixed
        assert!(
            fixed.contains("Some RUMDL text outside."),
            "Line outside disable block should be fixed"
        );
    }

    #[test]
    fn test_performance_with_many_names() {
        let mut names = vec![];
        for i in 0..50 {
            names.push(format!("ProperName{i}"));
        }

        let rule = MD044ProperNames::new(names, true);

        let content = "This has propername0, propername25, and propername49 incorrectly.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 3, "Should handle many configured names efficiently");
    }

    #[test]
    fn test_large_name_count_performance() {
        // Verify MD044 can handle large numbers of names without regex limitations
        // This test confirms that fancy-regex handles large patterns well
        let names = (0..1000).map(|i| format!("ProperName{i}")).collect::<Vec<_>>();

        let rule = MD044ProperNames::new(names, true);

        // The combined pattern should be created successfully
        assert!(rule.combined_pattern.is_some());

        // Should be able to check content without errors
        let content = "This has propername0 and propername999 in it.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Should detect both incorrect names
        assert_eq!(result.len(), 2, "Should handle 1000 names without issues");
    }

    #[test]
    fn test_cache_behavior() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        let content = "Using javascript here.";
        let ctx = create_context(content);

        // First check
        let result1 = rule.check(&ctx).unwrap();
        assert_eq!(result1.len(), 1);

        // Second check should use cache
        let result2 = rule.check(&ctx).unwrap();
        assert_eq!(result2.len(), 1);

        // Results should be identical
        assert_eq!(result1[0].line, result2[0].line);
        assert_eq!(result1[0].column, result2[0].column);
    }

    #[test]
    fn test_html_comments_not_checked_when_disabled() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            code_blocks: true,    // Check code blocks
            html_comments: false, // Don't check HTML comments
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = r#"Regular javascript here.
<!-- This javascript in HTML comment should be ignored -->
More javascript outside."#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should only flag javascript outside HTML comments");
        assert_eq!(result[0].line, 1);
        assert_eq!(result[1].line, 3);
    }

    #[test]
    fn test_html_comments_checked_when_enabled() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            code_blocks: true, // Check code blocks
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = r#"Regular javascript here.
<!-- This javascript in HTML comment should be checked -->
More javascript outside."#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            3,
            "Should flag all javascript occurrences including in HTML comments"
        );
    }

    #[test]
    fn test_indented_html_comment_escapes_via_link_and_backticks() {
        // Regression for #755: MD044 checks inside HTML comments by default, but
        // links and inline code inside a comment should escape the rule. That
        // protection depends on the line being recognised as an HTML comment,
        // which must hold whether or not the comment is indented.
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- see the [relevant page](test.md). -->\n<!-- see `test.md` -->\n  <!-- see the [relevant page](test.md). -->\n  <!-- see `test.md` -->\n";

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "'test' inside a link URL or backticks must be ignored in both column-0 and indented comments, got: {result:?}"
        );
    }

    #[test]
    fn test_indented_html_comment_still_checks_bare_prose() {
        // The indent fix must not suppress genuine violations: bare prose inside an
        // indented comment is still checked (only links/backticks escape).
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "  <!-- this is a test comment -->\n";

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            1,
            "bare 'test' in an indented comment is still a violation"
        );
        assert_eq!(result[0].line, 1);
    }

    #[test]
    fn test_multiline_html_comments() {
        let config = MD044Config {
            names: vec!["Python".to_string(), "JavaScript".to_string()],
            code_blocks: true,    // Check code blocks
            html_comments: false, // Don't check HTML comments
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = r#"Regular python here.
<!--
This is a multiline comment
with javascript and python
that should be ignored
-->
More javascript outside."#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should only flag names outside HTML comments");
        assert_eq!(result[0].line, 1); // python
        assert_eq!(result[1].line, 7); // javascript
    }

    #[test]
    fn test_fix_preserves_html_comments_when_disabled() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            code_blocks: true,    // Check code blocks
            html_comments: false, // Don't check HTML comments
            ..Default::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = r#"javascript here.
<!-- javascript in comment -->
More javascript."#;

        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        let expected = r#"JavaScript here.
<!-- javascript in comment -->
More JavaScript."#;

        assert_eq!(
            fixed, expected,
            "Should not fix names inside HTML comments when disabled"
        );
    }

    #[test]
    fn test_proper_names_in_link_text_are_flagged() {
        let rule = MD044ProperNames::new(
            vec!["JavaScript".to_string(), "Node.js".to_string(), "Python".to_string()],
            true,
        );

        let content = r#"Check this [javascript documentation](https://javascript.info) for info.

Visit [node.js homepage](https://nodejs.org) and [python tutorial](https://python.org).

Real javascript should be flagged.

Also see the [typescript guide][ts-ref] for more.

Real python should be flagged too.

[ts-ref]: https://typescript.org/handbook"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Link text should be checked, URLs should not be checked
        // Line 1: [javascript documentation] - "javascript" should be flagged
        // Line 3: [node.js homepage] - "node.js" should be flagged (matches "Node.js")
        // Line 3: [python tutorial] - "python" should be flagged
        // Line 5: standalone javascript
        // Line 9: standalone python
        assert_eq!(result.len(), 5, "Expected 5 warnings: 3 in link text + 2 standalone");

        // Verify line numbers for link text warnings
        let line_1_warnings: Vec<_> = result.iter().filter(|w| w.line == 1).collect();
        assert_eq!(line_1_warnings.len(), 1);
        assert!(
            line_1_warnings[0]
                .message
                .contains("'javascript' should be 'JavaScript'")
        );

        let line_3_warnings: Vec<_> = result.iter().filter(|w| w.line == 3).collect();
        assert_eq!(line_3_warnings.len(), 2); // node.js and python

        // Standalone warnings
        assert!(result.iter().any(|w| w.line == 5 && w.message.contains("'javascript'")));
        assert!(result.iter().any(|w| w.line == 9 && w.message.contains("'python'")));
    }

    #[test]
    fn test_link_urls_not_flagged() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        // URL contains "javascript" but should NOT be flagged
        let content = r#"[Link Text](https://javascript.info/guide)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // URL should not be checked
        assert!(result.is_empty(), "URLs should not be checked for proper names");
    }

    #[test]
    fn test_bare_urls_not_flagged() {
        let rule = MD044ProperNames::new(vec!["Foo".to_string(), "JavaScript".to_string()], true);

        // Bare URLs are not links in ctx.links(), but a proper-name "fix"
        // inside a domain or case-sensitive path would break the link.
        let content =
            "https://foo.com\n\nSee https://javascript.info/foo/guide for details.\n\nMail foo@foo.com about it.\n";

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Bare URLs and emails should not be checked for proper names: {result:?}"
        );
    }

    #[test]
    fn test_prose_around_bare_url_still_flagged() {
        let rule = MD044ProperNames::new(vec!["Foo".to_string()], true);

        // The word before and after the URL must still be flagged; only the
        // URL bytes themselves are exempt.
        let content = "Use foo at https://foo.com because foo is great.\n";

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            2,
            "Prose occurrences around a bare URL must still be flagged: {result:?}"
        );
        assert!(result.iter().all(|w| w.message.contains("'foo' should be 'Foo'")));
    }

    #[test]
    fn test_proper_names_in_image_alt_text_are_flagged() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        let content = r#"Here is a ![javascript logo](javascript.png "javascript icon") image.

Real javascript should be flagged."#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Image alt text should be checked, URL and title should not be checked
        // Line 1: ![javascript logo] - "javascript" should be flagged
        // Line 3: standalone javascript
        assert_eq!(result.len(), 2, "Expected 2 warnings: 1 in alt text + 1 standalone");
        assert!(result[0].message.contains("'javascript' should be 'JavaScript'"));
        assert!(result[0].line == 1); // "![javascript logo]"
        assert!(result[1].message.contains("'javascript' should be 'JavaScript'"));
        assert!(result[1].line == 3); // "Real javascript should be flagged."
    }

    #[test]
    fn test_image_urls_not_flagged() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        // URL contains "javascript" but should NOT be flagged
        let content = r#"![Logo](https://javascript.info/logo.png)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Image URL should not be checked
        assert!(result.is_empty(), "Image URLs should not be checked for proper names");
    }

    #[test]
    fn test_reference_link_text_flagged_but_definition_not() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string(), "TypeScript".to_string()], true);

        let content = r#"Check the [javascript guide][js-ref] for details.

Real javascript should be flagged.

[js-ref]: https://javascript.info/typescript/guide"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Link text should be checked, reference definitions should not
        // Line 1: [javascript guide] - should be flagged
        // Line 3: standalone javascript - should be flagged
        // Line 5: reference definition - should NOT be flagged
        assert_eq!(result.len(), 2, "Expected 2 warnings: 1 in link text + 1 standalone");
        assert!(result.iter().any(|w| w.line == 1 && w.message.contains("'javascript'")));
        assert!(result.iter().any(|w| w.line == 3 && w.message.contains("'javascript'")));
    }

    #[test]
    fn test_reference_definitions_not_flagged() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        // Reference definition should NOT be flagged
        let content = r#"[js-ref]: https://javascript.info/guide"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Reference definition URLs should not be checked
        assert!(result.is_empty(), "Reference definitions should not be checked");
    }

    #[test]
    fn test_wikilinks_text_is_flagged() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string()], true);

        // WikiLinks [[destination]] should have their text checked
        let content = r#"[[javascript]]

Regular javascript here.

