rumdl_lib/rules/

md033_no_inline_html.rs

//!
//! Rule MD033: No HTML tags
//!
//! See [docs/md033.md](../../docs/md033.md) for full documentation, configuration, and examples.

use crate::config::MarkdownFlavor;
use crate::lint_context::HtmlTag;
use crate::rule::{Fix, LintError, LintResult, LintWarning, Rule, RuleCategory, Severity};
use crate::utils::regex_cache::*;
use std::collections::{HashMap, HashSet};
use std::ops::Range;

mod md033_config;
use crate::utils::html_elements::is_void_element;
use md033_config::{MD033Config, MD033FixMode, is_permitted_without_markdown_equivalent};

/// Byte offsets of every literal closing tag (`</name>`) in a document, by
/// ASCII-lowercased tag name, so pairing an opening tag with its closing tag is a
/// binary search rather than a scan of the rest of the document.
#[derive(Default)]
struct ClosingTagIndex {
    positions: HashMap<String, Vec<usize>>,
}

impl ClosingTagIndex {
    fn new(content: &str) -> Self {
        let bytes = content.as_bytes();
        let mut positions: HashMap<String, Vec<usize>> = HashMap::new();
        let mut from = 0;
        while let Some(offset) = content[from..].find("</") {
            let start = from + offset;
            let name_start = start + 2;
            // A name ends at `>`; anything that cannot be part of one ends the
            // candidate. Stopping at `<` keeps every byte in at most one window.
            let name_len = bytes[name_start..]
                .iter()
                .position(|&b| b == b'>' || b == b'<' || b == b'/' || b.is_ascii_whitespace());
            match name_len {
                Some(len) if len > 0 && bytes[name_start + len] == b'>' => {
                    let name = content[name_start..name_start + len].to_ascii_lowercase();
                    positions.entry(name).or_default().push(start);
                    from = name_start + len;
                }
                Some(len) => from = name_start + len,
                None => break,
            }
        }
        Self { positions }
    }

    /// The offset of the first `</tag_name>` starting at or after `offset`.
    fn first_at_or_after(&self, tag_name: &str, offset: usize) -> Option<usize> {
        let positions = self.positions.get(tag_name)?;
        positions.get(positions.partition_point(|&p| p < offset)).copied()
    }
}

#[derive(Clone)]
pub struct MD033NoInlineHtml {
    config: MD033Config,
    allowed: HashSet<String>,
    allowed_inside: HashSet<String>,
    table_allowed: HashSet<String>,
    disallowed: HashSet<String>,
    drop_attributes: HashSet<String>,
    strip_wrapper_elements: HashSet<String>,
    allows_no_markdown_equivalent: bool,
    table_allows_no_markdown_equivalent: bool,
}

impl Default for MD033NoInlineHtml {
    fn default() -> Self {
        Self::from_config_struct(MD033Config::default())
    }
}

impl MD033NoInlineHtml {
    pub fn new() -> Self {
        Self::default()
    }

    pub fn with_allowed(allowed_vec: Vec<String>) -> Self {
        Self::from_config_struct(MD033Config {
            allowed: allowed_vec,
            ..MD033Config::default()
        })
    }

    pub fn with_disallowed(disallowed_vec: Vec<String>) -> Self {
        Self::from_config_struct(MD033Config {
            disallowed: disallowed_vec,
            ..MD033Config::default()
        })
    }

    /// Create a new rule with auto-fix enabled
    pub fn with_fix(fix: bool) -> Self {
        Self::from_config_struct(MD033Config {
            fix,
            ..MD033Config::default()
        })
    }

    /// Single source of truth for building an `MD033NoInlineHtml` from config.
    /// Pre-computes all lowercase HashSets so per-line lookups are O(1).
    pub fn from_config_struct(config: MD033Config) -> Self {
        let allowed = config.allowed_set();
        let allowed_inside = config.allowed_inside_set();
        let table_allowed = config.table_allowed_set();
        let disallowed = config.disallowed_set();
        let drop_attributes = config.drop_attributes_set();
        let strip_wrapper_elements = config.strip_wrapper_elements_set();
        let allows_no_markdown_equivalent = config.allows_no_markdown_equivalent();
        let table_allows_no_markdown_equivalent = config.table_allows_no_markdown_equivalent();
        Self {
            config,
            allowed,
            allowed_inside,
            table_allowed,
            disallowed,
            drop_attributes,
            strip_wrapper_elements,
            allows_no_markdown_equivalent,
            table_allows_no_markdown_equivalent,
        }
    }

    /// Extract the lowercase tag name from a raw tag string like `<br/>` or
    /// `</div >`. Strips angle brackets, leading slash, and stops at the first
    /// whitespace, `>`, or `/`. Returns an empty string if no name is present.
    #[inline]
    fn extract_tag_name(tag: &str) -> String {
        let trimmed = tag.trim_start_matches('<').trim_start_matches('/');
        trimmed
            .split(|c: char| c.is_whitespace() || c == '>' || c == '/')
            .next()
            .unwrap_or("")
            .to_lowercase()
    }

    /// Membership check against a precomputed lowercase set, returning false
    /// fast when the set is empty.
    #[inline]
    fn tag_in_set(set: &HashSet<String>, tag: &str) -> bool {
        if set.is_empty() {
            return false;
        }
        set.contains(&Self::extract_tag_name(tag))
    }

    /// Check whether the tag is permitted by the general `allowed_elements` list.
    #[inline]
    fn is_tag_allowed(&self, tag: &str, flavor: MarkdownFlavor) -> bool {
        if !self.allows_no_markdown_equivalent {
            return Self::tag_in_set(&self.allowed, tag);
        }
        let name = Self::extract_tag_name(tag);
        self.allowed.contains(&name) || is_permitted_without_markdown_equivalent(&name, flavor, false)
    }

    /// Check whether the tag is permitted inside a GFM table cell.
    /// Uses `table_allowed_elements` if configured, falling back to `allowed`.
    #[inline]
    fn is_tag_allowed_in_table(&self, tag: &str, flavor: MarkdownFlavor) -> bool {
        if !self.table_allows_no_markdown_equivalent {
            return Self::tag_in_set(&self.table_allowed, tag);
        }
        let name = Self::extract_tag_name(tag);
        self.table_allowed.contains(&name) || is_permitted_without_markdown_equivalent(&name, flavor, true)
    }

    /// Whether an element holds no content, so naming it in `allowed_inside`
    /// describes nothing.
    #[inline]
    fn is_void_element(tag_name: &str) -> bool {
        is_void_element(tag_name)
    }

    /// Whether this tag sits where its text is not markup: a code or math block,
    /// front matter, a comment, a link title, or a template shortcode tag.
    ///
    /// A shortcode tag is an argument list a template reads, so `<b>` written in
    /// one is a string the template receives rather than HTML the document emits.
    /// The range covers the tag only, so real HTML in the body between paired
    /// shortcode tags is still reported.
    fn is_inert_markup(ctx: &crate::lint_context::LintContext, html_tag: &HtmlTag) -> bool {
        ctx.line_info(html_tag.line).is_some_and(|info| {
            info.in_code_block
                || info.in_pymdown_block
                || info.is_kramdown_block_ial
                || info.in_front_matter
                || info.in_math_block
        }) || ctx.is_in_html_comment(html_tag.byte_offset)
            || ctx.is_in_mdx_comment(html_tag.byte_offset)
            || ctx.is_in_link_title(html_tag.byte_offset)
            || ctx.is_byte_offset_in_code_span(html_tag.byte_offset)
            || ctx.is_in_shortcode(html_tag.byte_offset)
    }

    /// Byte ranges spanned by the elements named in `allowed_inside`, each running
    /// from its opening tag through its matching closing tag.
    ///
    /// An element left unclosed reaches the end of the document, which is where a
    /// reader ends its content too.
    fn allowed_inside_ranges(&self, ctx: &crate::lint_context::LintContext) -> Vec<Range<usize>> {
        if self.allowed_inside.is_empty() {
            return Vec::new();
        }

        let html_tags = ctx.html_tags();
        let mut ranges = Vec::new();
        let mut open: Vec<(&str, usize)> = Vec::new();

        for html_tag in html_tags.iter() {
            if !self.allowed_inside.contains(&html_tag.tag_name) || Self::is_inert_markup(ctx, html_tag) {
                continue;
            }
            if html_tag.is_closing {
                // A closing tag ends the innermost element of that name, and with it
                // any element left open inside that one.
                if let Some(index) = open.iter().rposition(|(name, _)| *name == html_tag.tag_name) {
                    ranges.extend(open.drain(index..).map(|(_, start)| start..html_tag.byte_end));
                }
            } else if !html_tag.is_self_closing && !Self::is_void_element(&html_tag.tag_name) {
                open.push((html_tag.tag_name.as_str(), html_tag.byte_offset));
            }
        }

        ranges.extend(open.into_iter().map(|(_, start)| start..ctx.content.len()));
        ranges
    }

    /// Whether a byte offset falls in one of the ranges `allowed_inside` covers.
    #[inline]
    fn is_inside_allowed_element(ranges: &[Range<usize>], byte_offset: usize) -> bool {
        ranges.iter().any(|range| range.contains(&byte_offset))
    }

    /// Check if a tag is in the disallowed set (for disallowed-only mode).
    #[inline]
    fn is_tag_disallowed(&self, tag: &str) -> bool {
        Self::tag_in_set(&self.disallowed, tag)
    }

    /// Check if operating in disallowed-only mode
    #[inline]
    fn is_disallowed_mode(&self) -> bool {
        self.config.is_disallowed_mode()
    }

    // Check if a tag is an HTML comment
    #[inline]
    fn is_html_comment(&self, tag: &str) -> bool {
        tag.starts_with("<!--") && tag.ends_with("-->")
    }

    /// Check if a tag name is a valid HTML element or custom element.
    /// Returns false for placeholder syntax like `<NAME>`, `<resource>`, `<actual>`.
    ///
    /// Per HTML spec, custom elements must contain a hyphen (e.g., `<my-component>`).
    #[inline]
    fn is_html_element_or_custom(tag_name: &str) -> bool {
        // Sorted for binary search — must remain sorted when adding elements
        const HTML_ELEMENTS: &[&str] = &[
            "a",
            "abbr",
            "acronym",
            "address",
            "applet",
            "area",
            "article",
            "aside",
            "audio",
            "b",
            "base",
            "basefont",
            "bdi",
            "bdo",
            "big",
            "blockquote",
            "body",
            "br",
            "button",
            "canvas",
            "caption",
            "center",
            "cite",
            "code",
            "col",
            "colgroup",
            "data",
            "datalist",
            "dd",
            "del",
            "details",
            "dfn",
            "dialog",
            "dir",
            "div",
            "dl",
            "dt",
            "em",
            "embed",
            "fieldset",
            "figcaption",
            "figure",
            "font",
            "footer",
            "form",
            "frame",
            "frameset",
            "h1",
            "h2",
            "h3",
            "h4",
            "h5",
            "h6",
            "head",
            "header",
            "hgroup",
            "hr",
            "html",
            "i",
            "iframe",
            "img",
            "input",
            "ins",
            "isindex",
            "kbd",
            "label",
            "legend",
            "li",
            "link",
            "main",
            "map",
            "mark",
            "marquee",
            "math",
            "menu",
            "meta",
            "meter",
            "nav",
            "noembed",
            "noframes",
            "noscript",
            "object",
            "ol",
            "optgroup",
            "option",
            "output",
            "p",
            "param",
            "picture",
            "plaintext",
            "pre",
            "progress",
            "q",
            "rp",
            "rt",
            "ruby",
            "s",
            "samp",
            "script",
            "search",
            "section",
            "select",
            "slot",
            "small",
            "source",
            "span",
            "strike",
            "strong",
            "style",
            "sub",
            "summary",
            "sup",
            "svg",
            "table",
            "tbody",
            "td",
            "template",
            "textarea",
            "tfoot",
            "th",
            "thead",
            "time",
            "title",
            "tr",
            "track",
            "tt",
            "u",
            "ul",
            "var",
            "video",
            "wbr",
            "xmp",
        ];

        let lower = tag_name.to_ascii_lowercase();
        if HTML_ELEMENTS.binary_search(&lower.as_str()).is_ok() {
            return true;
        }
        // Custom elements must contain a hyphen per HTML spec
        tag_name.contains('-')
    }

    // Check if a tag is likely a programming type annotation rather than HTML
    #[inline]
    fn is_likely_type_annotation(&self, tag: &str) -> bool {
        // Sorted for binary search — must remain sorted when adding elements
        const COMMON_TYPES: &[&str] = &[
            "any",
            "apiresponse",
            "array",
            "bigint",
            "config",
            "data",
            "date",
            "e",
            "element",
            "error",
            "function",
            "generator",
            "item",
            "iterator",
            "k",
            "map",
            "node",
            "null",
            "number",
            "options",
            "params",
            "promise",
            "regexp",
            "request",
            "response",
            "result",
            "set",
            "string",
            "symbol",
            "t",
            "u",
            "undefined",
            "userdata",
            "v",
            "void",
            "weakmap",
            "weakset",
        ];

        let tag_content = tag
            .trim_start_matches('<')
            .trim_end_matches('>')
            .trim_start_matches('/');
        let tag_name = tag_content
            .split(|c: char| c.is_whitespace() || c == '>' || c == '/')
            .next()
            .unwrap_or("");

        // Check if it's a simple tag (no attributes) with a common type name
        if !tag_content.contains(' ') && !tag_content.contains('=') {
            let lower = tag_name.to_ascii_lowercase();
            COMMON_TYPES.binary_search(&lower.as_str()).is_ok()
        } else {
            false
        }
    }