[[JavaScript|display text]]"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Line 1: [[javascript]] - should be flagged (WikiLink text)
        // Line 3: standalone javascript - should be flagged
        // Line 5: [[JavaScript|display text]] - correct capitalization, no flag
        assert_eq!(result.len(), 2, "Expected 2 warnings: 1 in WikiLink + 1 standalone");
        assert!(
            result
                .iter()
                .any(|w| w.line == 1 && w.column == 3 && w.message.contains("'javascript'"))
        );
        assert!(result.iter().any(|w| w.line == 3 && w.message.contains("'javascript'")));
    }

    #[test]
    fn test_url_link_text_not_flagged() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);

        // Link text that is itself a URL should not be flagged
        let content = r#"[https://github.com/org/repo](https://github.com/org/repo)

[http://github.com/org/repo](http://github.com/org/repo)

[www.github.com/org/repo](https://www.github.com/org/repo)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "URL-like link text should not be flagged, got: {result:?}"
        );
    }

    #[test]
    fn test_url_link_text_with_leading_space_not_flagged() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);

        // Leading/trailing whitespace in link text should be trimmed before URL check
        let content = r#"[ https://github.com/org/repo](https://github.com/org/repo)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "URL-like link text with leading space should not be flagged, got: {result:?}"
        );
    }

    #[test]
    fn test_url_link_text_uppercase_scheme_not_flagged() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);

        let content = r#"[HTTPS://GITHUB.COM/org/repo](https://github.com/org/repo)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "URL-like link text with uppercase scheme should not be flagged, got: {result:?}"
        );
    }

    #[test]
    fn test_non_url_link_text_still_flagged() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);

        // Only prose descriptions in link text should be flagged.
        // Bare-domain, protocol-relative, and scheme-prefixed link texts that
        // match the destination URL are all URLs and must not be corrected.
        let content = r#"[github.com/org/repo](https://github.com/org/repo)

[Visit github](https://github.com/org/repo)

[//github.com/org/repo](//github.com/org/repo)

[ftp://github.com/org/repo](ftp://github.com/org/repo)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Line 1: bare-domain text matches destination — not flagged
        // Line 3: prose description — flagged
        // Line 5: protocol-relative URL text — not flagged
        // Line 7: ftp:// URL text matches destination — not flagged
        assert_eq!(
            result.len(),
            1,
            "Only prose link text should be flagged, got: {result:?}"
        );
        assert!(
            result.iter().any(|w| w.line == 3),
            "Expected 'Visit github' on line 3 to be flagged"
        );
    }

    #[test]
    fn test_url_link_text_fix_not_applied() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);

        let content = "[https://github.com/org/repo](https://github.com/org/repo)\n";

        let ctx = create_context(content);
        let result = rule.fix(&ctx).unwrap();

        assert_eq!(result, content, "Fix should not modify URL-like link text");
    }

    #[test]
    fn test_mixed_url_and_regular_link_text() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);

        // Mix of URL link text (should skip) and regular text (should flag)
        let content = r#"[https://github.com/org/repo](https://github.com/org/repo)

Visit [github documentation](https://github.com/docs) for details.

[www.github.com/pricing](https://www.github.com/pricing)"#;

        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Only line 3 should be flagged ("github documentation" is not a URL)
        assert_eq!(
            result.len(),
            1,
            "Only non-URL link text should be flagged, got: {result:?}"
        );
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_html_attribute_values_not_flagged() {
        // Matches inside HTML tag attributes (between `<` and `>`) are not flagged.
        // Attribute values are not prose — they hold URLs, class names, data values, etc.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);
        let content = "# Heading\n\ntest\n\n<img src=\"www.example.test/test_image.png\">\n";
        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Nothing on line 5 should be flagged — everything is inside the `<img ...>` tag
        let line5_violations: Vec<_> = result.iter().filter(|w| w.line == 5).collect();
        assert!(
            line5_violations.is_empty(),
            "Should not flag anything inside HTML tag attributes: {line5_violations:?}"
        );

        // Plain text on line 3 is still flagged
        let line3_violations: Vec<_> = result.iter().filter(|w| w.line == 3).collect();
        assert_eq!(line3_violations.len(), 1, "Plain 'test' on line 3 should be flagged");
    }

    #[test]
    fn test_html_text_content_still_flagged() {
        // Text between HTML tags (not inside `<...>`) is still checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);
        let content = "# Heading\n\n<a href=\"https://example.test/page\">test link</a>\n";
        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // "example.test" in the href attribute → not flagged (inside `<...>`)
        // "test link" in the anchor text → flagged (between `>` and `<`)
        assert_eq!(
            result.len(),
            1,
            "Should flag only 'test' in anchor text, not in href: {result:?}"
        );
        assert_eq!(result[0].column, 37, "Should flag col 37 ('test link' in anchor text)");
    }

    #[test]
    fn test_html_attribute_various_not_flagged() {
        // All attribute types are ignored: src, href, alt, class, data-*, title, etc.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);
        let content = concat!(
            "# Heading\n\n",
            "<img src=\"test.png\" alt=\"test image\">\n",
            "<span class=\"test-class\" data-test=\"value\">test content</span>\n",
        );
        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Only "test content" (between tags on line 4) should be flagged
        assert_eq!(
            result.len(),
            1,
            "Should flag only 'test content' between tags: {result:?}"
        );
        assert_eq!(result[0].line, 4);
    }

    #[test]
    fn test_plain_text_underscore_boundary_unchanged() {
        // Plain text (outside HTML tags) still uses original word boundary semantics where
        // underscore is a boundary character, matching markdownlint's behavior via AST splitting.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);
        let content = "# Heading\n\ntest_image is here and just_test ends here\n";
        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Both "test_image" (test at start) and "just_test" (test at end) are flagged
        // because in plain text, "_" is a word boundary
        assert_eq!(
            result.len(),
            2,
            "Should flag 'test' in both 'test_image' and 'just_test': {result:?}"
        );
        let cols: Vec<usize> = result.iter().map(|w| w.column).collect();
        assert!(cols.contains(&1), "Should flag col 1 (test_image): {cols:?}");
        assert!(cols.contains(&29), "Should flag col 29 (just_test): {cols:?}");
    }

    #[test]
    fn test_frontmatter_yaml_keys_not_flagged() {
        // YAML keys in frontmatter should NOT be checked for proper name violations.
        // Only values should be checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntitle: Heading\ntest: Some Test value\n---\n\nTest\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" in the YAML key (line 3) should NOT be flagged
        // "Test" in the YAML value (line 3) is correct capitalization, no flag
        // "Test" in body (line 6) is correct capitalization, no flag
        assert!(
            result.is_empty(),
            "Should not flag YAML keys or correctly capitalized values: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_yaml_values_flagged() {
        // Incorrectly capitalized names in YAML values should be flagged.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntitle: Heading\nkey: a test value\n---\n\nTest\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" in the YAML value (line 3) SHOULD be flagged
        assert_eq!(result.len(), 1, "Should flag 'test' in YAML value: {result:?}");
        assert_eq!(result[0].line, 3);
        assert_eq!(result[0].column, 8); // "key: a " = 7 chars, then "test" at column 8
    }

    #[test]
    fn test_frontmatter_key_matches_name_not_flagged() {
        // A YAML key that happens to match a configured name should NOT be flagged.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntest: other value\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag YAML key that matches configured name: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_empty_value_not_flagged() {
        // YAML key with no value should be skipped entirely.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntest:\ntest: \n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag YAML keys with empty values: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_nested_yaml_key_not_flagged() {
        // Nested/indented YAML keys should also be skipped.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\nparent:\n  test: nested value\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" as a nested key should NOT be flagged
        assert!(result.is_empty(), "Should not flag nested YAML keys: {result:?}");
    }

    #[test]
    fn test_frontmatter_list_items_checked() {
        // YAML list items are values and should be checked for proper names.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntags:\n  - test\n  - other\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" as a list item value SHOULD be flagged
        assert_eq!(result.len(), 1, "Should flag 'test' in YAML list item: {result:?}");
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_frontmatter_value_with_multiple_colons() {
        // For "key: value: more", key is before first colon.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntest: description: a test thing\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" as key should NOT be flagged
        // "test" in value portion ("description: a test thing") SHOULD be flagged
        assert_eq!(
            result.len(),
            1,
            "Should flag 'test' in value after first colon: {result:?}"
        );
        assert_eq!(result[0].line, 2);
        assert!(result[0].column > 6, "Violation column should be in value portion");
    }

    #[test]
    fn test_frontmatter_does_not_affect_body() {
        // Body text after frontmatter should still be fully checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntitle: Heading\n---\n\ntest should be flagged here\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should flag 'test' in body text: {result:?}");
        assert_eq!(result[0].line, 5);
    }

    #[test]
    fn test_frontmatter_fix_corrects_values_preserves_keys() {
        // Fix should correct YAML values but preserve keys.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntest: a test value\n---\n\ntest here\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // Key "test" should remain lowercase; value "test" should become "Test"
        assert_eq!(fixed, "---\ntest: a Test value\n---\n\nTest here\n");
    }

    #[test]
    fn test_frontmatter_multiword_value_flagged() {
        // Multiple proper names in a single YAML value should all be flagged.
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string(), "TypeScript".to_string()], true);

        let content = "---\ndescription: Learn javascript and typescript\n---\n\nBody\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should flag both names in YAML value: {result:?}");
        assert!(result.iter().all(|w| w.line == 2));
    }

    #[test]
    fn test_frontmatter_yaml_comments_not_checked() {
        // YAML comments inside frontmatter should be skipped entirely.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\n# test comment\ntitle: Heading\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(result.is_empty(), "Should not flag names in YAML comments: {result:?}");
    }

    #[test]
    fn test_frontmatter_delimiters_not_checked() {
        // Frontmatter delimiter lines (--- or +++) should never be checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntitle: Heading\n---\n\ntest here\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Only the body "test" on line 5 should be flagged
        assert_eq!(result.len(), 1, "Should only flag body text: {result:?}");
        assert_eq!(result[0].line, 5);
    }

    #[test]
    fn test_frontmatter_continuation_lines_checked() {
        // Continuation lines (indented, no colon) are value content and should be checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ndescription: >\n  a test value\n  continued here\n---\n\nBody\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" on the continuation line should be flagged
        assert_eq!(result.len(), 1, "Should flag 'test' in continuation line: {result:?}");
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_frontmatter_quoted_values_checked() {
        // Quoted YAML values should have their content checked (inside the quotes).
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntitle: \"a test title\"\n---\n\nBody\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should flag 'test' in quoted YAML value: {result:?}");
        assert_eq!(result[0].line, 2);
    }