    // Check if a tag is actually an email address in angle brackets
    #[inline]
    fn is_email_address(&self, tag: &str) -> bool {
        let content = tag.trim_start_matches('<').trim_end_matches('>');
        // Simple email pattern: contains @ and has reasonable structure
        content.contains('@')
            && content.chars().all(|c| c.is_alphanumeric() || "@.-_+".contains(c))
            && content.split('@').count() == 2
            && content.split('@').all(|part| !part.is_empty())
    }

    // Check if a tag has the markdown attribute (MkDocs/Material for MkDocs)
    #[inline]
    fn has_markdown_attribute(&self, tag: &str) -> bool {
        // Check for various forms of markdown attribute
        // Examples: <div markdown>, <div markdown="1">, <div class="result" markdown>
        tag.contains(" markdown>") || tag.contains(" markdown=") || tag.contains(" markdown ")
    }

    /// Check if a tag contains JSX-specific attributes that indicate it's JSX, not HTML
    /// JSX uses different attribute names than HTML:
    /// - `className` instead of `class`
    /// - `htmlFor` instead of `for`
    /// - camelCase event handlers (`onClick`, `onChange`, `onSubmit`, etc.)
    /// - JSX expression syntax `={...}` for dynamic values
    #[inline]
    fn has_jsx_attributes(tag: &str) -> bool {
        // JSX-specific attribute names (HTML uses class, for, onclick, etc.)
        tag.contains("className")
            || tag.contains("htmlFor")
            || tag.contains("dangerouslySetInnerHTML")
            // camelCase event handlers (JSX uses onClick, HTML uses onclick)
            || tag.contains("onClick")
            || tag.contains("onChange")
            || tag.contains("onSubmit")
            || tag.contains("onFocus")
            || tag.contains("onBlur")
            || tag.contains("onKeyDown")
            || tag.contains("onKeyUp")
            || tag.contains("onKeyPress")
            || tag.contains("onMouseDown")
            || tag.contains("onMouseUp")
            || tag.contains("onMouseEnter")
            || tag.contains("onMouseLeave")
            // JSX expression syntax: ={expression} or ={ expression }
            || tag.contains("={")
    }

    // Check if a tag is actually a URL in angle brackets
    #[inline]
    fn is_url_in_angle_brackets(&self, tag: &str) -> bool {
        let content = tag.trim_start_matches('<').trim_end_matches('>');
        // Check for common URL schemes
        content.starts_with("http://")
            || content.starts_with("https://")
            || content.starts_with("ftp://")
            || content.starts_with("ftps://")
            || content.starts_with("mailto:")
    }

    #[inline]
    fn is_relaxed_fix_mode(&self) -> bool {
        self.config.fix_mode == MD033FixMode::Relaxed
    }

    #[inline]
    fn is_droppable_attribute(&self, attr_name: &str) -> bool {
        // Event handler attributes (onclick, onload, etc.) are never droppable
        // because they can execute arbitrary JavaScript.
        if attr_name.starts_with("on") && attr_name.len() > 2 {
            return false;
        }
        self.drop_attributes.contains(attr_name)
            || (attr_name.starts_with("data-")
                && (self.drop_attributes.contains("data-*") || self.drop_attributes.contains("data-")))
    }

    #[inline]
    fn is_strippable_wrapper(&self, tag_name: &str) -> bool {
        self.is_relaxed_fix_mode() && self.strip_wrapper_elements.contains(tag_name)
    }

    /// Check whether `byte_offset` sits directly inside a top-level strippable
    /// wrapper element (e.g. `<p>`).  Returns `true` only when:
    ///  1. The nearest unclosed opening tag before the offset is a configured
    ///     wrapper element, AND
    ///  2. That wrapper is itself NOT nested inside another HTML element.
    ///
    /// Condition 2 prevents converting inner content when the wrapper cannot
    /// be stripped (e.g. `<div><p><img/></p></div>` -- stripping `<p>` is
    /// blocked because it is nested, so converting `<img>` would leave
    /// markdown inside an HTML block where it won't render).
    fn is_inside_strippable_wrapper(&self, content: &str, byte_offset: usize) -> bool {
        if byte_offset == 0 {
            return false;
        }
        let before = content[..byte_offset].trim_end();
        if !before.ends_with('>') || before.ends_with("->") {
            return false;
        }
        if let Some(last_lt) = before.rfind('<') {
            let potential_tag = &before[last_lt..];
            if potential_tag.starts_with("</") || potential_tag.starts_with("<!--") {
                return false;
            }
            let parent_name = potential_tag
                .trim_start_matches('<')
                .split(|c: char| c.is_whitespace() || c == '>' || c == '/')
                .next()
                .unwrap_or("")
                .to_lowercase();
            if !self.strip_wrapper_elements.contains(&parent_name) {
                return false;
            }
            // Verify the wrapper itself is not nested inside another element.
            let wrapper_before = before[..last_lt].trim_end();
            if wrapper_before.ends_with('>')
                && !wrapper_before.ends_with("->")
                && let Some(outer_lt) = wrapper_before.rfind('<')
                && let outer_tag = &wrapper_before[outer_lt..]
                && !outer_tag.starts_with("</")
                && !outer_tag.starts_with("<!--")
            {
                return false;
            }
            return true;
        }
        false
    }

    /// Convert paired HTML tags to their Markdown equivalents.
    /// Returns None if the tag cannot be safely converted (has nested tags, HTML entities, etc.)
    fn convert_to_markdown(tag_name: &str, inner_content: &str) -> Option<String> {
        // Skip if content contains nested HTML tags
        if inner_content.contains('<') {
            return None;
        }
        // Skip if content contains HTML entities (e.g., &vert;, &amp;, &lt;)
        // These need HTML context to render correctly; markdown won't process them
        if inner_content.contains('&') && inner_content.contains(';') {
            // Check for common HTML entity patterns
            let has_entity = inner_content
                .split('&')
                .skip(1)
                .any(|part| part.split(';').next().is_some_and(|e| !e.is_empty() && e.len() < 10));
            if has_entity {
                return None;
            }
        }
        match tag_name {
            "em" | "i" => Some(format!("*{inner_content}*")),
            "strong" | "b" => Some(format!("**{inner_content}**")),
            "code" => {
                // Handle backticks in content by using double backticks with padding
                if inner_content.contains('`') {
                    Some(format!("`` {inner_content} ``"))
                } else {
                    Some(format!("`{inner_content}`"))
                }
            }
            _ => None,
        }
    }

    /// Convert self-closing HTML tags to their Markdown equivalents.
    fn convert_self_closing_to_markdown(&self, tag_name: &str, opening_tag: &str) -> Option<String> {
        match tag_name {
            "br" => match self.config.br_style {
                md033_config::BrStyle::TrailingSpaces => Some("  \n".to_string()),
                md033_config::BrStyle::Backslash => Some("\\\n".to_string()),
            },
            "hr" => Some("\n---\n".to_string()),
            "img" => self.convert_img_to_markdown(opening_tag),
            _ => None,
        }
    }

    /// Parse all attributes from an HTML tag into a list of (name, value) pairs.
    /// This provides proper attribute parsing instead of naive string matching.
    fn parse_attributes(tag: &str) -> Vec<(String, Option<String>)> {
        let mut attrs = Vec::new();

        // Remove < and > and tag name
        let tag_content = tag.trim_start_matches('<').trim_end_matches('>').trim_end_matches('/');

        // Find first whitespace to skip tag name. Advance by the full UTF-8 width
        // of the whitespace char so multi-byte whitespace (e.g. U+00A0) does not
        // leave attr_start in the middle of a codepoint.
        let attr_start = tag_content
            .char_indices()
            .find(|(_, c)| c.is_whitespace())
            .map_or(tag_content.len(), |(i, c)| i + c.len_utf8());

        if attr_start >= tag_content.len() {
            return attrs;
        }

        let attr_str = &tag_content[attr_start..];
        let mut chars = attr_str.chars().peekable();

        while chars.peek().is_some() {
            // Skip whitespace
            while chars.peek().is_some_and(|c| c.is_whitespace()) {
                chars.next();
            }

            if chars.peek().is_none() {
                break;
            }

            // Read attribute name
            let mut attr_name = String::new();
            while let Some(&c) = chars.peek() {
                if c.is_whitespace() || c == '=' || c == '>' || c == '/' {
                    break;
                }
                attr_name.push(c);
                chars.next();
            }

            if attr_name.is_empty() {
                break;
            }

            // Skip whitespace before =
            while chars.peek().is_some_and(|c| c.is_whitespace()) {
                chars.next();
            }

            // Check for = and value
            if chars.peek() == Some(&'=') {
                chars.next(); // consume =

                // Skip whitespace after =
                while chars.peek().is_some_and(|c| c.is_whitespace()) {
                    chars.next();
                }

                // Read value
                let mut value = String::new();
                if let Some(&quote) = chars.peek() {
                    if quote == '"' || quote == '\'' {
                        chars.next(); // consume opening quote
                        for c in chars.by_ref() {
                            if c == quote {
                                break;
                            }
                            value.push(c);
                        }
                    } else {
                        // Unquoted value
                        while let Some(&c) = chars.peek() {
                            if c.is_whitespace() || c == '>' || c == '/' {
                                break;
                            }
                            value.push(c);
                            chars.next();
                        }
                    }
                }
                attrs.push((attr_name.to_ascii_lowercase(), Some(value)));
            } else {
                // Boolean attribute (no value)
                attrs.push((attr_name.to_ascii_lowercase(), None));
            }
        }

        attrs
    }

    /// Extract an HTML attribute value from a tag string.
    /// Handles double quotes, single quotes, and unquoted values.
    /// Returns None if the attribute is not found.
    fn extract_attribute(tag: &str, attr_name: &str) -> Option<String> {
        let attrs = Self::parse_attributes(tag);
        let attr_lower = attr_name.to_ascii_lowercase();

        attrs
            .into_iter()
            .find(|(name, _)| name == &attr_lower)
            .and_then(|(_, value)| value)
    }

    /// Check if an HTML tag has extra attributes beyond the specified allowed ones.
    /// Uses proper attribute parsing to avoid false positives from string matching.
    fn has_extra_attributes(&self, tag: &str, allowed_attrs: &[&str]) -> bool {
        let attrs = Self::parse_attributes(tag);

        // All event handlers (on*) are dangerous
        // Plus common attributes that would be lost in markdown conversion
        const DANGEROUS_ATTR_PREFIXES: &[&str] = &["on"]; // onclick, onload, onerror, etc.
        const DANGEROUS_ATTRS: &[&str] = &[
            "class",
            "id",
            "style",
            "target",
            "rel",
            "download",
            "referrerpolicy",
            "crossorigin",
            "loading",
            "decoding",
            "fetchpriority",
            "sizes",
            "srcset",
            "usemap",
            "ismap",
            "width",
            "height",
            "name",   // anchor names
            "data-*", // data attributes (checked separately)
        ];

        for (attr_name, _) in attrs {
            // Skip allowed attributes (list is small, linear scan is efficient)
            if allowed_attrs.iter().any(|a| a.to_ascii_lowercase() == attr_name) {
                continue;
            }

            if self.is_relaxed_fix_mode() {
                if self.is_droppable_attribute(&attr_name) {
                    continue;
                }
                return true;
            }

            // Check for event handlers (on*)
            for prefix in DANGEROUS_ATTR_PREFIXES {
                if attr_name.starts_with(prefix) && attr_name.len() > prefix.len() {
                    return true;
                }
            }

            // Check for data-* attributes
            if attr_name.starts_with("data-") {
                return true;
            }

            // Check for other dangerous attributes
            if DANGEROUS_ATTRS.contains(&attr_name.as_str()) {
                return true;
            }
        }

        false
    }

    /// Convert `<a href="url">text</a>` to `[text](url)` or `[text](url "title")`
    /// Returns None if conversion is not safe.
    fn convert_a_to_markdown(&self, opening_tag: &str, inner_content: &str) -> Option<String> {
        // Extract href attribute
        let href = Self::extract_attribute(opening_tag, "href")?;

        // Check URL is safe
        if !MD033Config::is_safe_url(&href) {
            return None;
        }

        // Check for nested HTML tags in content
        if inner_content.contains('<') {
            return None;
        }

        // Check for HTML entities that wouldn't render correctly in markdown
        if inner_content.contains('&') && inner_content.contains(';') {
            let has_entity = inner_content
                .split('&')
                .skip(1)
                .any(|part| part.split(';').next().is_some_and(|e| !e.is_empty() && e.len() < 10));
            if has_entity {
                return None;
            }
        }

        // Extract optional title attribute
        let title = Self::extract_attribute(opening_tag, "title");

        // Check for extra dangerous attributes (title is allowed)
        if self.has_extra_attributes(opening_tag, &["href", "title"]) {
            return None;
        }

        // If inner content is exactly a markdown image (from a prior <img> fix),
        // use it directly without bracket escaping to produce valid [![alt](src)](href).
        // Must verify the entire content is a single image — not mixed content like
        // "![](url) extra [text]" where trailing brackets still need escaping.
        let trimmed_inner = inner_content.trim();
        let is_markdown_image =
            trimmed_inner.starts_with("![") && trimmed_inner.contains("](") && trimmed_inner.ends_with(')') && {
                // Verify the closing ](url) accounts for the rest of the content
                // by finding the image's ]( and checking nothing follows the final )
                if let Some(bracket_close) = trimmed_inner.rfind("](") {
                    let after_paren = &trimmed_inner[bracket_close + 2..];
                    // The rest should be just "url)" — find the matching close paren
                    after_paren.ends_with(')')
                        && after_paren.chars().filter(|&c| c == ')').count()
                            >= after_paren.chars().filter(|&c| c == '(').count()
                } else {
                    false
                }
            };
        let escaped_text = if is_markdown_image {
            trimmed_inner.to_string()
        } else {
            // Escape special markdown characters in link text
            // Brackets need escaping to avoid breaking the link syntax
            inner_content.replace('[', r"\[").replace(']', r"\]")
        };