    #[test]
    fn test_frontmatter_single_quoted_values_checked() {
        // Single-quoted YAML values should have their content checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntitle: 'a test title'\n---\n\nBody\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            1,
            "Should flag 'test' in single-quoted YAML value: {result:?}"
        );
        assert_eq!(result[0].line, 2);
    }

    #[test]
    fn test_frontmatter_fix_multiword_values() {
        // Fix should correct all proper names in frontmatter values.
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string(), "TypeScript".to_string()], true);

        let content = "---\ndescription: Learn javascript and typescript\n---\n\nBody\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(
            fixed,
            "---\ndescription: Learn JavaScript and TypeScript\n---\n\nBody\n"
        );
    }

    #[test]
    fn test_frontmatter_fix_preserves_yaml_structure() {
        // Fix should preserve YAML structure while correcting values.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntags:\n  - test\n  - other\ntitle: a test doc\n---\n\ntest body\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(
            fixed,
            "---\ntags:\n  - Test\n  - other\ntitle: a Test doc\n---\n\nTest body\n"
        );
    }

    #[test]
    fn test_frontmatter_toml_delimiters_not_checked() {
        // TOML frontmatter with +++ delimiters should also be handled.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "+++\ntitle = \"a test title\"\n+++\n\ntest body\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "title" as TOML key should NOT be flagged
        // "test" in TOML quoted value SHOULD be flagged (line 2)
        // "test" in body SHOULD be flagged (line 5)
        assert_eq!(result.len(), 2, "Should flag TOML value and body: {result:?}");
        let fm_violations: Vec<_> = result.iter().filter(|w| w.line == 2).collect();
        assert_eq!(fm_violations.len(), 1, "Should flag 'test' in TOML value: {result:?}");
        let body_violations: Vec<_> = result.iter().filter(|w| w.line == 5).collect();
        assert_eq!(body_violations.len(), 1, "Should flag body 'test': {result:?}");
    }

    #[test]
    fn test_frontmatter_toml_key_not_flagged() {
        // TOML keys should NOT be flagged, only values.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "+++\ntest = \"other value\"\n+++\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag TOML key that matches configured name: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_toml_fix_preserves_keys() {
        // Fix should correct TOML values but preserve keys.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "+++\ntest = \"a test value\"\n+++\n\ntest here\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // Key "test" should remain lowercase; value "test" should become "Test"
        assert_eq!(fixed, "+++\ntest = \"a Test value\"\n+++\n\nTest here\n");
    }

    #[test]
    fn test_frontmatter_list_item_mapping_key_not_flagged() {
        // In "- test: nested value", "test" is a YAML key within a list-item mapping.
        // The key should NOT be flagged; only the value should be checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\nitems:\n  - test: nested value\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag YAML key in list-item mapping: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_list_item_mapping_value_flagged() {
        // In "- key: test value", the value portion should be checked.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\nitems:\n  - key: a test value\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            1,
            "Should flag 'test' in list-item mapping value: {result:?}"
        );
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_frontmatter_bare_list_item_still_flagged() {
        // Bare list items without a colon (e.g., "- test") are values and should be flagged.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\ntags:\n  - test\n  - other\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should flag 'test' in bare list item: {result:?}");
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_frontmatter_flow_mapping_not_flagged() {
        // Flow mappings like {test: value} contain YAML keys that should not be flagged.
        // The entire flow construct should be skipped.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\nflow_map: {test: value, other: test}\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag names inside flow mappings: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_flow_sequence_not_flagged() {
        // Flow sequences like [test, other] should also be skipped.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\nitems: [test, other, test]\n---\n\nBody text\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag names inside flow sequences: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_list_item_mapping_fix_preserves_key() {
        // Fix should correct values in list-item mappings but preserve keys.
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "---\nitems:\n  - test: a test value\n---\n\ntest here\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // "test" as list-item key should remain lowercase;
        // "test" in value portion should become "Test"
        assert_eq!(fixed, "---\nitems:\n  - test: a Test value\n---\n\nTest here\n");
    }

    #[test]
    fn test_frontmatter_backtick_code_not_flagged() {
        // Names inside backticks in frontmatter should NOT be flagged when code_blocks=false.
        let config = MD044Config {
            names: vec!["GoodApplication".to_string()],
            code_blocks: false,
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "---\ntitle: \"`goodapplication` CLI\"\n---\n\nIntroductory `goodapplication` CLI text.\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Neither the frontmatter nor the body backtick-wrapped name should be flagged
        assert!(
            result.is_empty(),
            "Should not flag names inside backticks in frontmatter or body: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_unquoted_backtick_code_not_flagged() {
        // Exact case from issue #513: unquoted YAML frontmatter with backticks
        let config = MD044Config {
            names: vec!["GoodApplication".to_string()],
            code_blocks: false,
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "---\ntitle: `goodapplication` CLI\n---\n\nIntroductory `goodapplication` CLI text.\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag names inside backticks in unquoted YAML frontmatter: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_bare_name_still_flagged_with_backtick_nearby() {
        // Names outside backticks in frontmatter should still be flagged.
        let config = MD044Config {
            names: vec!["GoodApplication".to_string()],
            code_blocks: false,
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "---\ntitle: goodapplication `goodapplication` CLI\n---\n\nBody\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Only the bare "goodapplication" (before backticks) should be flagged
        assert_eq!(
            result.len(),
            1,
            "Should flag bare name but not backtick-wrapped name: {result:?}"
        );
        assert_eq!(result[0].line, 2);
        assert_eq!(result[0].column, 8); // "title: " = 7 chars, name at column 8
    }

    #[test]
    fn test_frontmatter_backtick_code_with_code_blocks_true() {
        // When code_blocks=true, names inside backticks ARE checked.
        let config = MD044Config {
            names: vec!["GoodApplication".to_string()],
            code_blocks: true,
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "---\ntitle: \"`goodapplication` CLI\"\n---\n\nBody\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // With code_blocks=true, backtick-wrapped name SHOULD be flagged
        assert_eq!(
            result.len(),
            1,
            "Should flag backtick-wrapped name when code_blocks=true: {result:?}"
        );
        assert_eq!(result[0].line, 2);
    }

    #[test]
    fn test_frontmatter_fix_preserves_backtick_code() {
        // Fix should NOT change names inside backticks in frontmatter.
        let config = MD044Config {
            names: vec!["GoodApplication".to_string()],
            code_blocks: false,
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "---\ntitle: \"`goodapplication` CLI\"\n---\n\nIntroductory `goodapplication` CLI text.\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // Neither backtick-wrapped occurrence should be changed
        assert_eq!(
            fixed, content,
            "Fix should not modify names inside backticks in frontmatter"
        );
    }

    fn rule_ignoring(names: &[&str], ignore: &[&str]) -> MD044ProperNames {
        MD044ProperNames::from_config_struct(MD044Config {
            names: names.iter().map(ToString::to_string).collect(),
            ignore_frontmatter_fields: Some(ignore.iter().map(ToString::to_string).collect()),
            ..Default::default()
        })
    }

    #[test]
    fn test_ignore_frontmatter_field_suppresses_only_that_field() {
        let content = "---\ntitle: Heading for myapp\nslug: myapp-guide\n---\n";
        let rule = rule_ignoring(&["MyApp"], &["slug"]);
        let result = rule.check(&create_context(content)).unwrap();
        assert_eq!(result.len(), 1, "only title is flagged: {result:?}");
        assert_eq!(result[0].line, 2);
    }

    #[test]
    fn test_ignore_frontmatter_field_is_case_insensitive() {
        let content = "---\nSlug: myapp-guide\n---\n";
        let rule = rule_ignoring(&["MyApp"], &["SLUG"]);
        assert!(rule.check(&create_context(content)).unwrap().is_empty());
    }

    #[test]
    fn test_ignore_frontmatter_field_covers_nested_subtree() {
        let content = "---\nseo:\n  canonical: myapp\n  keywords:\n    - myapp\n---\n";
        let rule = rule_ignoring(&["MyApp"], &["seo"]);
        assert!(rule.check(&create_context(content)).unwrap().is_empty());
    }

    #[test]
    fn test_ignore_frontmatter_field_does_not_affect_body() {
        let content = "---\nslug: myapp\n---\n\nBody mentions myapp.\n";
        let rule = rule_ignoring(&["MyApp"], &["slug"]);
        let result = rule.check(&create_context(content)).unwrap();
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].line, 5);
    }

    #[test]
    fn test_ignore_frontmatter_field_toml_table() {
        let content = "+++\n[seo]\ncanonical = \"myapp\"\n+++\n";
        let rule = rule_ignoring(&["MyApp"], &["seo"]);
        assert!(rule.check(&create_context(content)).unwrap().is_empty());
    }

    // --- Angle-bracket URL tests (issue #457) ---

    #[test]
    fn test_angle_bracket_url_in_html_comment_not_flagged() {
        // Angle-bracket URLs inside HTML comments should be skipped
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "---\ntitle: Level 1 heading\n---\n\n<https://www.example.test>\n\n<!-- This is a Test https://www.example.test -->\n<!-- This is a Test <https://www.example.test> -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Line 7: "Test" in comment prose before bare URL -- already correct capitalization
        // Line 7: "test" in bare URL (not in angle brackets) -- but "test" is in URL domain, not prose.
        //   However, .example.test has "test" at a word boundary (after '.'), so it IS flagged.
        // Line 8: "Test" in comment prose -- correct capitalization, not flagged
        // Line 8: "test" in <https://www.example.test> -- inside angle-bracket URL, NOT flagged

        // The key assertion: line 8's angle-bracket URL should NOT produce a warning
        let line8_warnings: Vec<_> = result.iter().filter(|w| w.line == 8).collect();
        assert!(
            line8_warnings.is_empty(),
            "Should not flag names inside angle-bracket URLs in HTML comments: {line8_warnings:?}"
        );
    }

    #[test]
    fn test_bare_url_in_html_comment_still_flagged() {
        // Bare URLs (not in angle brackets) inside HTML comments should still be checked
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- This is a test https://www.example.test -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // "test" appears as prose text before URL and also in the bare URL domain
        // At minimum, the prose "test" should be flagged
        assert!(
            !result.is_empty(),
            "Should flag 'test' in prose text of HTML comment with bare URL"
        );
    }