        // Escape parentheses in URL
        let escaped_url = href.replace('(', "%28").replace(')', "%29");

        // Format with or without title
        if let Some(title_text) = title {
            // Escape quotes in title
            let escaped_title = title_text.replace('"', r#"\""#);
            Some(format!("[{escaped_text}]({escaped_url} \"{escaped_title}\")"))
        } else {
            Some(format!("[{escaped_text}]({escaped_url})"))
        }
    }

    /// Convert `<img src="url" alt="text">` to `![alt](src)` or `![alt](src "title")`
    /// Returns None if conversion is not safe.
    fn convert_img_to_markdown(&self, tag: &str) -> Option<String> {
        // Extract src attribute (required)
        let src = Self::extract_attribute(tag, "src")?;

        // Check URL is safe
        if !MD033Config::is_safe_url(&src) {
            return None;
        }

        // Extract alt attribute (optional, default to empty)
        let alt = Self::extract_attribute(tag, "alt").unwrap_or_default();

        // Extract optional title attribute
        let title = Self::extract_attribute(tag, "title");

        // Check for extra dangerous attributes (title is allowed)
        if self.has_extra_attributes(tag, &["src", "alt", "title"]) {
            return None;
        }

        // Escape special markdown characters in alt text
        let escaped_alt = alt.replace('[', r"\[").replace(']', r"\]");

        // Escape parentheses in URL
        let escaped_url = src.replace('(', "%28").replace(')', "%29");

        // Format with or without title
        if let Some(title_text) = title {
            // Escape quotes in title
            let escaped_title = title_text.replace('"', r#"\""#);
            Some(format!("![{escaped_alt}]({escaped_url} \"{escaped_title}\")"))
        } else {
            Some(format!("![{escaped_alt}]({escaped_url})"))
        }
    }

    /// Check if an HTML tag has attributes that would make conversion unsafe
    fn has_significant_attributes(opening_tag: &str) -> bool {
        // Tags with just whitespace or empty are fine
        let tag_content = opening_tag
            .trim_start_matches('<')
            .trim_end_matches('>')
            .trim_end_matches('/');

        // Split by whitespace; if there's more than the tag name, it has attributes
        let parts: Vec<&str> = tag_content.split_whitespace().collect();
        parts.len() > 1
    }

    /// Check if a tag appears to be nested inside another HTML element
    /// by looking at the surrounding context (e.g., `<code><em>text</em></code>`)
    fn is_nested_in_html(content: &str, tag_byte_start: usize, tag_byte_end: usize) -> bool {
        // Check if there's a `>` immediately before this tag (indicating inside another element)
        if tag_byte_start > 0 {
            let before = &content[..tag_byte_start];
            let before_trimmed = before.trim_end();
            if before_trimmed.ends_with('>') && !before_trimmed.ends_with("->") {
                // Check it's not a closing tag or comment
                if let Some(last_lt) = before_trimmed.rfind('<') {
                    let potential_tag = &before_trimmed[last_lt..];
                    // Skip if it's a closing tag (</...>) or comment (<!--)
                    if !potential_tag.starts_with("</") && !potential_tag.starts_with("<!--") {
                        return true;
                    }
                }
            }
        }
        // Check if there's a `<` immediately after the closing tag (indicating inside another element)
        if tag_byte_end < content.len() {
            let after = &content[tag_byte_end..];
            let after_trimmed = after.trim_start();
            if after_trimmed.starts_with("</") {
                return true;
            }
        }
        false
    }

    /// Whether the tag at `tag_byte_start` opens the HTML block its line starts:
    /// it is the first thing on the line, the line above is outside any HTML
    /// block, and its name is not one of the block-level elements (CommonMark
    /// start condition 7, a complete tag alone on its line).
    fn opens_an_unnamed_block(ctx: &crate::lint_context::LintContext, line_num: usize, tag_byte_start: usize) -> bool {
        let Some(info) = ctx.line_info(line_num) else {
            return false;
        };
        let line = info.content(ctx.content);
        let trimmed = line.trim_start();
        info.byte_offset + (line.len() - trimmed.len()) == tag_byte_start
            && !(line_num > 1 && ctx.is_in_html_block(line_num - 1))
            && crate::utils::html_block::parse_html_block_start(trimmed).is_none()
    }

    /// Calculate fix to remove HTML tags while keeping content.
    ///
    /// For self-closing tags like `<br/>`, returns a single fix to remove the tag.
    /// For paired tags like `<span>text</span>`, returns the replacement text (just the content).
    ///
    /// Returns (range, replacement_text) where range is the bytes to replace
    /// and replacement_text is what to put there (content without tags, or empty for self-closing).
    ///
    /// When `in_html_block` is true, returns None in conservative mode.  In
    /// relaxed mode two exceptions apply:
    /// - Strippable wrapper elements (e.g. `<p>`) bypass the block guard so
    ///   they can be stripped even though they ARE the HTML block.
    /// - Self-closing tags whose direct parent is a strippable wrapper also
    ///   bypass the guard so inner content can be converted first.
    fn calculate_fix(
        &self,
        content: &str,
        closing_tags: &ClosingTagIndex,
        opening_tag: &str,
        tag_byte_start: usize,
        in_html_block: bool,
    ) -> Option<(std::ops::Range<usize>, String)> {
        // Every fix below is opt-in.
        if !self.config.fix {
            return None;
        }

        // Extract tag name from opening tag
        let tag_name = opening_tag
            .trim_start_matches('<')
            .split(|c: char| c.is_whitespace() || c == '>' || c == '/')
            .next()?
            .to_lowercase();

        // Check if it's a self-closing tag (ends with /> or is a void element like <br>)
        let is_self_closing =
            opening_tag.ends_with("/>") || matches!(tag_name.as_str(), "br" | "hr" | "img" | "input" | "meta" | "link");

        if is_self_closing {
            // When fix is enabled, try to convert to Markdown equivalent.
            // Skip tags inside HTML blocks (would break structure), UNLESS we
            // are in relaxed mode and the containing block is a strippable
            // wrapper -- this lets the inner element be converted first so the
            // wrapper can be stripped on a subsequent pass.
            let block_ok = !in_html_block
                || (self.is_relaxed_fix_mode() && self.is_inside_strippable_wrapper(content, tag_byte_start));
            if self.config.fix
                && MD033Config::is_safe_fixable_tag(&tag_name)
                && block_ok
                && let Some(markdown) = self.convert_self_closing_to_markdown(&tag_name, opening_tag)
            {
                return Some((tag_byte_start..tag_byte_start + opening_tag.len(), markdown));
            }
            // Can't convert this self-closing tag to Markdown, don't provide a fix
            // (e.g., <input>, <meta> - these have no Markdown equivalent without the new img support)
            return None;
        }

        // The first closing tag after the opening tag (case-insensitive)
        let search_start = tag_byte_start + opening_tag.len();
        if let Some(closing_byte_start) = closing_tags.first_at_or_after(&tag_name, search_start) {
            let closing_byte_end = closing_byte_start + "</>".len() + tag_name.len();

            // Extract the content between tags
            let inner_content = &content[search_start..closing_byte_start];

            // In relaxed mode, check wrapper stripping BEFORE the in_html_block
            // guard because the wrapper element itself IS the HTML block. We only
            // strip when:
            //  - the wrapper is not nested inside another HTML element
            //  - the inner content no longer contains HTML tags (prevents
            //    overlapping byte-range replacements within a single fix pass)
            if self.config.fix && self.is_strippable_wrapper(&tag_name) {
                if Self::is_nested_in_html(content, tag_byte_start, closing_byte_end) {
                    return None;
                }
                if inner_content.contains('<') {
                    return None;
                }
                return Some((tag_byte_start..closing_byte_end, inner_content.trim().to_string()));
            }

            // Skip auto-fix if inside an HTML block (like <pre>, <div>, etc.)
            // Converting tags inside HTML blocks would break the intended structure
            if in_html_block {
                return None;
            }

            // Skip auto-fix if this tag is nested inside another HTML element
            // e.g., <code><em>text</em></code> - don't convert the inner <em>
            if Self::is_nested_in_html(content, tag_byte_start, closing_byte_end) {
                return None;
            }

            // When fix is enabled and tag is safe to convert, try markdown conversion
            if self.config.fix && MD033Config::is_safe_fixable_tag(&tag_name) {
                // Handle <a> tags specially - they require attribute extraction
                if tag_name == "a" {
                    if let Some(markdown) = self.convert_a_to_markdown(opening_tag, inner_content) {
                        return Some((tag_byte_start..closing_byte_end, markdown));
                    }
                    // convert_a_to_markdown returned None - unsafe URL, nested HTML, etc.
                    return None;
                }

                // For simple tags (em, strong, code, etc.) - no attributes allowed
                if Self::has_significant_attributes(opening_tag) {
                    // Don't provide a fix for tags with attributes
                    // User may want to keep the attributes, so leave as-is
                    return None;
                }
                if let Some(markdown) = Self::convert_to_markdown(&tag_name, inner_content) {
                    return Some((tag_byte_start..closing_byte_end, markdown));
                }
                // convert_to_markdown returned None, meaning content has nested tags or
                // HTML entities that shouldn't be converted - leave as-is
                return None;
            }

            // For non-fixable tags, don't provide a fix
            // (e.g., <div>content</div>, <span>text</span>)
            return None;
        }

        // If no closing tag found, don't provide a fix (malformed HTML)
        None
    }
}

impl Rule for MD033NoInlineHtml {
    fn name(&self) -> &'static str {
        "MD033"
    }

    fn description(&self) -> &'static str {
        "Inline HTML is not allowed"
    }

    fn check(&self, ctx: &crate::lint_context::LintContext) -> LintResult {
        let content = ctx.content;

        // Early return: if no HTML tags at all, skip processing
        if content.is_empty() || !ctx.likely_has_html() {
            return Ok(Vec::new());
        }

        // Quick check for HTML tag pattern before expensive processing
        if !HTML_TAG_QUICK_CHECK.is_match(content) {
            return Ok(Vec::new());
        }

        let mut warnings = Vec::new();

        // Use centralized HTML parser to get all HTML tags (including multiline)
        let html_tags = ctx.html_tags();

        // Disallowed-only mode is a denylist, which the allowlists take no part in.
        let allowed_inside_ranges = if self.is_disallowed_mode() {
            Vec::new()
        } else {
            self.allowed_inside_ranges(ctx)
        };

        let closing_tags = if self.config.fix {
            ClosingTagIndex::new(content)
        } else {
            ClosingTagIndex::default()
        };

        for html_tag in html_tags.iter() {
            // Skip closing tags (only warn on opening tags)
            if html_tag.is_closing {
                continue;
            }

            let line_num = html_tag.line;
            let tag_byte_start = html_tag.byte_offset;

            // Reconstruct tag string from byte offsets
            let tag = &content[html_tag.byte_offset..html_tag.byte_end];

            // Skip tags whose text is not markup: code and math blocks, PyMdown
            // blocks, block IALs, front matter, comments, code spans, and the titles
            // of link reference definitions.
            if Self::is_inert_markup(ctx, html_tag) {
                continue;
            }

            // Skip HTML comments themselves
            if self.is_html_comment(tag) {
                continue;
            }

            // Skip JSX components in MDX files (e.g., <Chart />, <MyComponent>)
            if ctx.flavor.supports_jsx() && html_tag.tag_name.chars().next().is_some_and(char::is_uppercase) {
                continue;
            }

            // Skip JSX fragments in MDX files (<> and </>)
            if ctx.flavor.supports_jsx() && (html_tag.tag_name.is_empty() || tag == "<>" || tag == "</>") {
                continue;
            }

            // Skip elements with JSX-specific attributes in MDX files
            // e.g., <div className="...">, <button onClick={handler}>
            if ctx.flavor.supports_jsx() && Self::has_jsx_attributes(tag) {
                continue;
            }

            // Skip non-HTML elements (placeholder syntax like <NAME>, <resource>)
            if !Self::is_html_element_or_custom(&html_tag.tag_name) {
                continue;
            }

            // Skip likely programming type annotations
            if self.is_likely_type_annotation(tag) {
                continue;
            }

            // Skip email addresses in angle brackets
            if self.is_email_address(tag) {
                continue;
            }

            // Skip URLs in angle brackets
            if self.is_url_in_angle_brackets(tag) {
                continue;
            }

            // Determine whether to report this tag based on mode:
            // - Disallowed mode: only report tags in the disallowed list
            // - Default mode: report all tags except those in the allowed list or
            //   inside an element named by `allowed_inside`, with `table_allowed`
            //   taking precedence inside GFM table cells.
            if self.is_disallowed_mode() {
                if !self.is_tag_disallowed(tag) {
                    continue;
                }
            } else if ctx.is_in_table_block(line_num) {
                // An explicit table allowlist is the whole answer for a table cell,
                // so it decides there instead of the surrounding element.
                let inside_allowed_element = self.config.table_allowed_elements.is_none()
                    && Self::is_inside_allowed_element(&allowed_inside_ranges, tag_byte_start);
                if self.is_tag_allowed_in_table(tag, ctx.flavor) || inside_allowed_element {
                    continue;
                }
            } else if self.is_tag_allowed(tag, ctx.flavor)
                || Self::is_inside_allowed_element(&allowed_inside_ranges, tag_byte_start)
            {
                continue;
            }

            // Skip tags with markdown attribute in MkDocs mode
            if ctx.flavor == crate::config::MarkdownFlavor::MkDocs && self.has_markdown_attribute(tag) {
                continue;
            }

            // Check if we're inside an HTML block (like <pre>, <div>, etc.). A tag
            // alone on its line that opens a block with no block-level name (such
            // as `<img src="x.png">`) is the block itself, not part of an
            // enclosing structure a conversion could break.
            let in_html_block =
                ctx.is_in_html_block(line_num) && !Self::opens_an_unnamed_block(ctx, line_num, tag_byte_start);

            // Calculate fix to remove HTML tags but keep content
            let fix = self
                .calculate_fix(content, &closing_tags, tag, tag_byte_start, in_html_block)
                .map(|(range, replacement)| Fix::new(range, replacement));

            // Calculate actual end line and column for multiline tags
            // Use byte_end - 1 to get the last character position of the tag
            let (end_line, end_col) = if html_tag.byte_end > 0 {
                ctx.offset_to_line_col(html_tag.byte_end - 1)
            } else {
                (line_num, html_tag.end_col + 1)
            };

            // Report the HTML tag
            warnings.push(LintWarning {
                rule_name: Some(self.name().to_string()),
                line: line_num,
                column: html_tag.start_col + 1, // Convert to 1-indexed
                end_line,                       // Actual end line for multiline tags
                end_column: end_col + 1,        // Actual end column
                message: format!("Inline HTML found: {tag}"),
                severity: Severity::Warning,
                fix,
            });
        }

        Ok(warnings)
    }

    fn fix(&self, ctx: &crate::lint_context::LintContext) -> Result<String, LintError> {
        // Auto-fix is opt-in: only apply if explicitly enabled in config
        if !self.config.fix {
            return Ok(ctx.content.to_string());
        }

        // Get warnings with their inline fixes
        let warnings = self.check(ctx)?;
        let warnings =
            crate::utils::fix_utils::filter_warnings_by_inline_config(warnings, ctx.inline_config(), self.name());

        // If no warnings with fixes, return original content
        if warnings.is_empty() || !warnings.iter().any(|w| w.fix.is_some()) {
            return Ok(ctx.content.to_string());
        }

        // Collect all fixes and sort by range start (descending) to apply from end to beginning
        let mut fixes: Vec<_> = warnings
            .iter()
            .filter_map(|w| w.fix.as_ref().map(|f| (f.range.start, f.range.end, &f.replacement)))
            .collect();
        fixes.sort_by_key(|f| std::cmp::Reverse(f.0));

        // Apply fixes from end to beginning to preserve byte offsets
        let mut result = ctx.content.to_string();
        for (start, end, replacement) in fixes {
            if start < result.len() && end <= result.len() && start <= end {
                result.replace_range(start..end, replacement);
            }
        }

        Ok(result)
    }

    fn fix_capability(&self) -> crate::rule::FixCapability {
        if self.config.fix {
            crate::rule::FixCapability::FullyFixable
        } else {
            crate::rule::FixCapability::Unfixable
        }
    }

    /// Get the category of this rule for selective processing
    fn category(&self) -> RuleCategory {
        RuleCategory::Html
    }

    /// Check if this rule should be skipped
    fn should_skip(&self, ctx: &crate::lint_context::LintContext) -> bool {
        ctx.content.is_empty() || !ctx.likely_has_html()
    }

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

    crate::impl_rule_config_methods!(MD033Config);

    fn config_aliases(&self) -> Option<std::collections::HashMap<String, String>> {
        let mut aliases = std::collections::HashMap::new();
        // Shorthand aliases for allowed-elements/disallowed-elements
        aliases.insert("allowed".to_string(), "allowed-elements".to_string());
        aliases.insert("disallowed".to_string(), "disallowed-elements".to_string());
        // Documented in docs/md033.md; kept in step with the `alias` attributes on
        // `MD033Config::table_allowed_elements`.
        aliases.insert("table-allowed".to_string(), "table-allowed-elements".to_string());
        aliases.insert("table_allowed".to_string(), "table-allowed-elements".to_string());
        Some(aliases)
    }
}

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

    #[test]
    fn closing_tag_index_finds_what_a_case_insensitive_search_finds() {
        let documents = [
            "<em>a</em> b </EM> </Em>",
            "</</em> <//em> </ em> </em > </em/> </>",
            "x</em",
            "</em></em></strong></em>",
            "</span-x></span></span:y>",
            "</é></É></em>",
            "",
        ];
        for content in documents {
            let index = ClosingTagIndex::new(content);
            for tag_name in ["em", "strong", "span", "span-x", "span:y", "é"] {
                let needle = format!("</{tag_name}>");
                let lowered = content.to_ascii_lowercase();
                for offset in 0..=content.len() {
                    if !content.is_char_boundary(offset) {
                        continue;
                    }
                    let expected = lowered[offset..].find(&needle).map(|pos| offset + pos);
                    assert_eq!(
                        index.first_at_or_after(tag_name, offset),
                        expected,
                        "{tag_name:?} from {offset} in {content:?}"
                    );
                }
            }
        }
    }

    #[test]
    fn check_without_fix_enabled_offers_no_fixes() {
        let content = "Text <em>x</em> and <br/> and <p>y</p>\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let warnings = MD033NoInlineHtml::default().check(&ctx).unwrap();
        assert!(!warnings.is_empty());
        assert!(warnings.iter().all(|w| w.fix.is_none()));
    }

    fn relaxed_fix_rule() -> MD033NoInlineHtml {
        let config = MD033Config {
            fix: true,
            fix_mode: MD033FixMode::Relaxed,
            ..MD033Config::default()
        };
        MD033NoInlineHtml::from_config_struct(config)
    }

    #[test]
    fn test_md033_basic_html() {
        let rule = MD033NoInlineHtml::default();
        let content = "<div>Some content</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Only reports opening tags, not closing tags
        assert_eq!(result.len(), 1); // Only <div>, not </div>
        assert!(result[0].message.starts_with("Inline HTML found: <div>"));
    }

    #[test]
    fn test_md033_ignores_backslash_escaped_custom_element() {
        let rule = MD033NoInlineHtml::default();
        let content = r"Press \<x-keyboard> to continue, but flag <x-keyboard>.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);

        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].column, 43);
        assert_eq!(result[0].message, "Inline HTML found: <x-keyboard>");
    }

    #[test]
    fn test_md033_front_matter() {
        let rule = MD033NoInlineHtml::default();
        let content = "---\ndescription: <div class=\"test\">hello</div>\n---\n# Title\n<div>body</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Should only report <div>body</div> (line 5), not <div class="test"> (line 2)
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].line, 5);
        assert_eq!(result[0].message, "Inline HTML found: <div>");
    }

    #[test]
    fn test_md033_math_block() {
        let rule = MD033NoInlineHtml::default();
        let content = "$$\nx < y && y > z\n$$\n<div>body</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Should only report <div>body</div> (line 4), not the fake tag <y> in math (line 2)
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].line, 4);
    }

    #[test]
    fn test_md033_case_insensitive() {
        let rule = MD033NoInlineHtml::default();
        let content = "<DiV>Some <B>content</B></dIv>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Only reports opening tags, not closing tags
        assert_eq!(result.len(), 2); // <DiV>, <B> (not </B>, </dIv>)
        assert_eq!(result[0].message, "Inline HTML found: <DiV>");
        assert_eq!(result[1].message, "Inline HTML found: <B>");
    }

    #[test]
    fn test_md033_multibyte_whitespace_in_tag_does_not_panic() {
        // A non-ASCII whitespace (U+00A0 NO-BREAK SPACE) before the attributes
        // must not cause a non-char-boundary slice panic while parsing attributes.
        let rule = relaxed_fix_rule();
        let content = "<img\u{00A0}src=\"test.png\" alt=\"x\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        // check() and fix() both reach parse_attributes; neither may panic.
        let _ = rule.check(&ctx).unwrap();
        let _ = rule.fix(&ctx).unwrap();
    }

    #[test]
    fn test_md033_allowed_tags() {
        let rule = MD033NoInlineHtml::with_allowed(vec!["div".to_string(), "br".to_string()]);
        let content = "<div>Allowed</div><p>Not allowed</p><br/>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Only warnings for non-allowed opening tags (<p> only, div and br are allowed)
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].message, "Inline HTML found: <p>");

        // Test case-insensitivity of allowed tags
        let content2 = "<DIV>Allowed</DIV><P>Not allowed</P><BR/>";
        let ctx2 = LintContext::new(content2, crate::config::MarkdownFlavor::Standard, None);
        let result2 = rule.check(&ctx2).unwrap();
        assert_eq!(result2.len(), 1); // Only <P> flagged
        assert_eq!(result2[0].message, "Inline HTML found: <P>");
    }

    #[test]
    fn test_md033_html_comments() {
        let rule = MD033NoInlineHtml::default();
        let content = "<!-- This is a comment --> <p>Not a comment</p>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Should detect warnings for HTML opening tags (comments are skipped, closing tags not reported)
        assert_eq!(result.len(), 1); // Only <p>
        assert_eq!(result[0].message, "Inline HTML found: <p>");
    }

    #[test]
    fn test_md033_tags_in_links() {
        let rule = MD033NoInlineHtml::default();
        let content = "[Link](http://example.com/<div>)";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // The <div> in the URL should be detected as HTML (not skipped)
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].message, "Inline HTML found: <div>");

        let content2 = "[Link <a>text</a>](url)";
        let ctx2 = LintContext::new(content2, crate::config::MarkdownFlavor::Standard, None);
        let result2 = rule.check(&ctx2).unwrap();
        // Only reports opening tags
        assert_eq!(result2.len(), 1); // Only <a>
        assert_eq!(result2[0].message, "Inline HTML found: <a>");
    }

    #[test]
    fn test_md033_fix_escaping() {
        let rule = MD033NoInlineHtml::default();
        let content = "Text with <div> and <br/> tags.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed_content = rule.fix(&ctx).unwrap();
        // No fix for HTML tags; output should be unchanged
        assert_eq!(fixed_content, content);
    }

    #[test]
    fn test_md033_in_code_blocks() {
        let rule = MD033NoInlineHtml::default();
        let content = "```html\n<div>Code</div>\n```\n<div>Not code</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Only reports opening tags outside code block
        assert_eq!(result.len(), 1); // Only <div> outside code block
        assert_eq!(result[0].message, "Inline HTML found: <div>");
    }

    #[test]
    fn test_md033_in_code_spans() {
        let rule = MD033NoInlineHtml::default();
        let content = "Text with `<p>in code</p>` span. <br/> Not in span.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Should detect <br/> outside code span, but not tags inside code span
        assert_eq!(result.len(), 1);
        assert_eq!(result[0].message, "Inline HTML found: <br/>");
    }

    #[test]
    fn test_md033_issue_90_code_span_with_diff_block() {
        // Test for issue #90: inline code span followed by diff code block
        let rule = MD033NoInlineHtml::default();
        let content = r#"# Heading

`<env>`

```diff
- this
+ that
```"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        // Should NOT detect <env> as HTML since it's inside backticks
        assert_eq!(result.len(), 0, "Should not report HTML tags inside code spans");
    }

    #[test]
    fn test_md033_multiple_code_spans_with_angle_brackets() {
        // Test multiple code spans on same line
        let rule = MD033NoInlineHtml::default();
        let content = "`<one>` and `<two>` and `<three>` are all code spans";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 0, "Should not report HTML tags inside any code spans");
    }

    #[test]
    fn test_md033_nested_angle_brackets_in_code_span() {
        // Test nested angle brackets
        let rule = MD033NoInlineHtml::default();
        let content = "Text with `<<nested>>` brackets";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 0, "Should handle nested angle brackets in code spans");
    }

    #[test]
    fn test_md033_code_span_at_end_before_code_block() {
        // Test code span at end of line before code block
        let rule = MD033NoInlineHtml::default();
        let content = "Testing `<test>`\n```\ncode here\n```";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 0, "Should handle code span before code block");
    }

    #[test]
    fn test_md033_quick_fix_inline_tag() {
        // Test that non-fixable tags (like <span>) do NOT get a fix
        // Only safe fixable tags (em, i, strong, b, code, br, hr) with fix=true get fixes
        let rule = MD033NoInlineHtml::default();
        let content = "This has <span>inline text</span> that should keep content.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        // <span> is NOT a safe fixable tag, so no fix should be provided
        assert!(
            result[0].fix.is_none(),
            "Non-fixable tags like <span> should not have a fix"
        );
    }

    #[test]
    fn test_md033_quick_fix_multiline_tag() {
        // HTML block elements like <div> are intentionally NOT auto-fixed
        // Removing them would change document structure significantly
        let rule = MD033NoInlineHtml::default();
        let content = "<div>\nBlock content\n</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        // HTML block elements should NOT have auto-fix
        assert!(result[0].fix.is_none(), "HTML block elements should NOT have auto-fix");
    }

    #[test]
    fn test_md033_quick_fix_self_closing_tag() {
        // Test that self-closing tags with fix=false (default) do NOT get a fix
        let rule = MD033NoInlineHtml::default();
        let content = "Self-closing: <br/>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        // Default config has fix=false, so no fix should be provided
        assert!(
            result[0].fix.is_none(),
            "Self-closing tags should not have a fix when fix config is false"
        );
    }