    #[test]
    fn test_angle_bracket_url_in_regular_markdown_not_flagged() {
        // Angle-bracket URLs in regular markdown are already handled by the link parser,
        // but the angle-bracket check provides a safety net
        let rule = MD044ProperNames::new(vec!["Test".to_string()], true);

        let content = "<https://www.example.test>\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag names inside angle-bracket URLs in regular markdown: {result:?}"
        );
    }

    #[test]
    fn test_multiple_angle_bracket_urls_in_one_comment() {
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- See <https://test.example.com> and <https://www.example.test> for details -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Both URLs are inside angle brackets, so "test" inside them should NOT be flagged
        assert!(
            result.is_empty(),
            "Should not flag names inside multiple angle-bracket URLs: {result:?}"
        );
    }

    #[test]
    fn test_angle_bracket_non_url_still_flagged() {
        // <Test> is NOT a URL (no scheme), so is_in_angle_bracket_url does NOT protect it.
        // Whether it gets flagged depends on HTML tag detection, not on our URL check.
        assert!(
            !MD044ProperNames::is_in_angle_bracket_url("<test> which is not a URL.", 1),
            "is_in_angle_bracket_url should return false for non-URL angle brackets"
        );
    }

    #[test]
    fn test_angle_bracket_mailto_url_not_flagged() {
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- Contact <mailto:test@example.com> for help -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag names inside angle-bracket mailto URLs: {result:?}"
        );
    }

    #[test]
    fn test_angle_bracket_ftp_url_not_flagged() {
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- Download from <ftp://test.example.com/file> -->\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert!(
            result.is_empty(),
            "Should not flag names inside angle-bracket FTP URLs: {result:?}"
        );
    }

    #[test]
    fn test_angle_bracket_url_fix_preserves_url() {
        // Fix should not modify text inside angle-bracket URLs
        let config = MD044Config {
            names: vec!["Test".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "<!-- test text <https://www.example.test> -->\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // "test" in prose should be fixed, URL should be preserved
        assert!(
            fixed.contains("<https://www.example.test>"),
            "Fix should preserve angle-bracket URLs: {fixed}"
        );
        assert!(
            fixed.contains("Test text"),
            "Fix should correct prose 'test' to 'Test': {fixed}"
        );
    }

    #[test]
    fn test_fallback_delimiter_index_matches_direct_scans() {
        let alphabet = [b'a', b'\\', b'[', b']', b'(', b')'];
        for mut value in 0..alphabet.len().pow(6) {
            let mut bytes = [0; 6];
            for byte in &mut bytes {
                *byte = alphabet[value % alphabet.len()];
                value /= alphabet.len();
            }
            for (open, close) in [(b'[', b']'), (b'(', b')')] {
                let mut lookup = FallbackDelimiters {
                    scanned_bytes: bytes.len() * 2,
                    ..Default::default()
                };
                for from in (1..=bytes.len()).chain((1..=bytes.len()).rev()) {
                    if bytes[from - 1] == open {
                        assert_eq!(
                            lookup.find(&bytes, from, open, close),
                            FallbackDelimiters::scan(&bytes, from, open, close),
                            "{bytes:?}, {from}, {open}"
                        );
                    }
                }
            }
        }
    }

    #[test]
    fn test_fallback_ranges_match_scalar_queries() {
        fn scalar_links(line: &str, pos: usize) -> bool {
            let bytes = line.as_bytes();
            let len = bytes.len();
            let mut i = 0;

            while i < len {
                // Look for unescaped '[' (handle double-escaped \\[ as unescaped)
                if bytes[i] == b'[' && (i == 0 || bytes[i - 1] != b'\\' || (i >= 2 && bytes[i - 2] == b'\\')) {
                    // Find matching ']' handling nested brackets
                    let mut depth: u32 = 1;
                    let mut j = i + 1;
                    while j < len && depth > 0 {
                        match bytes[j] {
                            b'\\' => {
                                j += 1; // skip escaped char
                            }
                            b'[' => depth += 1,
                            b']' => depth -= 1,
                            _ => {}
                        }
                        j += 1;
                    }

                    // j is now one past the ']'
                    if depth == 0 && j < len {
                        if bytes[j] == b'(' {
                            // Inline link: [text](url)
                            let url_start = j;
                            let mut paren_depth: u32 = 1;
                            let mut k = j + 1;
                            while k < len && paren_depth > 0 {
                                match bytes[k] {
                                    b'\\' => {
                                        k += 1; // skip escaped char
                                    }
                                    b'(' => paren_depth += 1,
                                    b')' => paren_depth -= 1,
                                    _ => {}
                                }
                                k += 1;
                            }

                            if paren_depth == 0 {
                                if pos > url_start && pos < k {
                                    return true;
                                }
                                i = k;
                                continue;
                            }
                        } else if bytes[j] == b'[' {
                            // Reference link: [text][ref]
                            let ref_start = j;
                            let mut ref_depth: u32 = 1;
                            let mut k = j + 1;
                            while k < len && ref_depth > 0 {
                                match bytes[k] {
                                    b'\\' => {
                                        k += 1;
                                    }
                                    b'[' => ref_depth += 1,
                                    b']' => ref_depth -= 1,
                                    _ => {}
                                }
                                k += 1;
                            }

                            if ref_depth == 0 {
                                if pos > ref_start && pos < k {
                                    return true;
                                }
                                i = k;
                                continue;
                            }
                        }
                    }
                }
                i += 1;
            }
            false
        }

        fn scalar_code(line: &str, pos: usize) -> bool {
            let bytes = line.as_bytes();
            let len = bytes.len();
            let mut i = 0;
            while i < len {
                if bytes[i] == b'`' {
                    // Count the opening backtick sequence length
                    let open_start = i;
                    while i < len && bytes[i] == b'`' {
                        i += 1;
                    }
                    let tick_len = i - open_start;

                    // Scan forward for a closing sequence of exactly tick_len backticks
                    while i < len {
                        if bytes[i] == b'`' {
                            let close_start = i;
                            while i < len && bytes[i] == b'`' {
                                i += 1;
                            }
                            if i - close_start == tick_len {
                                // Matched pair found; the code span content is between
                                // the end of the opening backticks and the start of the
                                // closing backticks (exclusive of the backticks themselves).
                                let content_start = open_start + tick_len;
                                let content_end = close_start;
                                if pos >= content_start && pos < content_end {
                                    return true;
                                }
                                // Continue scanning after this pair
                                break;
                            }
                            // Not the right length; keep scanning
                        } else {
                            i += 1;
                        }
                    }
                } else {
                    i += 1;
                }
            }
            false
        }

        let tokens = [
            "javascript",
            "`",
            "``",
            "```",
            "`javascript`",
            "``javascript``",
            "`javascript``",
            "\\`",
            "[javascript](url)",
            "[text][javascript]",
            "[[x]](url)",
            "[a [b] c](x(y)z)",
            "[a](x\\)y)",
            "[a][x\\]y]",
            "[a](x y)",
            "[a](x",
            "[a][x",
            "[[",
            "[",
            "]",
            "(",
            ")",
            "\\[",
            "\\\\[",
            "\\\\\\[",
            "\\\\",
            "é日本🙂",
            "<!--",
            "-->",
        ];
        for first in tokens {
            for second in tokens {
                let text = format!("{first}{second}");
                let urls = MD044ProperNames::markdown_link_url_ranges(&text);
                let code = MD044ProperNames::backtick_code_ranges(&text);
                for ranges in [&urls, &code] {
                    assert!(ranges.windows(2).all(|pair| pair[0].end <= pair[1].start));
                }
                for byte in 0..=text.len() + 1 {
                    assert_eq!(
                        crate::inline_config::offset_in_ranges(&urls, byte),
                        scalar_links(&text, byte),
                        "links {text:?} at {byte}: {urls:?}"
                    );
                    assert_eq!(
                        crate::inline_config::offset_in_ranges(&code, byte),
                        scalar_code(&text, byte),
                        "code {text:?} at {byte}: {code:?}"
                    );
                }
            }
        }
    }

    #[test]
    fn test_fallback_ranges_preserve_code_and_link_fixes() {
        let body =
            "javascript `javascript` [javascript](javascript) [javascript][javascript] <custom:javascript> CAFÉ ";
        let fixed =
            "JavaScript `javascript` [JavaScript](javascript) [JavaScript][javascript] <custom:javascript> Café ";
        for (prefix, suffix) in [
            ("<!--\n", "\n-->"),
            ("<div>\n", "\n</div>"),
            ("---\ntitle: ", "\n---\n"),
        ] {
            for ending in ["\n", "\r\n"] {
                let content = format!("{prefix}{}{suffix}", body.repeat(32)).replace('\n', ending);
                let expected = format!("{prefix}{}{suffix}", fixed.repeat(32)).replace('\n', ending);
                let ctx = create_context(&content);
                let rule = MD044ProperNames::new(vec!["JavaScript".into(), "Café".into()], false);
                let warnings = rule.check(&ctx).unwrap();
                assert_eq!(warnings.len(), 128, "{prefix:?} {ending:?}");
                assert_eq!(rule.check(&ctx).unwrap(), warnings);
                assert_eq!(rule.fix(&ctx).unwrap(), expected);
            }
        }
    }

    #[test]
    fn test_wikilink_url_ranges_match_scalar_queries() {
        fn scalar(ctx: &crate::lint_context::LintContext, byte_pos: usize) -> bool {
            use pulldown_cmark::LinkType;
            let content = ctx.content.as_bytes();

            for link in ctx.links_starting_before_or_at(byte_pos) {
                if !matches!(link.link_type, LinkType::WikiLink { .. }) {
                    continue;
                }
                let wiki_end = link.byte_end;
                // The WikiLink must end before byte_pos and be immediately followed by '('.
                if wiki_end >= byte_pos || wiki_end >= content.len() || content[wiki_end] != b'(' {
                    continue;
                }
                // Scan to the matching ')' tracking nested parens and backslash escapes.
                // Per CommonMark, an unquoted inline link destination cannot contain
                // spaces, tabs, or newlines. If we encounter one, this is parenthesised
                // prose rather than a URL, and pulldown-cmark will not parse it as a link.
                let mut depth: u32 = 1;
                let mut k = wiki_end + 1;
                let mut valid_destination = true;
                while k < content.len() && depth > 0 {
                    match content[k] {
                        b'\\' => {
                            k += 1; // skip escaped character
                        }
                        b'(' => depth += 1,
                        b')' => depth -= 1,
                        b' ' | b'\t' | b'\n' | b'\r' => {
                            valid_destination = false;
                            break;
                        }
                        _ => {}
                    }
                    k += 1;
                }
                // byte_pos is inside the URL if it falls between '(' and the matching ')'
                // and the destination is valid (no unescaped whitespace).
                if valid_destination && depth == 0 && byte_pos > wiki_end && byte_pos < k {
                    return true;
                }
            }
            false
        }

        use crate::config::MarkdownFlavor;
        let tokens = [
            "[[target]](javascript)",
            "[[target]](a(b)c)",
            "[[target]](a\\)b)",
            "[[target]](a\\ b)",
            "[[target]](a b)",
            "[[target]](a\tb)",
            "[[target]](a\nb)",
            "[[target]](a\rb)",
            "[[target]](a",
            "[[target]](a[[inner]](b)c)",
            "[[target]](a[[inner]](b)",
            "\\[[target]](path)",
            "`[[target]](path)`",
            "![x](path)",
            "é🙂 ",
            "[[x]]",
            "\\)",
            "([[target]](x))",
        ];
        for flavor in [
            MarkdownFlavor::Standard,
            MarkdownFlavor::Obsidian,
            MarkdownFlavor::MDX,
            MarkdownFlavor::Quarto,
        ] {
            for first in tokens {
                for second in tokens {
                    let content = format!("{first}{second}");
                    let ctx = LintContext::new(&content, flavor, None);
                    let ranges = MD044ProperNames::wikilink_url_ranges(&ctx);
                    assert!(ranges.windows(2).all(|pair| pair[0].end <= pair[1].start));
                    for byte in 0..=content.len() + 1 {
                        assert_eq!(
                            crate::inline_config::offset_in_ranges(&ranges, byte),
                            scalar(&ctx, byte),
                            "{flavor:?} {content:?} at {byte}: {ranges:?}"
                        );
                    }
                }
            }
        }
    }

    #[test]
    fn test_wikilink_destinations_preserve_prose_fixes() {
        use crate::config::MarkdownFlavor;
        for flavor in [MarkdownFlavor::Standard, MarkdownFlavor::Obsidian] {
            let content = "javascript [[target]](javascript) javascript [[other]](a(javascript)b) CAFÉ";
            let ctx = LintContext::new(content, flavor, None);
            let rule = MD044ProperNames::new(vec!["JavaScript".into(), "Café".into()], false);
            assert_eq!(rule.check(&ctx).unwrap().len(), 3);
            assert_eq!(
                rule.fix(&ctx).unwrap(),
                "JavaScript [[target]](javascript) JavaScript [[other]](a(javascript)b) Café"
            );
        }
    }

    #[test]
    fn test_repeated_name_columns_and_fixes_with_unicode() {
        for ending in ["\n", "\r\n"] {
            let line = "é日本🙂 javascript CAFÉ JavaScript <custom:javascript> ".repeat(64);
            let content = format!("{line}{ending}{ending}{line}");
            let ctx = create_context(&content);
            let rule = MD044ProperNames::new(vec!["JavaScript".into(), "Café".into()], false);
            let warnings = rule.check(&ctx).unwrap();
            let mut expected = Vec::new();
            for (line_idx, line_info) in ctx.lines.iter().enumerate() {
                let text = line_info.content(ctx.content);
                for (byte, _) in text.match_indices("javascript CAFÉ") {
                    for (offset, found, replacement) in [(0, "javascript", "JavaScript"), (11, "CAFÉ", "Café")] {
                        let start = byte + offset;
                        let column = text[..start].chars().count() + 1;
                        expected.push((
                            line_idx + 1,
                            column,
                            column + found.chars().count(),
                            line_info.byte_offset + start..line_info.byte_offset + start + found.len(),
                            replacement,
                        ));
                    }
                }
            }
            assert_eq!(warnings.len(), 256);
            assert_eq!(warnings.len(), expected.len());
            for (warning, (line, column, end_column, range, replacement)) in warnings.iter().zip(expected) {
                assert_eq!(
                    (warning.line, warning.column, warning.end_line, warning.end_column),
                    (line, column, line, end_column)
                );
                assert_eq!(warning.fix, Some(Fix::new(range, replacement.to_owned())));
            }
            assert_eq!(rule.check(&ctx).unwrap(), warnings, "cached diagnostics");
            assert_eq!(rule.clone().check(&ctx).unwrap(), warnings, "shared cache");
            assert_eq!(
                rule.fix(&ctx).unwrap(),
                content.replace("javascript CAFÉ", "JavaScript Café")
            );
        }
    }

    #[test]
    fn test_angle_url_ranges_preserve_boundaries() {
        let line = "<bad space> <x:one> <y:two\tthree> <z:unclosed <good:last> tail";
        let ranges = MD044ProperNames::angle_bracket_url_ranges(line);
        let expected = ["<x:one>", "<y:two\tthree>", "<good:last>"];
        assert_eq!(ranges.len(), expected.len());
        for (range, text) in ranges.iter().zip(expected) {
            assert_eq!(&line[range.clone()], text);
            assert!(MD044ProperNames::is_in_angle_bracket_url(line, range.start));
            assert!(MD044ProperNames::is_in_angle_bracket_url(line, range.end - 1));
            assert!(!MD044ProperNames::is_in_angle_bracket_url(line, range.end));
        }
    }

    #[test]
    fn test_is_in_angle_bracket_url_helper() {
        // Direct tests of the helper function
        let line = "text <https://example.test> more text";

        // Inside the URL
        assert!(MD044ProperNames::is_in_angle_bracket_url(line, 5)); // '<'
        assert!(MD044ProperNames::is_in_angle_bracket_url(line, 6)); // 'h'
        assert!(MD044ProperNames::is_in_angle_bracket_url(line, 15)); // middle of URL
        assert!(MD044ProperNames::is_in_angle_bracket_url(line, 26)); // '>'

        // Outside the URL
        assert!(!MD044ProperNames::is_in_angle_bracket_url(line, 0)); // 't' at start
        assert!(!MD044ProperNames::is_in_angle_bracket_url(line, 4)); // space before '<'
        assert!(!MD044ProperNames::is_in_angle_bracket_url(line, 27)); // space after '>'

        // Non-URL angle brackets
        assert!(!MD044ProperNames::is_in_angle_bracket_url("<notaurl>", 1));

        // mailto scheme
        assert!(MD044ProperNames::is_in_angle_bracket_url(
            "<mailto:test@example.com>",
            10
        ));

        // ftp scheme
        assert!(MD044ProperNames::is_in_angle_bracket_url(
            "<ftp://test.example.com>",
            10
        ));
    }

    #[test]
    fn test_is_in_angle_bracket_url_uppercase_scheme() {
        // RFC 3986: URI schemes are case-insensitive
        assert!(MD044ProperNames::is_in_angle_bracket_url(
            "<HTTPS://test.example.com>",
            10
        ));
        assert!(MD044ProperNames::is_in_angle_bracket_url(
            "<Http://test.example.com>",
            10
        ));
    }

    #[test]
    fn test_is_in_angle_bracket_url_uncommon_schemes() {
        // ssh scheme
        assert!(MD044ProperNames::is_in_angle_bracket_url(
            "<ssh://test@example.com>",
            10
        ));
        // file scheme
        assert!(MD044ProperNames::is_in_angle_bracket_url("<file:///test/path>", 10));
        // data scheme (no authority, just colon)
        assert!(MD044ProperNames::is_in_angle_bracket_url("<data:text/plain;test>", 10));
    }

    #[test]
    fn test_is_in_angle_bracket_url_unclosed() {
        // Unclosed angle bracket should NOT match
        assert!(!MD044ProperNames::is_in_angle_bracket_url(
            "<https://test.example.com",
            10
        ));
    }

    #[test]
    fn test_vale_inline_config_comments_not_flagged() {
        let config = MD044Config {
            names: vec!["Vale".to_string(), "JavaScript".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "\
<!-- vale off -->
Some javascript text here.
<!-- vale on -->
<!-- vale Style.Rule = NO -->
More javascript text.
<!-- vale Style.Rule = YES -->
<!-- vale JavaScript.Grammar = NO -->
";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Only the body text lines (2, 5) should be flagged for "javascript"
        assert_eq!(result.len(), 2, "Should only flag body lines, not Vale config comments");
        assert_eq!(result[0].line, 2);
        assert_eq!(result[1].line, 5);
    }

    #[test]
    fn test_remark_lint_inline_config_comments_not_flagged() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "\
<!-- lint disable remark-lint-some-rule -->
Some javascript text here.
<!-- lint enable remark-lint-some-rule -->
<!-- lint ignore remark-lint-some-rule -->
More javascript text.
";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(
            result.len(),
            2,
            "Should only flag body lines, not remark-lint config comments"
        );
        assert_eq!(result[0].line, 2);
        assert_eq!(result[1].line, 5);
    }

    #[test]
    fn test_fix_does_not_modify_vale_remark_lint_comments() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string(), "Vale".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "\
<!-- vale off -->
Some javascript text.
<!-- vale on -->
<!-- lint disable remark-lint-some-rule -->
More javascript text.
<!-- lint enable remark-lint-some-rule -->
";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();

        // Config directive lines must be preserved unchanged
        assert!(fixed.contains("<!-- vale off -->"));
        assert!(fixed.contains("<!-- vale on -->"));
        assert!(fixed.contains("<!-- lint disable remark-lint-some-rule -->"));
        assert!(fixed.contains("<!-- lint enable remark-lint-some-rule -->"));
        // Body text should be fixed
        assert!(fixed.contains("Some JavaScript text."));
        assert!(fixed.contains("More JavaScript text."));
    }