    #[test]
    fn test_md033_quick_fix_multiple_tags() {
        // Test that multiple tags without fix=true do NOT get fixes
        // <span> is not a safe fixable tag, <strong> is but fix=false by default
        let rule = MD033NoInlineHtml::default();
        let content = "<span>first</span> and <strong>second</strong>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 2, "Should find two HTML tags");
        // Neither should have a fix: <span> is not fixable, <strong> is but fix=false
        assert!(result[0].fix.is_none(), "Non-fixable <span> should not have a fix");
        assert!(
            result[1].fix.is_none(),
            "<strong> should not have a fix when fix config is false"
        );
    }

    #[test]
    fn test_md033_skip_angle_brackets_in_link_titles() {
        // Angle brackets inside link reference definition titles should not be flagged as HTML
        let rule = MD033NoInlineHtml::default();
        let content = r#"# Test

[example]: <https://example.com> "Title with <Angle Brackets> inside"

Regular text with <div>content</div> HTML tag.
"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should only flag <div>, not <Angle Brackets> in the title (not a valid HTML element)
        // Opening tag only (markdownlint behavior)
        assert_eq!(result.len(), 1, "Should find opening div tag");
        assert!(
            result[0].message.contains("<div>"),
            "Should flag <div>, got: {}",
            result[0].message
        );
    }

    #[test]
    fn test_md033_skip_angle_brackets_in_link_title_single_quotes() {
        // Test with single-quoted title
        let rule = MD033NoInlineHtml::default();
        let content = r#"[ref]: url 'Title <Help Wanted> here'

<span>text</span> here
"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // <Help Wanted> is not a valid HTML element, so only <span> is flagged
        // Opening tag only (markdownlint behavior)
        assert_eq!(result.len(), 1, "Should find opening span tag");
        assert!(
            result[0].message.contains("<span>"),
            "Should flag <span>, got: {}",
            result[0].message
        );
    }

    #[test]
    fn test_md033_multiline_tag_end_line_calculation() {
        // Test that multiline HTML tags report correct end_line
        let rule = MD033NoInlineHtml::default();
        let content = "<div\n  class=\"test\"\n  id=\"example\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        // Tag starts on line 1
        assert_eq!(result[0].line, 1, "Start line should be 1");
        // Tag ends on line 3 (where the closing > is)
        assert_eq!(result[0].end_line, 3, "End line should be 3");
    }

    #[test]
    fn test_md033_single_line_tag_same_start_end_line() {
        // Test that single-line HTML tags have same start and end line
        let rule = MD033NoInlineHtml::default();
        let content = "Some text <div class=\"test\"> more text";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        assert_eq!(result[0].line, 1, "Start line should be 1");
        assert_eq!(result[0].end_line, 1, "End line should be 1 for single-line tag");
    }

    #[test]
    fn test_md033_multiline_tag_with_many_attributes() {
        // Test multiline tag spanning multiple lines
        let rule = MD033NoInlineHtml::default();
        let content =
            "Text\n<div\n  data-attr1=\"value1\"\n  data-attr2=\"value2\"\n  data-attr3=\"value3\">\nMore text";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        // Tag starts on line 2 (first line is "Text")
        assert_eq!(result[0].line, 2, "Start line should be 2");
        // Tag ends on line 5 (where the closing > is)
        assert_eq!(result[0].end_line, 5, "End line should be 5");
    }

    #[test]
    fn test_md033_disallowed_mode_basic() {
        // Test disallowed mode: only flags tags in the disallowed list
        let rule = MD033NoInlineHtml::with_disallowed(vec!["script".to_string(), "iframe".to_string()]);
        let content = "<div>Safe content</div><script>alert('xss')</script>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should only flag <script>, not <div>
        assert_eq!(result.len(), 1, "Should only flag disallowed tags");
        assert!(result[0].message.contains("<script>"), "Should flag script tag");
    }

    #[test]
    fn test_md033_disallowed_gfm_security_tags() {
        // Test GFM security tags expansion
        let rule = MD033NoInlineHtml::with_disallowed(vec!["gfm".to_string()]);
        let content = r#"
<div>Safe</div>
<title>Bad title</title>
<textarea>Bad textarea</textarea>
<style>.bad{}</style>
<iframe src="evil"></iframe>
<script>evil()</script>
<plaintext>old tag</plaintext>
<span>Safe span</span>
"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should flag: title, textarea, style, iframe, script, plaintext
        // Should NOT flag: div, span
        assert_eq!(result.len(), 6, "Should flag 6 GFM security tags");

        let flagged_tags: Vec<&str> = result
            .iter()
            .filter_map(|w| w.message.split('<').nth(1))
            .filter_map(|s| s.split('>').next())
            .filter_map(|s| s.split_whitespace().next())
            .collect();

        assert!(flagged_tags.contains(&"title"), "Should flag title");
        assert!(flagged_tags.contains(&"textarea"), "Should flag textarea");
        assert!(flagged_tags.contains(&"style"), "Should flag style");
        assert!(flagged_tags.contains(&"iframe"), "Should flag iframe");
        assert!(flagged_tags.contains(&"script"), "Should flag script");
        assert!(flagged_tags.contains(&"plaintext"), "Should flag plaintext");
        assert!(!flagged_tags.contains(&"div"), "Should NOT flag div");
        assert!(!flagged_tags.contains(&"span"), "Should NOT flag span");
    }

    #[test]
    fn test_md033_disallowed_case_insensitive() {
        // Test that disallowed check is case-insensitive
        let rule = MD033NoInlineHtml::with_disallowed(vec!["script".to_string()]);
        let content = "<SCRIPT>alert('xss')</SCRIPT><Script>alert('xss')</Script>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should flag both <SCRIPT> and <Script>
        assert_eq!(result.len(), 2, "Should flag both case variants");
    }

    #[test]
    fn test_md033_disallowed_with_attributes() {
        // Test that disallowed mode works with tags that have attributes
        let rule = MD033NoInlineHtml::with_disallowed(vec!["iframe".to_string()]);
        let content = r#"<iframe src="https://evil.com" width="100" height="100"></iframe>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should flag iframe with attributes");
        assert!(result[0].message.contains("iframe"), "Should flag iframe");
    }

    #[test]
    fn test_md033_disallowed_all_gfm_tags() {
        // Verify all GFM disallowed tags are covered
        use md033_config::GFM_DISALLOWED_TAGS;
        let rule = MD033NoInlineHtml::with_disallowed(vec!["gfm".to_string()]);

        for tag in GFM_DISALLOWED_TAGS {
            let content = format!("<{tag}>content</{tag}>");
            let ctx = LintContext::new(&content, crate::config::MarkdownFlavor::Standard, None);
            let result = rule.check(&ctx).unwrap();

            assert_eq!(result.len(), 1, "GFM tag <{tag}> should be flagged");
        }
    }

    #[test]
    fn test_md033_disallowed_mixed_with_custom() {
        // Test mixing "gfm" with custom disallowed tags
        let rule = MD033NoInlineHtml::with_disallowed(vec![
            "gfm".to_string(),
            "marquee".to_string(), // Custom disallowed tag
        ]);
        let content = r#"<script>bad</script><marquee>annoying</marquee><div>ok</div>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should flag script (gfm) and marquee (custom)
        assert_eq!(result.len(), 2, "Should flag both gfm and custom tags");
    }

    #[test]
    fn test_md033_disallowed_empty_means_default_mode() {
        // Empty disallowed list means default mode (flag all HTML)
        let rule = MD033NoInlineHtml::with_disallowed(vec![]);
        let content = "<div>content</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should flag <div> in default mode
        assert_eq!(result.len(), 1, "Empty disallowed = default mode");
    }

    #[test]
    fn test_md033_jsx_fragments_in_mdx() {
        // JSX fragments (<> and </>) should not trigger warnings in MDX
        let rule = MD033NoInlineHtml::default();
        let content = r#"# MDX Document

<>
  <Heading />
  <Content />
</>

<div>Regular HTML should still be flagged</div>
"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::MDX, None);
        let result = rule.check(&ctx).unwrap();

        // Should only flag <div>, not the fragments or JSX components
        assert_eq!(result.len(), 1, "Should only find one HTML tag (the div)");
        assert!(
            result[0].message.contains("<div>"),
            "Should flag <div>, not JSX fragments"
        );
    }

    #[test]
    fn test_md033_jsx_components_in_mdx() {
        // JSX components (capitalized) should not trigger warnings in MDX
        let rule = MD033NoInlineHtml::default();
        let content = r#"<CustomComponent prop="value">
  Content
</CustomComponent>

<MyButton onClick={handler}>Click</MyButton>
"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::MDX, None);
        let result = rule.check(&ctx).unwrap();

        // No warnings - all are JSX components
        assert_eq!(result.len(), 0, "Should not flag JSX components in MDX");
    }

    #[test]
    fn test_md033_jsx_not_skipped_in_standard_markdown() {
        // In standard markdown, capitalized tags should still be flagged if they're valid HTML
        let rule = MD033NoInlineHtml::default();
        let content = "<Script>alert(1)</Script>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should flag <Script> in standard markdown (it's a valid HTML element)
        assert_eq!(result.len(), 1, "Should flag <Script> in standard markdown");
    }

    #[test]
    fn test_md033_jsx_attributes_in_mdx() {
        // Elements with JSX-specific attributes should not trigger warnings in MDX
        let rule = MD033NoInlineHtml::default();
        let content = r#"# MDX with JSX Attributes

<div className="card big">Content</div>

<button onClick={handleClick}>Click me</button>

<label htmlFor="input-id">Label</label>

<input onChange={handleChange} />

<div class="html-class">Regular HTML should be flagged</div>
"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::MDX, None);
        let result = rule.check(&ctx).unwrap();

        // Should only flag the div with regular HTML "class" attribute
        assert_eq!(
            result.len(),
            1,
            "Should only flag HTML element without JSX attributes, got: {result:?}"
        );
        assert!(
            result[0].message.contains("<div class="),
            "Should flag the div with HTML class attribute"
        );
    }

    #[test]
    fn test_md033_jsx_attributes_not_skipped_in_standard() {
        // In standard markdown, JSX attributes should still be flagged
        let rule = MD033NoInlineHtml::default();
        let content = r#"<div className="card">Content</div>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        // Should flag in standard markdown
        assert_eq!(result.len(), 1, "Should flag JSX-style elements in standard markdown");
    }

    // Auto-fix tests for MD033

    #[test]
    fn test_md033_fix_disabled_by_default() {
        // Auto-fix should be disabled by default
        let rule = MD033NoInlineHtml::default();
        assert!(!rule.config.fix, "Fix should be disabled by default");
        assert_eq!(rule.fix_capability(), crate::rule::FixCapability::Unfixable);
    }

    #[test]
    fn test_md033_fix_enabled_em_to_italic() {
        // When fix is enabled, <em>text</em> should convert to *text*
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <em>emphasized text</em> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has *emphasized text* here.");
    }

    #[test]
    fn test_md033_fix_enabled_i_to_italic() {
        // <i>text</i> should convert to *text*
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <i>italic text</i> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has *italic text* here.");
    }

    #[test]
    fn test_md033_fix_enabled_strong_to_bold() {
        // <strong>text</strong> should convert to **text**
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <strong>bold text</strong> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has **bold text** here.");
    }

    #[test]
    fn test_md033_fix_enabled_b_to_bold() {
        // <b>text</b> should convert to **text**
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <b>bold text</b> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has **bold text** here.");
    }

    #[test]
    fn test_md033_fix_enabled_code_to_backticks() {
        // <code>text</code> should convert to `text`
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <code>inline code</code> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has `inline code` here.");
    }

    #[test]
    fn test_md033_fix_enabled_code_with_backticks() {
        // <code>text with `backticks`</code> should use double backticks
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <code>text with `backticks`</code> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has `` text with `backticks` `` here.");
    }

    #[test]
    fn test_md033_fix_enabled_br_trailing_spaces() {
        // <br> should convert to two trailing spaces + newline (default)
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "First line<br>Second line";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "First line  \nSecond line");
    }

    #[test]
    fn test_md033_fix_enabled_br_self_closing() {
        // <br/> and <br /> should also convert
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "First<br/>second<br />third";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "First  \nsecond  \nthird");
    }

    #[test]
    fn test_md033_fix_enabled_br_backslash_style() {
        // With br_style = backslash, <br> should convert to backslash + newline
        let config = MD033Config {
            allowed: Vec::new(),
            disallowed: Vec::new(),
            fix: true,
            br_style: md033_config::BrStyle::Backslash,
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "First line<br>Second line";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "First line\\\nSecond line");
    }

    #[test]
    fn test_md033_fix_enabled_hr() {
        // <hr> should convert to horizontal rule
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "Above<hr>Below";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Above\n---\nBelow");
    }

    #[test]
    fn test_md033_fix_enabled_hr_self_closing() {
        // <hr/> should also convert
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "Above<hr/>Below";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Above\n---\nBelow");
    }

    #[test]
    fn test_md033_fix_skips_nested_tags() {
        // Tags with nested HTML - outer tags may not be fully fixed due to overlapping ranges
        // The inner tags are processed first, which can invalidate outer tag ranges
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <em>text with <strong>nested</strong> tags</em> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Inner <strong> is converted to markdown, outer <em> range becomes invalid
        // This is expected behavior - user should run fix multiple times for nested tags
        assert_eq!(fixed, "This has <em>text with **nested** tags</em> here.");
    }