    #[test]
    fn test_mixed_tool_directives_all_skipped() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string(), "Vale".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        let content = "\
<!-- rumdl-disable MD044 -->
Some javascript text.
<!-- markdownlint-disable -->
More javascript text.
<!-- vale off -->
Even more javascript text.
<!-- lint disable some-rule -->
Final javascript text.
<!-- rumdl-enable MD044 -->
<!-- markdownlint-enable -->
<!-- vale on -->
<!-- lint enable some-rule -->
";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Only body text lines should be flagged (lines 2, 4, 6, 8)
        assert_eq!(
            result.len(),
            4,
            "Should only flag body lines, not any tool directive comments"
        );
        assert_eq!(result[0].line, 2);
        assert_eq!(result[1].line, 4);
        assert_eq!(result[2].line, 6);
        assert_eq!(result[3].line, 8);
    }

    #[test]
    fn test_vale_remark_lint_edge_cases_not_matched() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string(), "Vale".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        // These are regular HTML comments, NOT tool directives:
        // - "<!-- vale -->" is not a valid Vale directive (no action keyword)
        // - "<!-- vale is a tool -->" starts with "vale" but is prose, not a directive
        // - "<!-- valedictorian javascript -->" does not start with "<!-- vale "
        // - "<!-- linting javascript tips -->" does not start with "<!-- lint "
        // - "<!-- vale javascript -->" starts with "vale" but has no action keyword
        // - "<!-- lint your javascript code -->" starts with "lint" but has no action keyword
        let content = "\
<!-- vale -->
<!-- vale is a tool for writing -->
<!-- valedictorian javascript -->
<!-- linting javascript tips -->
<!-- vale javascript -->
<!-- lint your javascript code -->
";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Line 1: "<!-- vale -->" contains "vale" (wrong case for "Vale") -> flagged
        // Line 2: "<!-- vale is a tool for writing -->" contains "vale" -> flagged
        // Line 3: "<!-- valedictorian javascript -->" contains "javascript" -> flagged
        // Line 4: "<!-- linting javascript tips -->" contains "javascript" -> flagged
        // Line 5: "<!-- vale javascript -->" contains "vale" and "javascript" -> flagged for both
        // Line 6: "<!-- lint your javascript code -->" contains "javascript" -> flagged
        assert_eq!(
            result.len(),
            7,
            "Should flag proper names in non-directive HTML comments: got {result:?}"
        );
        assert_eq!(result[0].line, 1); // "vale" in <!-- vale -->
        assert_eq!(result[1].line, 2); // "vale" in <!-- vale is a tool -->
        assert_eq!(result[2].line, 3); // "javascript" in <!-- valedictorian javascript -->
        assert_eq!(result[3].line, 4); // "javascript" in <!-- linting javascript tips -->
        assert_eq!(result[4].line, 5); // "vale" in <!-- vale javascript -->
        assert_eq!(result[5].line, 5); // "javascript" in <!-- vale javascript -->
        assert_eq!(result[6].line, 6); // "javascript" in <!-- lint your javascript code -->
    }

    #[test]
    fn test_vale_style_directives_skipped() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string(), "Vale".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);

        // These ARE valid Vale directives and should be skipped:
        let content = "\
<!-- vale style = MyStyle -->
<!-- vale styles = Style1, Style2 -->
<!-- vale MyRule.Name = YES -->
<!-- vale MyRule.Name = NO -->
Some javascript text.
";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();

        // Only line 5 (body text) should be flagged
        assert_eq!(
            result.len(),
            1,
            "Should only flag body lines, not Vale style/rule directives: got {result:?}"
        );
        assert_eq!(result[0].line, 5);
    }

    // --- is_in_backtick_code_in_line unit tests ---

    #[test]
    fn test_backtick_code_single_backticks() {
        let line = "hello `world` bye";
        // 'w' is at index 7, inside the backtick span (content between backticks at 6 and 12)
        assert!(MD044ProperNames::is_in_backtick_code_in_line(line, 7));
        // 'h' at index 0 is outside
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 0));
        // 'b' at index 14 is outside
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 14));
    }

    #[test]
    fn test_backtick_code_double_backticks() {
        let line = "a ``code`` b";
        // 'c' is at index 4, inside ``...``
        assert!(MD044ProperNames::is_in_backtick_code_in_line(line, 4));
        // 'a' at index 0 is outside
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 0));
        // 'b' at index 11 is outside
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 11));
    }

    #[test]
    fn test_backtick_code_unclosed() {
        let line = "a `code b";
        // No closing backtick, so nothing is a code span
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 3));
    }

    #[test]
    fn test_backtick_code_mismatched_count() {
        // Single backtick opening, double backtick is not a match
        let line = "a `code`` b";
        // The single ` at index 2 doesn't match `` at index 7-8
        // So 'c' at index 3 is NOT in a code span
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 3));
    }

    #[test]
    fn test_backtick_code_multiple_spans() {
        let line = "`first` and `second`";
        // 'f' at index 1 (inside first span)
        assert!(MD044ProperNames::is_in_backtick_code_in_line(line, 1));
        // 'a' at index 8 (between spans)
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 8));
        // 's' at index 13 (inside second span)
        assert!(MD044ProperNames::is_in_backtick_code_in_line(line, 13));
    }

    #[test]
    fn test_backtick_code_on_backtick_boundary() {
        let line = "`code`";
        // Position 0 is the opening backtick itself, not inside the span
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 0));
        // Position 5 is the closing backtick, not inside the span
        assert!(!MD044ProperNames::is_in_backtick_code_in_line(line, 5));
        // Position 1-4 are inside the span
        assert!(MD044ProperNames::is_in_backtick_code_in_line(line, 1));
        assert!(MD044ProperNames::is_in_backtick_code_in_line(line, 4));
    }

    // Double-bracket WikiLink + URL: [[text]](url)
    // pulldown-cmark parses [[text]] as a WikiLink but leaves the (url)
    // as plain text, so ctx.links() does not cover the URL portion.
    // MD044 must fall back to markdown_link_url_ranges for all lines.

    #[test]
    fn test_double_bracket_link_url_not_flagged() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        // Exact reproduction from issue #564
        let content = "[[rumdl]](https://github.com/rvben/rumdl)";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "URL inside [[text]](url) must not be flagged, got: {result:?}"
        );
    }

    #[test]
    fn test_double_bracket_link_url_not_fixed() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "[[rumdl]](https://github.com/rvben/rumdl)\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(
            fixed, content,
            "fix() must leave the URL inside [[text]](url) unchanged"
        );
    }

    #[test]
    fn test_double_bracket_link_text_still_flagged() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        // The link text portion [[github]](url) should still be checked.
        let content = "[[github]](https://example.com)";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "Incorrect name in [[text]] link text should still be flagged, got: {result:?}"
        );
        assert_eq!(result[0].message, "Proper name 'github' should be 'GitHub'");
    }

    #[test]
    fn test_double_bracket_link_mixed_line() {
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        // URL must be skipped, standalone text must be flagged.
        let content = "See [[rumdl]](https://github.com/rvben/rumdl) and github for more.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "Only the standalone 'github' after the link should be flagged, got: {result:?}"
        );
        assert!(result[0].message.contains("'github'"));
        // "See " (4) + "[[rumdl]](https://github.com/rvben/rumdl)" (42) + " and " (4) = column 51
        assert_eq!(
            result[0].column, 51,
            "Flagged column should be the trailing 'github', not the one in the URL"
        );
    }

    #[test]
    fn test_regular_link_url_still_not_flagged() {
        // Confirm existing [text](url) behavior is unaffected by the fix.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "[rumdl](https://github.com/rvben/rumdl)";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "URL inside regular [text](url) must still not be flagged, got: {result:?}"
        );
    }

    #[test]
    fn test_link_like_text_in_code_span_still_flagged_when_code_blocks_enabled() {
        // When code-blocks = true the user explicitly opts into checking code spans.
        // A code span containing link-like text (`[foo](https://github.com)`) must
        // NOT be silently suppressed by markdown_link_url_ranges: the content is
        // literal characters, not a real Markdown link.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], true);
        let content = "`[foo](https://github.com/org/repo)`";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "Proper name inside a code span must be flagged when code-blocks=true, got: {result:?}"
        );
        assert!(result[0].message.contains("'github'"));
    }

    #[test]
    fn test_malformed_link_not_treated_as_url() {
        // [text](url with spaces) is NOT a valid Markdown link; pulldown-cmark
        // does not parse it, so the name inside must still be flagged.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "See [rumdl](github repo) for details.";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "Name inside malformed [text](url with spaces) must still be flagged, got: {result:?}"
        );
        assert!(result[0].message.contains("'github'"));
    }

    #[test]
    fn test_wikilink_followed_by_prose_parens_still_flagged() {
        // [[note]](github repo) — WikiLink followed by parenthesised prose, NOT
        // a valid link URL (space in destination). pulldown-cmark does not parse
        // it as a link, so the name inside must still be flagged.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "[[note]](github repo)";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "Name inside [[wikilink]](prose with spaces) must still be flagged, got: {result:?}"
        );
        assert!(result[0].message.contains("'github'"));
    }

    /// Roundtrip safety: fix() output must produce zero warnings on re-check.
    #[test]
    fn test_roundtrip_fix_then_check_basic() {
        let rule = MD044ProperNames::new(
            vec![
                "JavaScript".to_string(),
                "TypeScript".to_string(),
                "Node.js".to_string(),
            ],
            true,
        );
        let content = "I love javascript, typescript, and nodejs!";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        let ctx2 = create_context(&fixed);
        let warnings = rule.check(&ctx2).unwrap();
        assert!(
            warnings.is_empty(),
            "Re-check after fix should produce zero warnings, got: {warnings:?}"
        );
    }

    /// Roundtrip safety: fix() output must produce zero warnings for multiline content.
    #[test]
    fn test_roundtrip_fix_then_check_multiline() {
        let rule = MD044ProperNames::new(vec!["Rust".to_string(), "Python".to_string()], true);
        let content = "First line with rust.\nSecond line with python.\nThird line with RUST and PYTHON.\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        let ctx2 = create_context(&fixed);
        let warnings = rule.check(&ctx2).unwrap();
        assert!(
            warnings.is_empty(),
            "Re-check after fix should produce zero warnings, got: {warnings:?}"
        );
    }

    /// Roundtrip safety: fix() with inline config disable blocks.
    #[test]
    fn test_roundtrip_fix_then_check_inline_config() {
        let config = MD044Config {
            names: vec!["RUMDL".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);
        let content =
            "<!-- rumdl-disable MD044 -->\nSome rumdl text.\n<!-- rumdl-enable MD044 -->\n\nSome rumdl text outside.\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        // The disabled block should be preserved, the outside text fixed
        assert!(
            fixed.contains("Some rumdl text.\n"),
            "Disabled block text should be preserved"
        );
        assert!(
            fixed.contains("Some RUMDL text outside."),
            "Outside text should be fixed"
        );
    }

    /// Roundtrip safety: fix() with HTML comment content.
    #[test]
    fn test_roundtrip_fix_then_check_html_comments() {
        let config = MD044Config {
            names: vec!["JavaScript".to_string()],
            ..MD044Config::default()
        };
        let rule = MD044ProperNames::from_config_struct(config);
        let content = "# Guide\n\n<!-- javascript mentioned here -->\n\njavascript outside\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        let ctx2 = create_context(&fixed);
        let warnings = rule.check(&ctx2).unwrap();
        assert!(
            warnings.is_empty(),
            "Re-check after fix should produce zero warnings, got: {warnings:?}"
        );
    }

    /// Roundtrip safety: fix() preserves content when no violations exist.
    #[test]
    fn test_roundtrip_no_op_when_correct() {
        let rule = MD044ProperNames::new(vec!["JavaScript".to_string(), "TypeScript".to_string()], true);
        let content = "This uses JavaScript and TypeScript correctly.\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Fix should be a no-op when content is already correct");
    }

    // --- Bare-domain link text: display text is the destination URL with scheme stripped ---

    #[test]
    fn test_bare_domain_link_text_not_flagged() {
        // `[ravencentric.github.io](https://ravencentric.github.io)` — the display text
        // is the URL with the scheme stripped; "github" here is a domain label, not a
        // reference to "GitHub" the product, and must not be corrected.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "My site is [ravencentric.github.io](https://ravencentric.github.io).\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Should not flag 'github' in a bare-domain link text that matches the link URL: {result:?}"
        );
    }

    #[test]
    fn test_bare_domain_link_text_not_fixed() {
        // fix() must not rewrite the link text when it is the bare URL hostname.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "My site is [ravencentric.github.io](https://ravencentric.github.io).\n";
        let ctx = create_context(content);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(
            fixed, content,
            "fix() must not alter bare-domain link text that matches the destination URL"
        );
    }

    #[test]
    fn test_bare_domain_link_text_with_path_not_flagged() {
        // Display text is the hostname only; destination has a path.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "Visit [ravencentric.github.io](https://ravencentric.github.io/projects).\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Should not flag 'github' when bare-domain text is the hostname of its destination URL: {result:?}"
        );
    }

    #[test]
    fn test_bare_domain_link_text_full_path_not_flagged() {
        // Display text is the full URL-without-scheme including a path.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "See [ravencentric.github.io/blog](https://ravencentric.github.io/blog).\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Should not flag 'github' when link text is the full URL path without scheme: {result:?}"
        );
    }

    #[test]
    fn test_github_product_name_in_link_text_still_flagged() {
        // `[github pages](https://pages.github.com)` — the display text is a human
        // description, not a bare domain; "github" should still be corrected to "GitHub".
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "Hosted on [github pages](https://pages.github.com).\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            !result.is_empty(),
            "Should still flag 'github' in descriptive link text that does not match the destination URL"
        );
    }

    #[test]
    fn test_protocol_relative_bare_domain_link_text_not_flagged() {
        // Protocol-relative URL `[github.io](//github.io)`.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "See [github.io](//github.io).\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Should not flag 'github' in bare-domain text matching a protocol-relative destination: {result:?}"
        );
    }

    #[test]
    fn test_dotted_wikilink_target_still_flagged() {
        // `[[node.js]]` is a WikiLink whose page name contains a dot.
        // The dot guard alone does not protect it because text == url == "node.js".
        // The is_in_link WikiLink guard must prevent bare-domain suppression,
        // so the improper capitalization is still caught.
        let rule = MD044ProperNames::new(vec!["Node.js".to_string()], false);
        let content = "See [[node.js]] for details.\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            !result.is_empty(),
            "Should flag 'node.js' in a dotted WikiLink target: {result:?}"
        );
    }

    #[test]
    fn test_bare_domain_link_text_case_insensitive_url() {
        // URL with uppercase scheme `[github.io](HTTPS://github.io)` — the scheme is
        // case-insensitive, so the display text should still be recognised as a bare domain.
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let content = "See [github.io](HTTPS://github.io).\n";
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Should not flag bare-domain text when destination URL has an uppercase scheme: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_value_span_strips_trailing_comment() {
        let line = "link: docs/guide/myapp # canonical path";
        let (s, e) = frontmatter_values::value_span(line).unwrap();
        assert_eq!(&line[s..e], "docs/guide/myapp");
    }

    #[test]
    fn test_frontmatter_value_span_quoted_keeps_hash_and_spaces() {
        let line = "link: 'docs/My App/a#b'";
        let (s, e) = frontmatter_values::value_span(line).unwrap();
        assert_eq!(&line[s..e], "docs/My App/a#b");
    }

    #[test]
    fn test_frontmatter_value_span_plain_value() {
        let line = "title: Heading for myapp";
        let (s, e) = frontmatter_values::value_span(line).unwrap();
        assert_eq!(&line[s..e], "Heading for myapp");
    }

    #[test]
    fn test_frontmatter_value_span_none_for_key_only() {
        assert!(frontmatter_values::value_span("seo:").is_none());
        assert!(frontmatter_values::value_span("---").is_none());
    }

    #[test]
    fn test_frontmatter_value_span_quoted_strips_trailing_comment() {
        let line = "link: 'docs/guide' # canonical path";
        let (s, e) = frontmatter_values::value_span(line).unwrap();
        assert_eq!(&line[s..e], "docs/guide");
    }

    #[test]
    fn test_frontmatter_value_span_empty_quoted_value_is_none() {
        assert!(frontmatter_values::value_span("key: ''").is_none());
    }

    #[test]
    fn test_frontmatter_value_span_unterminated_quote_strips_leading_quote() {
        let line = "link: 'docs/a";
        let (s, e) = frontmatter_values::value_span(line).unwrap();
        assert_eq!(&line[s..e], "docs/a");
    }

    /// Byte offset of `needle` in `line`, for locating the match under test.
    fn at(line: &str, needle: &str) -> usize {
        line.find(needle).expect("needle present")
    }

    #[test]
    fn test_path_like_exempts_single_token_frontmatter_paths() {
        for line in [
            "link: this/is/a/link/to/myapp.md",
            "link: docs/myapp.md",
            "link: /abs/path/myapp.md",
            "link: ./myapp.md",
            "link: ../shared/myapp.md",
        ] {
            let span = frontmatter_values::value_span(line).unwrap();
            let pos = at(line, "myapp");
            assert!(
                MD044ProperNames::is_in_path_like_token(line, pos, span),
                "should treat as a path: {line}"
            );
        }
    }

    #[test]
    fn test_path_like_does_not_exempt_slash_conjunction_prose() {
        // Extra words around the slash-separated token mean it is not the
        // sole frontmatter value, so the 3+ segment path signal must not fire.
        let line = "description: We support github/gitlab/bitbucket imports.";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(
            !MD044ProperNames::is_in_path_like_token(line, at(line, "github"), span),
            "slash-separated prose is not a path"
        );

        let line = "description: The javascript/typescript ecosystem is large.";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(!MD044ProperNames::is_in_path_like_token(
            line,
            at(line, "javascript"),
            span
        ));
    }

    #[test]
    fn test_path_like_requires_a_slash_so_dotted_names_survive() {
        let line = "title: Use nodejs and myapp.md today.";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(!MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));
    }

    #[test]
    fn test_path_like_no_slash_frontmatter_value_still_flagged() {
        // A frontmatter value with no slash at all is never a path signal,
        // regardless of it being the sole value; the mandatory slash is what
        // protects dotted proper names like `Node.js` without over-exempting
        // plain slugs.
        let line = "slug: myapp-guide";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(!MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));
    }

    #[test]
    fn test_path_like_returns_false_outside_value_span() {
        // A match outside the frontmatter value span (e.g. in the key) is
        // rejected immediately, before any token-bound scanning happens.
        let line = "myapp: docs/guide/myapp";
        let span = frontmatter_values::value_span(line).unwrap();
        let key_pos = 0;
        assert!(!MD044ProperNames::is_in_path_like_token(line, key_pos, span));
    }

    #[test]
    fn test_path_like_three_segments_only_as_sole_frontmatter_value() {
        let line = "link: docs/guide/myapp";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));

        let line = "description: We support github/gitlab/bitbucket now";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(
            !MD044ProperNames::is_in_path_like_token(line, at(line, "github"), span),
            "multi-token value gets body treatment"
        );
    }

    #[test]
    fn test_path_like_quoted_value_with_spaces() {
        // The collapsed token has a real extension on its last segment, so
        // signal (b) exempts it regardless of the multi-word collapse.
        let line = "link: 'docs/My App/myapp.md'";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));
    }

    #[test]
    fn test_path_like_quoted_value_with_spaces_no_extension_not_exempt() {
        // Same shape as above but without an extension: the collapsed token
        // only has the 3+ segment signal available, which is deliberately
        // restricted to single-token values (see `is_multi_word_collapse` in
        // `is_in_path_like_token`). An extensionless path containing a
        // literal space is rare enough that this is an accepted narrowing,
        // not an oversight.
        let line = "link: 'docs/My App/myapp'";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(!MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));
    }