    #[test]
    fn test_md033_fix_skips_tags_with_attributes() {
        // Tags with attributes should NOT be fixed at all - leave as-is
        // User may want to keep the attributes (e.g., class="highlight" for styling)
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <em class=\"highlight\">emphasized</em> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Content should remain unchanged - we don't know if attributes matter
        assert_eq!(fixed, content);
    }

    #[test]
    fn test_md033_fix_disabled_no_changes() {
        // When fix is disabled, original content should be returned
        let rule = MD033NoInlineHtml::default(); // fix is false by default
        let content = "This has <em>emphasized text</em> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Should return original content when fix is disabled");
    }

    #[test]
    fn test_md033_fix_capability_enabled() {
        let rule = MD033NoInlineHtml::with_fix(true);
        assert_eq!(rule.fix_capability(), crate::rule::FixCapability::FullyFixable);
    }

    #[test]
    fn test_md033_fix_multiple_tags() {
        // Test fixing multiple HTML tags in one document
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "Here is <em>italic</em> and <strong>bold</strong> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Here is *italic* and **bold** text.");
    }

    #[test]
    fn test_md033_fix_uppercase_tags() {
        // HTML tags are case-insensitive
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <EM>emphasized</EM> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This has *emphasized* text.");
    }

    #[test]
    fn test_md033_fix_unsafe_tags_not_modified() {
        // Tags without safe markdown equivalents should NOT be modified
        // Only safe fixable tags (em, i, strong, b, code, br, hr) get converted
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This has <div>a div</div> content.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // <div> is not a safe fixable tag, so content should be unchanged
        assert_eq!(fixed, "This has <div>a div</div> content.");
    }

    #[test]
    fn test_md033_fix_img_tag_converted() {
        // <img> tags with simple src/alt attributes are converted to markdown images
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "Image: <img src=\"photo.jpg\" alt=\"My Photo\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // <img> is converted to ![alt](src) format
        assert_eq!(fixed, "Image: ![My Photo](photo.jpg)");
    }

    #[test]
    fn test_md033_fix_img_tag_with_extra_attrs_not_converted() {
        // <img> tags with width/height/style attributes are NOT converted
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "Image: <img src=\"photo.jpg\" alt=\"My Photo\" width=\"100\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Has width attribute - not safe to convert
        assert_eq!(fixed, "Image: <img src=\"photo.jpg\" alt=\"My Photo\" width=\"100\">");
    }

    #[test]
    fn test_md033_fix_relaxed_a_with_target_is_converted() {
        let rule = relaxed_fix_rule();
        let content = "Link: <a href=\"https://example.com\" target=\"_blank\">Example</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Link: [Example](https://example.com)");
    }

    #[test]
    fn test_md033_fix_relaxed_img_with_width_is_converted() {
        let rule = relaxed_fix_rule();
        let content = "Image: <img src=\"photo.jpg\" alt=\"My Photo\" width=\"100\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Image: ![My Photo](photo.jpg)");
    }

    #[test]
    fn test_md033_fix_relaxed_rejects_unknown_extra_attributes() {
        let rule = relaxed_fix_rule();
        let content = "Image: <img src=\"photo.jpg\" alt=\"My Photo\" aria-label=\"hero\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Unknown attributes should not be dropped by default");
    }

    #[test]
    fn test_md033_fix_relaxed_still_blocks_unsafe_schemes() {
        let rule = relaxed_fix_rule();
        let content = "Link: <a href=\"javascript:alert(1)\" target=\"_blank\">Example</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Unsafe URL schemes must never be converted");
    }

    #[test]
    fn test_md033_fix_relaxed_wrapper_strip_requires_second_pass_for_nested_html() {
        let rule = relaxed_fix_rule();
        let content = "<p align=\"center\">\n  <img src=\"logo.svg\" alt=\"Logo\" width=\"120\" />\n</p>";
        let ctx1 = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed_once = rule.fix(&ctx1).unwrap();
        assert!(
            fixed_once.contains("<p"),
            "First pass should keep wrapper when inner HTML is still present: {fixed_once}"
        );
        assert!(
            fixed_once.contains("![Logo](logo.svg)"),
            "Inner image should be converted on first pass: {fixed_once}"
        );

        let ctx2 = LintContext::new(&fixed_once, crate::config::MarkdownFlavor::Standard, None);
        let fixed_twice = rule.fix(&ctx2).unwrap();
        assert!(
            !fixed_twice.contains("<p"),
            "Second pass should strip configured wrapper: {fixed_twice}"
        );
        assert!(fixed_twice.contains("![Logo](logo.svg)"));
    }

    #[test]
    fn test_md033_fix_relaxed_multiple_droppable_attrs() {
        let rule = relaxed_fix_rule();
        let content = "<a href=\"https://example.com\" target=\"_blank\" rel=\"noopener\" class=\"btn\">Click</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "[Click](https://example.com)");
    }

    #[test]
    fn test_md033_fix_relaxed_img_multiple_droppable_attrs() {
        let rule = relaxed_fix_rule();
        let content = "<img src=\"logo.png\" alt=\"Logo\" width=\"120\" height=\"40\" style=\"border:none\" />";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "![Logo](logo.png)");
    }

    #[test]
    fn test_md033_fix_relaxed_event_handler_never_dropped() {
        let rule = relaxed_fix_rule();
        let content = "<a href=\"https://example.com\" onclick=\"track()\">Link</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Event handler attributes must block conversion");
    }

    #[test]
    fn test_md033_fix_relaxed_event_handler_even_with_custom_config() {
        // Even if someone adds on* to drop-attributes, event handlers must be rejected
        let config = MD033Config {
            fix: true,
            fix_mode: MD033FixMode::Relaxed,
            drop_attributes: vec!["on*".to_string(), "target".to_string()],
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "<a href=\"https://example.com\" onclick=\"alert(1)\">Link</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "on* event handlers must never be dropped");
    }

    #[test]
    fn test_md033_fix_relaxed_custom_drop_attributes() {
        let config = MD033Config {
            fix: true,
            fix_mode: MD033FixMode::Relaxed,
            drop_attributes: vec!["loading".to_string()],
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        // "loading" is in the custom list, "width" is NOT
        let content = "<img src=\"x.jpg\" alt=\"\" loading=\"lazy\">";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "![](x.jpg)", "Custom drop-attributes should be respected");

        let content2 = "<img src=\"x.jpg\" alt=\"\" width=\"100\">";
        let ctx2 = LintContext::new(content2, crate::config::MarkdownFlavor::Standard, None);
        let fixed2 = rule.fix(&ctx2).unwrap();
        assert_eq!(
            fixed2, content2,
            "Attributes not in custom list should block conversion"
        );
    }

    #[test]
    fn test_md033_fix_relaxed_custom_strip_wrapper() {
        let config = MD033Config {
            fix: true,
            fix_mode: MD033FixMode::Relaxed,
            strip_wrapper_elements: vec!["div".to_string()],
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "<div>Some text content</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Some text content");
    }

    #[test]
    fn test_md033_fix_relaxed_wrapper_with_plain_text() {
        let rule = relaxed_fix_rule();
        let content = "<p align=\"center\">Just some text</p>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Just some text");
    }

    #[test]
    fn test_md033_fix_relaxed_data_attr_with_wildcard() {
        let config = MD033Config {
            fix: true,
            fix_mode: MD033FixMode::Relaxed,
            drop_attributes: vec!["data-*".to_string(), "target".to_string()],
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "<a href=\"https://example.com\" data-tracking=\"abc\" target=\"_blank\">Link</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "[Link](https://example.com)");
    }

    #[test]
    fn test_md033_fix_relaxed_mixed_droppable_and_blocking_attrs() {
        let rule = relaxed_fix_rule();
        // "target" is droppable, "aria-label" is not in the default list
        let content = "<a href=\"https://example.com\" target=\"_blank\" aria-label=\"nav\">Link</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Non-droppable attribute should block conversion");
    }

    #[test]
    fn test_md033_fix_relaxed_badge_pattern() {
        // Common GitHub README badge pattern
        let rule = relaxed_fix_rule();
        let content = "<a href=\"https://crates.io/crates/rumdl\" target=\"_blank\"><img src=\"https://img.shields.io/crates/v/rumdl.svg\" alt=\"Crate\" width=\"120\" /></a>";
        let ctx1 = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed_once = rule.fix(&ctx1).unwrap();
        // First pass should convert the inner <img>
        assert!(
            fixed_once.contains("![Crate](https://img.shields.io/crates/v/rumdl.svg)"),
            "Inner img should be converted: {fixed_once}"
        );

        // Second pass converts the <a> wrapper
        let ctx2 = LintContext::new(&fixed_once, crate::config::MarkdownFlavor::Standard, None);
        let fixed_twice = rule.fix(&ctx2).unwrap();
        assert!(
            fixed_twice
                .contains("[![Crate](https://img.shields.io/crates/v/rumdl.svg)](https://crates.io/crates/rumdl)"),
            "Badge should produce nested markdown image link: {fixed_twice}"
        );
    }

    #[test]
    fn test_md033_fix_relaxed_conservative_mode_unchanged() {
        // Verify conservative mode (default) is unaffected by the relaxed logic
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<a href=\"https://example.com\" target=\"_blank\">Link</a>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, content, "Conservative mode should not drop target attribute");
    }

    #[test]
    fn test_md033_fix_relaxed_img_inside_pre_not_converted() {
        // <img> inside <pre> must NOT be converted, even in relaxed mode
        let rule = relaxed_fix_rule();
        let content = "<pre>\n  <img src=\"diagram.png\" alt=\"d\" width=\"100\" />\n</pre>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert!(fixed.contains("<img"), "img inside pre must not be converted: {fixed}");
    }

    #[test]
    fn test_md033_fix_relaxed_wrapper_nested_inside_div_not_stripped() {
        // <p> nested inside <div> should not be stripped
        let rule = relaxed_fix_rule();
        let content = "<div><p>text</p></div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert!(
            fixed.contains("<p>text</p>") || fixed.contains("<p>"),
            "Nested <p> inside <div> should not be stripped: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_relaxed_img_inside_nested_wrapper_not_converted() {
        // <img> inside <div><p>...</p></div> must NOT be converted because the
        // <p> wrapper can't be stripped (it's nested), so the markdown would be
        // stuck inside an HTML block where it won't render.
        let rule = relaxed_fix_rule();
        let content = "<div><p><img src=\"x.jpg\" alt=\"pic\" width=\"100\" /></p></div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert!(
            fixed.contains("<img"),
            "img inside nested wrapper must not be converted: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_mixed_safe_tags() {
        // All tags are now safe fixable (em, img, strong)
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<em>italic</em> and <img src=\"x.jpg\"> and <strong>bold</strong>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // All are converted
        assert_eq!(fixed, "*italic* and ![](x.jpg) and **bold**");
    }

    #[test]
    fn test_md033_fix_multiple_tags_same_line() {
        // Multiple tags on the same line should all be fixed correctly
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "Regular text <i>italic</i> and <b>bold</b> here.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "Regular text *italic* and **bold** here.");
    }

    #[test]
    fn test_md033_fix_multiple_em_tags_same_line() {
        // Multiple em/strong tags on the same line
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<em>first</em> and <strong>second</strong> and <code>third</code>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "*first* and **second** and `third`");
    }

    #[test]
    fn test_md033_fix_skips_tags_inside_pre() {
        // Tags inside <pre> blocks should NOT be fixed (would break structure)
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<pre><code><em>VALUE</em></code></pre>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // The <em> inside <pre><code> should NOT be converted
        // Only the outer structure might be changed
        assert!(
            !fixed.contains("*VALUE*"),
            "Tags inside <pre> should not be converted to markdown. Got: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_skips_tags_inside_div() {
        // Tags inside HTML block elements should not be fixed
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<div>\n<em>emphasized</em>\n</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // The <em> inside <div> should not be converted to *emphasized*
        assert!(
            !fixed.contains("*emphasized*"),
            "Tags inside HTML blocks should not be converted. Got: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_outside_html_block() {
        // Tags outside HTML blocks should still be fixed
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<div>\ncontent\n</div>\n\nOutside <em>emphasized</em> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // The <em> outside the div should be converted
        assert!(
            fixed.contains("*emphasized*"),
            "Tags outside HTML blocks should be converted. Got: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_with_id_attribute() {
        // Tags with id attributes should not be fixed (id might be used for anchors)
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "See <em id=\"important\">this note</em> for details.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should remain unchanged - id attribute matters for linking
        assert_eq!(fixed, content);
    }

    #[test]
    fn test_md033_fix_with_style_attribute() {
        // Tags with style attributes should not be fixed
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This is <strong style=\"color: red\">important</strong> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should remain unchanged - style attribute provides formatting
        assert_eq!(fixed, content);
    }

    #[test]
    fn test_md033_fix_mixed_with_and_without_attributes() {
        // Mix of tags with and without attributes
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<em>normal</em> and <em class=\"special\">styled</em> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Only the tag without attributes should be fixed
        assert_eq!(fixed, "*normal* and <em class=\"special\">styled</em> text.");
    }

    #[test]
    fn test_md033_quick_fix_tag_with_attributes_no_fix() {
        // Quick fix should not be provided for tags with attributes
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<em class=\"test\">emphasized</em>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();

        assert_eq!(result.len(), 1, "Should find one HTML tag");
        // No fix should be provided for tags with attributes
        assert!(
            result[0].fix.is_none(),
            "Should NOT have a fix for tags with attributes"
        );
    }

    #[test]
    fn test_md033_fix_skips_html_entities() {
        // Tags containing HTML entities should NOT be fixed
        // HTML entities need HTML context to render; markdown won't process them
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<code>&vert;</code>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should remain unchanged - converting would break rendering
        assert_eq!(fixed, content);
    }

    #[test]
    fn test_md033_fix_skips_multiple_html_entities() {
        // Multiple HTML entities should also be skipped
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<code>&lt;T&gt;</code>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should remain unchanged
        assert_eq!(fixed, content);
    }

    #[test]
    fn test_md033_fix_allows_ampersand_without_entity() {
        // Content with & but no semicolon should still be fixed
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<code>a & b</code>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should be converted since & is not part of an entity
        assert_eq!(fixed, "`a & b`");
    }

    #[test]
    fn test_md033_fix_em_with_entities_skipped() {
        // <em> with entities should also be skipped
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<em>&nbsp;text</em>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should remain unchanged
        assert_eq!(fixed, content);
    }

    #[test]
    fn test_md033_fix_skips_nested_em_in_code() {
        // Tags nested inside other HTML elements should NOT be fixed
        // e.g., <code><em>n</em></code> - the <em> should not be converted
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<code><em>n</em></code>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // The inner <em> should NOT be converted to *n* because it's nested
        // The whole structure should be left as-is (or outer code converted, but not inner)
        assert!(
            !fixed.contains("*n*"),
            "Nested <em> should not be converted to markdown. Got: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_skips_nested_in_table() {
        // Tags nested in HTML structures in tables should not be fixed
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "| <code>><em>n</em></code> | description |";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        // Should not convert nested <em> to *n*
        assert!(
            !fixed.contains("*n*"),
            "Nested tags in table should not be converted. Got: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_standalone_em_still_converted() {
        // Standalone (non-nested) <em> should still be converted
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "This is <em>emphasized</em> text.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();
        assert_eq!(fixed, "This is *emphasized* text.");
    }

    // ==========================================================================
    // Obsidian Templater Plugin Syntax Tests
    //
    // Templater is a popular Obsidian plugin that uses `<% ... %>` syntax for
    // template interpolation. The `<%` pattern is NOT captured by the HTML tag
    // parser because `%` is not a valid HTML tag name character (tags must start
    // with a letter). This behavior is documented here with comprehensive tests.
    //
    // Reference: https://silentvoid13.github.io/Templater/
    // ==========================================================================

    #[test]
    fn test_md033_templater_basic_interpolation_not_flagged() {
        // Basic Templater interpolation: <% expr %>
        // Should NOT be flagged because `%` is not a valid HTML tag character
        let rule = MD033NoInlineHtml::default();
        let content = "Today is <% tp.date.now() %> which is nice.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater basic interpolation should not be flagged as HTML. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_file_functions_not_flagged() {
        // Templater file functions: <% tp.file.* %>
        let rule = MD033NoInlineHtml::default();
        let content = "File: <% tp.file.title %>\nCreated: <% tp.file.creation_date() %>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater file functions should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_with_arguments_not_flagged() {
        // Templater with function arguments
        let rule = MD033NoInlineHtml::default();
        let content = r#"Date: <% tp.date.now("YYYY-MM-DD") %>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater with arguments should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_javascript_execution_not_flagged() {
        // Templater JavaScript execution block: <%* code %>
        let rule = MD033NoInlineHtml::default();
        let content = "<%* const today = tp.date.now(); tR += today; %>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater JS execution block should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_dynamic_execution_not_flagged() {
        // Templater dynamic/preview execution: <%+ expr %>
        let rule = MD033NoInlineHtml::default();
        let content = "Dynamic: <%+ tp.date.now() %>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater dynamic execution should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_whitespace_trim_all_not_flagged() {
        // Templater whitespace control - trim all: <%_ expr _%>
        let rule = MD033NoInlineHtml::default();
        let content = "<%_ tp.date.now() _%>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater trim-all whitespace should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_whitespace_trim_newline_not_flagged() {
        // Templater whitespace control - trim newline: <%- expr -%>
        let rule = MD033NoInlineHtml::default();
        let content = "<%- tp.date.now() -%>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater trim-newline should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_combined_modifiers_not_flagged() {
        // Templater combined whitespace and execution modifiers
        let rule = MD033NoInlineHtml::default();
        let contents = [
            "<%-* const x = 1; -%>",  // trim + JS execution
            "<%_+ tp.date.now() _%>", // trim-all + dynamic
            "<%- tp.file.title -%>",  // trim-newline only
            "<%_ tp.file.title _%>",  // trim-all only
        ];
        for content in contents {
            let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
            let result = rule.check(&ctx).unwrap();
            assert!(
                result.is_empty(),
                "Templater combined modifiers should not be flagged: {content}. Got: {result:?}"
            );
        }
    }

    #[test]
    fn test_md033_templater_multiline_block_not_flagged() {
        // Multi-line Templater JavaScript block
        let rule = MD033NoInlineHtml::default();
        let content = r#"<%*
const x = 1;
const y = 2;
tR += x + y;
%>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater multi-line block should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_with_angle_brackets_in_condition_not_flagged() {
        // Templater with angle brackets in JavaScript condition
        // This is a key edge case: `<` inside Templater should not trigger HTML detection
        let rule = MD033NoInlineHtml::default();
        let content = "<%* if (x < 5) { tR += 'small'; } %>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater with angle brackets in conditions should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_mixed_with_html_only_html_flagged() {
        // Templater syntax mixed with actual HTML - only HTML should be flagged
        let rule = MD033NoInlineHtml::default();
        let content = "<% tp.date.now() %> is today's date. <div>This is HTML</div>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(result.len(), 1, "Should only flag the HTML div tag");
        assert!(
            result[0].message.contains("<div>"),
            "Should flag <div>, got: {}",
            result[0].message
        );
    }

    #[test]
    fn test_md033_templater_in_heading_not_flagged() {
        // Templater in markdown heading
        let rule = MD033NoInlineHtml::default();
        let content = "# <% tp.file.title %>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater in heading should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_multiple_on_same_line_not_flagged() {
        // Multiple Templater blocks on same line
        let rule = MD033NoInlineHtml::default();
        let content = "From <% tp.date.now() %> to <% tp.date.tomorrow() %> we have meetings.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Multiple Templater blocks should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_in_code_block_not_flagged() {
        // Templater syntax in code blocks should not be flagged (code blocks are skipped)
        let rule = MD033NoInlineHtml::default();
        let content = "```\n<% tp.date.now() %>\n```";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater in code block should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_in_inline_code_not_flagged() {
        // Templater syntax in inline code span should not be flagged
        let rule = MD033NoInlineHtml::default();
        let content = "Use `<% tp.date.now() %>` for current date.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater in inline code should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_also_works_in_standard_flavor() {
        // Templater syntax should also not be flagged in Standard flavor
        // because the HTML parser doesn't recognize `<%` as a valid tag
        let rule = MD033NoInlineHtml::default();
        let content = "<% tp.date.now() %> works everywhere.";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater should not be flagged even in Standard flavor. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_empty_tag_not_flagged() {
        // Empty Templater tags
        let rule = MD033NoInlineHtml::default();
        let content = "<%>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Empty Templater-like tag should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_unclosed_not_flagged() {
        // Unclosed Templater tags - these are template errors, not HTML
        let rule = MD033NoInlineHtml::default();
        let content = "<% tp.date.now() without closing tag";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Unclosed Templater should not be flagged as HTML. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_with_newlines_inside_not_flagged() {
        // Templater with newlines inside the expression
        let rule = MD033NoInlineHtml::default();
        let content = r#"<% tp.date.now("YYYY") +
"-" +
tp.date.now("MM") %>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Templater with internal newlines should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_erb_style_tags_not_flagged() {
        // ERB/EJS style tags (similar to Templater) are also not HTML
        // This documents the general principle that `<%` is not valid HTML
        let rule = MD033NoInlineHtml::default();
        let content = "<%= variable %> and <% code %> and <%# comment %>";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "ERB/EJS style tags should not be flagged as HTML. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_templater_complex_expression_not_flagged() {
        // Complex Templater expression with multiple function calls
        let rule = MD033NoInlineHtml::default();
        let content = r#"<%*
const file = tp.file.title;
const date = tp.date.now("YYYY-MM-DD");
const folder = tp.file.folder();
tR += `# ${file}\n\nCreated: ${date}\nIn: ${folder}`;
%>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Obsidian, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "Complex Templater expression should not be flagged. Got: {result:?}"
        );
    }

    #[test]
    fn test_md033_percent_sign_variations_not_flagged() {
        // Various patterns starting with <% that should all be safe
        let rule = MD033NoInlineHtml::default();
        let patterns = [
            "<%=",  // ERB output
            "<%#",  // ERB comment
            "<%%",  // Double percent
            "<%!",  // Some template engines
            "<%@",  // JSP directive
            "<%--", // JSP comment
        ];
        for pattern in patterns {
            let content = format!("{pattern} content %>");
            let ctx = LintContext::new(&content, crate::config::MarkdownFlavor::Standard, None);
            let result = rule.check(&ctx).unwrap();
            assert!(
                result.is_empty(),
                "Pattern {pattern} should not be flagged. Got: {result:?}"
            );
        }
    }

    // ───── Bug #3: Bracket escaping in image-inside-link conversion ─────
    //
    // When <a> wraps already-converted markdown image text, the bracket escaping
    // must be skipped to produce valid [![alt](url)](href) instead of !\[\](url)

    #[test]
    fn test_md033_fix_a_wrapping_markdown_image_no_escaped_brackets() {
        // When <a> wraps a markdown image (from a prior fix iteration),
        // the result should be [![](url)](href) — no escaped brackets
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = r#"<a href="https://example.com">![](https://example.com/image.png)</a>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(fixed, "[![](https://example.com/image.png)](https://example.com)",);
        assert!(!fixed.contains(r"\["), "Must not escape brackets: {fixed}");
        assert!(!fixed.contains(r"\]"), "Must not escape brackets: {fixed}");
    }

    #[test]
    fn test_md033_fix_a_wrapping_markdown_image_with_alt() {
        // <a> wrapping ![alt](url) preserves alt text in linked image
        let rule = MD033NoInlineHtml::with_fix(true);
        let content =
            r#"<a href="https://github.com/repo">![Contributors](https://contrib.rocks/image?repo=org/repo)</a>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(
            fixed,
            "[![Contributors](https://contrib.rocks/image?repo=org/repo)](https://github.com/repo)"
        );
    }

    #[test]
    fn test_md033_fix_img_without_alt_produces_empty_alt() {
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = r#"<img src="photo.jpg" />"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();

        assert_eq!(fixed, "![](photo.jpg)");
    }

    #[test]
    fn test_md033_fix_leaves_a_tag_inside_an_unnamed_html_block() {
        // `<span>` alone on its line opens an HTML block running to the next
        // blank line, so the `<img>` below it is raw HTML, where Markdown image
        // syntax would render as literal text. The block's own opening tag, and
        // an image after the blank line, are still converted.
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = "<span>\n<img src=\"a.png\" />\n\n<img src=\"b.png\" />\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        assert_eq!(rule.fix(&ctx).unwrap(), "<span>\n<img src=\"a.png\" />\n\n![](b.png)\n");
    }

    #[test]
    fn test_md033_fix_a_with_plain_text_still_escapes_brackets() {
        // Plain text brackets inside <a> SHOULD be escaped
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = r#"<a href="https://example.com">text with [brackets]</a>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();

        assert!(
            fixed.contains(r"\[brackets\]"),
            "Plain text brackets should be escaped: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_a_with_image_plus_extra_text_escapes_brackets() {
        // Mixed content: image followed by bracketed text — brackets must be escaped
        // The image detection must NOT match partial content
        let rule = MD033NoInlineHtml::with_fix(true);
        let content = r#"<a href="/link">![](img.png) see [docs]</a>"#;
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let fixed = rule.fix(&ctx).unwrap();

        // "see [docs]" brackets should be escaped since inner content is mixed
        assert!(
            fixed.contains(r"\[docs\]"),
            "Brackets in mixed image+text content should be escaped: {fixed}"
        );
    }

    #[test]
    fn test_md033_fix_img_in_a_end_to_end() {
        // End-to-end: verify that iterative fixing of <a><img></a>
        // produces the correct final result through the fix coordinator
        use crate::config::Config;
        use crate::fix_coordinator::FixCoordinator;

        let rule = MD033NoInlineHtml::with_fix(true);
        let rules: Vec<Box<dyn crate::rule::Rule>> = vec![Box::new(rule)];

        let mut content =
            r#"<a href="https://github.com/org/repo"><img src="https://contrib.rocks/image?repo=org/repo" /></a>"#
                .to_string();
        let config = Config::default();
        let coordinator = FixCoordinator::new();

        let result = coordinator
            .apply_fixes_iterative(&rules, &[], &mut content, &config, 10, None)
            .unwrap();

        assert_eq!(
            content, "[![](https://contrib.rocks/image?repo=org/repo)](https://github.com/org/repo)",
            "End-to-end: <a><img></a> should become valid linked image"
        );
        assert!(result.converged);
        assert!(!content.contains(r"\["), "No escaped brackets: {content}");
    }