    #[test]
    fn test_path_like_trailing_comment_is_still_sole_value() {
        let line = "link: docs/guide/myapp # canonical path";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));
    }

    #[test]
    fn test_path_like_trailing_punctuation_trimmed() {
        let line = "link: docs/myapp.md, then leave.";
        let span = frontmatter_values::value_span(line).unwrap();
        assert!(MD044ProperNames::is_in_path_like_token(line, at(line, "myapp"), span));
    }

    #[test]
    fn test_trim_token_bounds_reaches_fixpoint_after_punctuation_exposes_wrapper() {
        let line = r#"See "docs/myapp.md", then leave."#;
        let raw_start = at(line, "\"docs");
        let raw_end = raw_start + r#""docs/myapp.md","#.len();
        assert_eq!(&line[raw_start..raw_end], r#""docs/myapp.md","#);
        let (start, end) = frontmatter_values::trim_token_bounds(line, raw_start, raw_end);
        assert_eq!(&line[start..end], "docs/myapp.md");
    }

    #[test]
    fn test_trim_token_bounds_reaches_fixpoint_with_multiple_trailing_wrappers() {
        let line = r#"("docs/myapp.md")."#;
        let (start, end) = frontmatter_values::trim_token_bounds(line, 0, line.len());
        assert_eq!(&line[start..end], "docs/myapp.md");
    }

    #[test]
    fn test_frontmatter_link_path_not_flagged() {
        let content = "---\ntitle: Heading for MyApp\nlink: 'this/is/a/link/to/myapp.md'\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "path in a frontmatter value must not be flagged: {result:?}"
        );
    }

    #[test]
    fn test_fix_does_not_corrupt_frontmatter_link_path() {
        let content = "---\nlink: 'this/is/a/link/to/myapp.md'\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        assert_eq!(rule.fix(&ctx).unwrap(), content, "fix must not rewrite a path");
    }

    // The next few tests document the deliberate frontmatter-only scope: a
    // body occurrence that sits inside text shaped like a file path (a
    // parenthesised disambiguator, a bracketed dynamic segment, a Next.js
    // catch-all route) is corrected exactly like any other prose occurrence,
    // and only that single word changes. Earlier attempts at exempting
    // body-prose paths corrupted these exact shapes (splitting mid-token on
    // wrapper characters, or on `[[`/`]]`/`](` sequences); asserting the
    // plain single-word correction here guards against that class of bug
    // reappearing without silently reintroducing the abandoned exemption.

    #[test]
    fn test_body_prose_parenthesized_disambiguator_is_case_corrected() {
        let content = "See docs/myapp(1).md here.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 1, "body occurrence is flagged: {result:?}");
        assert_eq!(rule.fix(&ctx).unwrap(), "See docs/MyApp(1).md here.\n");
    }

    #[test]
    fn test_body_prose_bracketed_dynamic_segment_is_case_corrected() {
        let content = "See docs/[myapp].md here.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 1, "body occurrence is flagged: {result:?}");
        assert_eq!(rule.fix(&ctx).unwrap(), "See docs/[MyApp].md here.\n");
    }

    #[test]
    fn test_body_prose_nextjs_catch_all_segment_is_case_corrected() {
        // `[[...myapp]]` is a Next.js optional catch-all route segment, not
        // WikiLink syntax; nothing in the parser treats it specially here.
        let content = "pages/[[...myapp]].tsx are catch-all routes.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 1, "body occurrence is flagged: {result:?}");
        assert_eq!(
            rule.fix(&ctx).unwrap(),
            "pages/[[...MyApp]].tsx are catch-all routes.\n"
        );
    }

    #[test]
    fn test_two_adjacent_whitespace_free_links_both_flagged() {
        // Two links with no whitespace between them, `[myapp](url)[github](url)`.
        // A body-prose tokenizer that split on markdown syntax previously let
        // the first link's boundary swallow the second, losing its flag.
        let content = "[myapp](https://a.com)[github](https://b.com)\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string(), "GitHub".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 2, "both link texts must be flagged: {result:?}");
        assert!(result.iter().any(|w| w.message.contains("'myapp'")));
        assert!(result.iter().any(|w| w.message.contains("'github'")));
    }

    #[test]
    fn test_fix_does_not_corrupt_frontmatter_path_with_route_group_named_after_proper_name() {
        // Next.js-style route group: the parenthesised directory segment
        // itself is the proper name, e.g. `src/(myapp)/page.tsx`.
        let content = "---\nlink: src/(myapp)/page.tsx\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        assert_eq!(
            rule.fix(&ctx).unwrap(),
            content,
            "fix must not rewrite a frontmatter path whose route-group directory name is the proper name"
        );
    }

    #[test]
    fn test_quoted_frontmatter_value_slash_conjunction_prose_still_flagged() {
        // A quoted value is not automatically a single path token just
        // because it is quoted: "We" carries no slash, so per-word
        // tokenization applies and the slash-separated list is judged as
        // prose, not the sole value.
        let content = "---\ndescription: \"We support github/gitlab/bitbucket now\"\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "quoted prose value must still flag 'github': {result:?}"
        );
    }

    #[test]
    fn test_quoted_frontmatter_value_single_slash_word_with_unrelated_dot_still_flagged() {
        // One slash-bearing word plus an unrelated later dot (in "1.0" or
        // "e.g.") must not make the whole quoted sentence exempt as a path.
        let content = "---\ndescription: \"We use myapp/gitlab and version 1.0 e.g. weekly\"\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "quoted prose value must still flag 'myapp': {result:?}"
        );
    }

    #[test]
    fn test_quoted_toml_frontmatter_value_slash_conjunction_prose_still_flagged() {
        // TOML string values are essentially always quoted, so this class of
        // bug affects TOML frontmatter systematically.
        let content = "+++\ndescription = \"We support github/gitlab/bitbucket now\"\n+++\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "TOML quoted prose value must still flag 'github': {result:?}"
        );
    }

    #[test]
    fn test_path_like_collapsed_multiword_no_extension_not_exempt() {
        // Every word in these quoted values carries a slash, so they collapse
        // to one token per `is_single_quoted_path`. None of the collapsed
        // tokens starts with a path prefix or ends in an extension, so the
        // 3+ segment sole-value signal must not exempt them either: it is
        // restricted to genuinely single-token values, not tokens formed by
        // collapsing several whitespace-separated words together.
        for line in [
            r#"description: "myapp/gitlab github/bitbucket""#,
            r#"description: "and/or this/that myapp/gitlab""#,
            r#"description: "he/him she/her myapp/gitlab""#,
        ] {
            let span = frontmatter_values::value_span(line).unwrap();
            for needle in ["myapp", "gitlab"] {
                if let Some(byte_pos) = line.find(needle) {
                    assert!(
                        !MD044ProperNames::is_in_path_like_token(line, byte_pos, span),
                        "collapsed multi-word value must not exempt '{needle}': {line}"
                    );
                }
            }
        }
    }

    #[test]
    fn test_path_like_collapsed_multiword_no_extension_not_exempt_toml() {
        let line = r#"description = "myapp/gitlab github/bitbucket""#;
        let span = frontmatter_values::value_span(line).unwrap();
        for needle in ["myapp", "gitlab", "github", "bitbucket"] {
            let byte_pos = at(line, needle);
            assert!(
                !MD044ProperNames::is_in_path_like_token(line, byte_pos, span),
                "collapsed multi-word TOML value must not exempt '{needle}'"
            );
        }
    }

    #[test]
    fn test_frontmatter_collapsed_multiword_names_all_flagged_yaml() {
        let content = "---\ndescription: \"myapp/gitlab github/bitbucket\"\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(
            vec![
                "MyApp".to_string(),
                "GitLab".to_string(),
                "GitHub".to_string(),
                "Bitbucket".to_string(),
            ],
            false,
        );
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            4,
            "all four names in the collapsed multi-word value must be flagged: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_collapsed_multiword_names_all_flagged_toml() {
        let content = "+++\ndescription = \"myapp/gitlab github/bitbucket\"\n+++\n\nBody.\n";
        let rule = MD044ProperNames::new(
            vec![
                "MyApp".to_string(),
                "GitLab".to_string(),
                "GitHub".to_string(),
                "Bitbucket".to_string(),
            ],
            false,
        );
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            4,
            "all four names in the collapsed multi-word TOML value must be flagged: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_collapsed_multiword_conjunction_pairs_flagged() {
        let content = "---\ndescription: \"and/or this/that myapp/gitlab\"\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string(), "GitLab".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            2,
            "myapp and gitlab must both be flagged despite the surrounding slash pairs: {result:?}"
        );
    }

    #[test]
    fn test_frontmatter_collapsed_multiword_pronoun_pairs_flagged() {
        let content = "---\ndescription: \"he/him she/her myapp/gitlab\"\n---\n\nBody.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string(), "GitLab".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            2,
            "myapp and gitlab must both be flagged despite the surrounding slash pairs: {result:?}"
        );
    }

    /// The path exemption is scoped to frontmatter only, by design: a body
    /// occurrence sitting inside what looks like a file path is still
    /// flagged and fixed like any other prose occurrence. This is a
    /// deliberate limit (see `is_in_path_like_token`), not an oversight, so a
    /// future reader does not "fix" it back into hand-rolled body tokenizing.
    #[test]
    fn test_body_prose_path_is_flagged_frontmatter_only_scope() {
        let content = "See docs/myapp.md for details about myapp.\n";
        let rule = MD044ProperNames::new(vec!["MyApp".to_string()], false);
        let ctx = create_context(content);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            2,
            "both the path occurrence and the prose occurrence are flagged in body text: {result:?}"
        );
        assert_eq!(rule.fix(&ctx).unwrap(), "See docs/MyApp.md for details about MyApp.\n");
    }

    #[test]
    fn test_slash_conjunction_prose_still_flagged() {
        let content = "We support github/gitlab imports.\n";
        let rule = MD044ProperNames::new(vec!["GitHub".to_string()], false);
        let ctx = create_context(content);
        assert_eq!(rule.check(&ctx).unwrap().len(), 1);
    }
}