    #[test]
    fn test_md033_fix_img_in_a_with_alt_end_to_end() {
        use crate::config::Config;
        use crate::fix_coordinator::FixCoordinator;

        let rule = MD033NoInlineHtml::with_fix(true);
        let rules: Vec<Box<dyn crate::rule::Rule>> = vec![Box::new(rule)];

        let mut content =
            r#"<a href="https://github.com/org/repo"><img src="https://contrib.rocks/image" alt="Contributors" /></a>"#
                .to_string();
        let config = Config::default();
        let coordinator = FixCoordinator::new();

        let result = coordinator
            .apply_fixes_iterative(&rules, &[], &mut content, &config, 10, None)
            .unwrap();

        assert_eq!(
            content,
            "[![Contributors](https://contrib.rocks/image)](https://github.com/org/repo)",
        );
        assert!(result.converged);
    }

    // =========================================================================
    // table_allowed_elements config option tests
    //
    // Mirrors markdownlint's `table_allowed_elements`: when unset, the in-table
    // allowlist falls back to `allowed_elements`; when explicitly set (even to
    // []), it overrides for tags inside GFM table cells. Out-of-table tags are
    // never affected by this option.
    // =========================================================================

    #[test]
    fn test_md033_table_allowed_unset_falls_back_to_allowed() {
        let config = MD033Config {
            allowed: vec!["br".to_string()],
            table_allowed_elements: None,
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "| col |\n|-----|\n| a<br>b |\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "<br> in table cell should be allowed via fallback to `allowed`, got {result:?}"
        );
    }

    #[test]
    fn test_md033_table_allowed_explicit_empty_rejects_in_tables() {
        let config = MD033Config {
            allowed: vec!["br".to_string()],
            table_allowed_elements: Some(Vec::new()),
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "| col |\n|-----|\n| a<br>b |\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "Explicit empty table_allowed should reject <br> in tables even if it's in `allowed`, got {result:?}"
        );
        assert_eq!(result[0].line, 3);
    }

    #[test]
    fn test_md033_table_allowed_explicit_list_overrides_in_tables() {
        let config = MD033Config {
            allowed: vec!["br".to_string()],
            table_allowed_elements: Some(vec!["img".to_string()]),
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        // <br> is in allowed but NOT in table_allowed, so it should be flagged in table.
        // <img> is in table_allowed only, so it should be permitted in table.
        let content = "| col |\n|-----|\n| <br><img src=\"x\"/> |\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert_eq!(
            result.len(),
            1,
            "table_allowed should override `allowed` inside tables, got {result:?}"
        );
        assert!(
            result[0].message.contains("br"),
            "expected the flagged tag to be <br>, got {:?}",
            result[0].message
        );
    }

    #[test]
    fn test_md033_table_allowed_does_not_affect_out_of_table_tags() {
        let config = MD033Config {
            allowed: vec!["br".to_string()],
            table_allowed_elements: Some(Vec::new()),
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        // <br> outside a table — should still be allowed via `allowed`.
        let content = "Paragraph with <br> tag.\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "<br> outside tables must still be allowed by `allowed`, got {result:?}"
        );
    }

    #[test]
    fn test_md033_table_allowed_kebab_case_parses() {
        let toml_str = r#"
            allowed-elements = ["br"]
            table-allowed-elements = ["img"]
        "#;
        let config: MD033Config = toml::from_str(toml_str).unwrap();
        assert_eq!(config.allowed, vec!["br"]);
        assert_eq!(
            config.table_allowed_elements.as_deref(),
            Some(["img".to_string()].as_slice())
        );
    }

    #[test]
    fn test_md033_table_allowed_snake_case_alias_parses() {
        let toml_str = r#"
            allowed_elements = ["br"]
            table_allowed_elements = ["img"]
        "#;
        let config: MD033Config = toml::from_str(toml_str).unwrap();
        assert_eq!(config.allowed, vec!["br"]);
        assert_eq!(
            config.table_allowed_elements.as_deref(),
            Some(["img".to_string()].as_slice())
        );
    }

    #[test]
    fn test_md033_table_allowed_default_is_none() {
        let cfg = MD033Config::default();
        assert!(
            cfg.table_allowed_elements.is_none(),
            "Default for table_allowed_elements should be None (so it falls back to `allowed`)"
        );
    }

    #[test]
    fn test_md033_table_allowed_case_insensitive() {
        let config = MD033Config {
            allowed: Vec::new(),
            table_allowed_elements: Some(vec!["BR".to_string()]),
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "| col |\n|-----|\n| a<br>b |\n";
        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard, None);
        let result = rule.check(&ctx).unwrap();
        assert!(
            result.is_empty(),
            "table_allowed should be case-insensitive, got {result:?}"
        );
    }

    // =========================================================================
    // allowed_inside config option tests
    //
    // An element named here is permitted, and so is everything between its
    // opening and closing tag.
    // =========================================================================

    /// The raw tags MD033 reports, in document order.
    fn reported_tags_in(rule: &MD033NoInlineHtml, content: &str, flavor: crate::config::MarkdownFlavor) -> Vec<String> {
        let ctx = LintContext::new(content, flavor, None);
        rule.check(&ctx)
            .unwrap()
            .into_iter()
            .map(|warning| {
                warning
                    .message
                    .strip_prefix("Inline HTML found: ")
                    .expect("MD033 reports the tag it found")
                    .to_string()
            })
            .collect()
    }

    fn reported_tags(rule: &MD033NoInlineHtml, content: &str) -> Vec<String> {
        reported_tags_in(rule, content, crate::config::MarkdownFlavor::Standard)
    }

    fn rule_allowing_inside(elements: &[&str]) -> MD033NoInlineHtml {
        MD033NoInlineHtml::from_config_struct(MD033Config {
            allowed_inside: elements.iter().map(ToString::to_string).collect(),
            ..MD033Config::default()
        })
    }

    #[test]
    fn test_md033_allowed_inside_permits_the_element_and_its_contents() {
        let content = "<details>\n<summary>\n<a href=\"./other.md\">Go there</a>\n</summary>\n<b>bold</b>\n</details>\n\nOutside <b>bold</b>.\n";

        let without = reported_tags(&MD033NoInlineHtml::default(), content);
        assert_eq!(
            without,
            vec!["<details>", "<summary>", "<a href=\"./other.md\">", "<b>", "<b>"],
            "control: every tag is reported without the option"
        );

        let with = reported_tags(&rule_allowing_inside(&["details"]), content);
        assert_eq!(
            with,
            vec!["<b>"],
            "only the tag outside the details element is left, got {with:?}"
        );
    }

    #[test]
    fn test_md033_allowed_inside_ends_at_the_matching_closing_tag() {
        // The closing tag of the inner element does not end the outer one.
        let content = "<details>\n<details>\n<b>x</b>\n</details>\n<i>y</i>\n</details>\n<b>z</b>\n";
        let reported = reported_tags(&rule_allowing_inside(&["details"]), content);
        assert_eq!(reported, vec!["<b>"], "got {reported:?}");

        // An element left open covers the rest of the document.
        let unclosed = "<details>\n\n<b>x</b>\n\n<i>y</i>\n";
        assert!(
            reported_tags(&rule_allowing_inside(&["details"]), unclosed).is_empty(),
            "an unclosed element reaches the end of the document"
        );
    }

    #[test]
    fn test_md033_allowed_inside_ignores_an_element_quoted_in_a_code_block() {
        let content = "```html\n<details>\n```\n\n<b>x</b>\n";
        let reported = reported_tags(&rule_allowing_inside(&["details"]), content);
        assert_eq!(
            reported,
            vec!["<b>"],
            "a code block quotes the element, it does not open one: {reported:?}"
        );
    }

    #[test]
    fn test_md033_allowed_inside_a_void_element_holds_nothing() {
        // <br> has no contents, so naming it permits the <br> itself and nothing else.
        let content = "a<br>b <b>x</b>\n";
        let reported = reported_tags(&rule_allowing_inside(&["br"]), content);
        assert_eq!(
            reported,
            vec!["<br>", "<b>"],
            "a void element must not swallow the rest of the document: {reported:?}"
        );
    }

    #[test]
    fn test_md033_allowed_inside_is_case_insensitive() {
        let content = "<DETAILS>\n<b>x</b>\n</DETAILS>\n<i>y</i>\n";
        let reported = reported_tags(&rule_allowing_inside(&["Details"]), content);
        assert_eq!(reported, vec!["<i>"], "got {reported:?}");
    }

    #[test]
    fn test_md033_allowed_inside_takes_no_part_in_disallowed_mode() {
        let config = MD033Config {
            allowed_inside: vec!["details".to_string()],
            disallowed: vec!["b".to_string()],
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let content = "<details>\n<b>x</b>\n<kbd>k</kbd>\n</details>\n";
        let reported = reported_tags(&rule, content);
        assert_eq!(
            reported,
            vec!["<b>"],
            "a denylist names what is wrong wherever it appears: {reported:?}"
        );
    }

    #[test]
    fn test_md033_allowed_inside_yields_to_an_explicit_table_allowlist() {
        let content = "| col |\n|-----|\n| <details><b>x</b></details> |\n";

        let unset = reported_tags(&rule_allowing_inside(&["details"]), content);
        assert!(
            unset.is_empty(),
            "without a table allowlist the element applies inside a cell too: {unset:?}"
        );

        let config = MD033Config {
            allowed_inside: vec!["details".to_string()],
            table_allowed_elements: Some(Vec::new()),
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let reported = reported_tags(&rule, content);
        assert_eq!(
            reported,
            vec!["<details>", "<b>"],
            "an explicit table allowlist decides inside a cell: {reported:?}"
        );
    }

    // =========================================================================
    // no-markdown-equivalent sentinel tests
    //
    // Permits every element Markdown has no syntax for, so only elements a
    // reader could have written in Markdown are reported.
    // =========================================================================

    fn rule_allowing(elements: &[&str]) -> MD033NoInlineHtml {
        MD033NoInlineHtml::with_allowed(elements.iter().map(ToString::to_string).collect())
    }

    #[test]
    fn test_md033_no_markdown_equivalent_reports_only_what_markdown_can_write() {
        let content = "Press <kbd>Ctrl</kbd>, <b>bold</b>, <mark>hi</mark>, <abbr title=\"x\">A</abbr>, <em>i</em>, <details>d</details>\n";

        let control = reported_tags(&MD033NoInlineHtml::default(), content);
        assert_eq!(control.len(), 6, "control: every tag is reported: {control:?}");

        let reported = reported_tags(&rule_allowing(&["no-markdown-equivalent"]), content);
        assert_eq!(
            reported,
            vec!["<b>", "<em>"],
            "only the elements with Markdown syntax are left: {reported:?}"
        );
    }

    #[test]
    fn test_md033_no_markdown_equivalent_keeps_reporting_gfm_filtered_tags() {
        // Nothing renders these, so permitting them is never about expressiveness.
        let content = "<kbd>k</kbd>\n\n<script>alert(1)</script>\n\n<iframe src=\"x\"></iframe>\n";
        let reported = reported_tags(&rule_allowing(&["no-markdown-equivalent"]), content);
        assert_eq!(reported, vec!["<script>", "<iframe src=\"x\">"], "got {reported:?}");
    }

    #[test]
    fn test_md033_no_markdown_equivalent_follows_the_flavor() {
        let content = "H<sub>2</sub>O, x<sup>2</sup>, <mark>hi</mark>, <kbd>k</kbd>\n";
        let rule = rule_allowing(&["no-markdown-equivalent"]);

        assert!(
            reported_tags_in(&rule, content, crate::config::MarkdownFlavor::Standard).is_empty(),
            "standard Markdown writes none of these"
        );
        assert_eq!(
            reported_tags_in(&rule, content, crate::config::MarkdownFlavor::Pandoc),
            vec!["<sub>", "<sup>"],
            "Pandoc writes ~sub~ and ^sup^"
        );
        assert_eq!(
            reported_tags_in(&rule, content, crate::config::MarkdownFlavor::Obsidian),
            vec!["<mark>"],
            "Obsidian writes ==highlight=="
        );
    }

    #[test]
    fn test_md033_no_markdown_equivalent_permits_a_line_break_inside_a_table_cell() {
        // Two trailing spaces do not survive inside a cell, so <br> has no
        // equivalent there and every other one it does.
        let content = "| col |\n|-----|\n| a<br>b |\n\nOutside a<br>b.\n";
        let reported = reported_tags(&rule_allowing(&["no-markdown-equivalent"]), content);
        assert_eq!(reported, vec!["<br>"], "got {reported:?}");
    }

    #[test]
    fn test_md033_no_markdown_equivalent_composes_with_named_elements() {
        let content = "<kbd>k</kbd> <b>bold</b> <em>italic</em>\n";
        let reported = reported_tags(&rule_allowing(&["no-markdown-equivalent", "b"]), content);
        assert_eq!(
            reported,
            vec!["<em>"],
            "a named element joins the ones the sentinel permits: {reported:?}"
        );
    }

    #[test]
    fn test_md033_no_markdown_equivalent_reads_the_snake_case_spelling() {
        // Config values are not normalized on the way in, unlike config keys.
        let content = "<kbd>k</kbd> <b>bold</b>\n";
        let reported = reported_tags(&rule_allowing(&["no_markdown_equivalent"]), content);
        assert_eq!(reported, vec!["<b>"], "got {reported:?}");
    }

    #[test]
    fn test_md033_no_markdown_equivalent_takes_no_part_in_disallowed_mode() {
        let config = MD033Config {
            allowed: vec!["no-markdown-equivalent".to_string()],
            disallowed: vec!["kbd".to_string()],
            ..MD033Config::default()
        };
        let rule = MD033NoInlineHtml::from_config_struct(config);
        let reported = reported_tags(&rule, "<kbd>k</kbd> <b>bold</b>\n");
        assert_eq!(
            reported,
            vec!["<kbd>"],
            "a denylist names what is wrong wherever it appears: {reported:?}"
        );
    }
}