rumdl_lib/rules/
md053_link_image_reference_definitions.rs

1use crate::rule::{LintError, LintResult, LintWarning, Rule, Severity};
2use crate::rule_config_serde::RuleConfig;
3use crate::utils::range_utils::calculate_line_range;
4use fancy_regex::Regex as FancyRegex;
5use lazy_static::lazy_static;
6use regex::Regex;
7use serde::{Deserialize, Serialize};
8use std::collections::{HashMap, HashSet};
9
10lazy_static! {
11    // Link reference format: [text][reference]
12    // REMOVED: static ref LINK_REFERENCE_REGEX: FancyRegex = FancyRegex::new(r"\[([^\]]*)\]\s*\[([^\]]*)\]").unwrap();
13
14    // Image reference format: ![text][reference]
15    // REMOVED: static ref IMAGE_REFERENCE_REGEX: FancyRegex = FancyRegex::new(r"!\[([^\]]*)\]\s*\[([^\]]*)\]").unwrap();
16
17    // Shortcut reference links: [reference] - must not be followed by another bracket
18    // Allow references followed by punctuation like colon, period, comma (e.g., "[reference]:", "[reference].")
19    // Don't exclude references followed by ": " in the middle of a line (only at start of line)
20    static ref SHORTCUT_REFERENCE_REGEX: FancyRegex =
21        FancyRegex::new(r"(?<!\!)\[([^\]]+)\](?!\[)").unwrap();
22
23    // REMOVED: Empty reference links: [text][] or ![text][]
24    // static ref EMPTY_LINK_REFERENCE_REGEX: Regex = Regex::new(r"\[([^\]]+)\]\s*\[\s*\]").unwrap();
25    // static ref EMPTY_IMAGE_REFERENCE_REGEX: Regex = Regex::new(r"!\[([^\]]+)\]\s*\[\s*\]").unwrap();
26
27    // Link/image reference definition format: [reference]: URL
28    static ref REFERENCE_DEFINITION_REGEX: Regex =
29        Regex::new(r"^\s*\[([^\]]+)\]:\s+(.+)$").unwrap();
30
31    // Multi-line reference definition continuation pattern
32    static ref CONTINUATION_REGEX: Regex = Regex::new(r"^\s+(.+)$").unwrap();
33
34    // Code block regex - support indented code blocks for MkDocs tabs
35    static ref CODE_BLOCK_START_REGEX: Regex = Regex::new(r"^(\s*)(`{3,}|~{3,})").unwrap();
36    static ref CODE_BLOCK_END_REGEX: Regex = Regex::new(r"^(\s*)(`{3,}|~{3,})\s*$").unwrap();
37}
38
39/// Configuration for MD053 rule
40#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
41#[serde(rename_all = "kebab-case")]
42pub struct MD053Config {
43    /// List of reference names to keep even if unused
44    #[serde(default = "default_ignored_definitions")]
45    pub ignored_definitions: Vec<String>,
46}
47
48impl Default for MD053Config {
49    fn default() -> Self {
50        Self {
51            ignored_definitions: default_ignored_definitions(),
52        }
53    }
54}
55
56fn default_ignored_definitions() -> Vec<String> {
57    Vec::new()
58}
59
60impl RuleConfig for MD053Config {
61    const RULE_NAME: &'static str = "MD053";
62}
63
64/// Rule MD053: Link and image reference definitions should be used
65///
66/// See [docs/md053.md](../../docs/md053.md) for full documentation, configuration, and examples.
67///
68/// This rule is triggered when a link or image reference definition is declared but not used
69/// anywhere in the document. Unused reference definitions can create confusion and clutter.
70///
71/// ## Supported Reference Formats
72///
73/// This rule handles the following reference formats:
74///
75/// - **Full reference links/images**: `[text][reference]` or `![text][reference]`
76/// - **Collapsed reference links/images**: `[text][]` or `![text][]`
77/// - **Shortcut reference links**: `[reference]` (must be defined elsewhere)
78/// - **Reference definitions**: `[reference]: URL "Optional Title"`
79/// - **Multi-line reference definitions**:
80///   ```markdown
81///   [reference]: URL
82///      "Optional title continued on next line"
83///   ```
84///
85/// ## Configuration Options
86///
87/// The rule supports the following configuration options:
88///
89/// ```yaml
90/// MD053:
91///   ignored_definitions: []  # List of reference definitions to ignore (never report as unused)
92/// ```
93///
94/// ## Performance Optimizations
95///
96/// This rule implements various performance optimizations for handling large documents:
97///
98/// 1. **Caching**: The rule caches parsed definitions and references based on content hashing
99/// 2. **Efficient Reference Matching**: Uses HashMaps for O(1) lookups of definitions
100/// 3. **Smart Code Block Handling**: Efficiently skips references inside code blocks/spans
101/// 4. **Lazy Evaluation**: Only processes necessary portions of the document
102///
103/// ## Edge Cases Handled
104///
105/// - **Case insensitivity**: References are matched case-insensitively
106/// - **Escaped characters**: Properly processes escaped characters in references
107/// - **Unicode support**: Handles non-ASCII characters in references and URLs
108/// - **Code blocks**: Ignores references inside code blocks and spans
109/// - **Special characters**: Properly handles references with special characters
110///
111/// ## Fix Behavior
112///
113/// This rule does not provide automatic fixes. Unused references must be manually reviewed
114/// and removed, as they may be intentionally kept for future use or as templates.
115#[derive(Clone)]
116pub struct MD053LinkImageReferenceDefinitions {
117    config: MD053Config,
118}
119
120impl MD053LinkImageReferenceDefinitions {
121    /// Create a new instance of the MD053 rule
122    pub fn new() -> Self {
123        Self {
124            config: MD053Config::default(),
125        }
126    }
127
128    /// Create a new instance with the given configuration
129    pub fn from_config_struct(config: MD053Config) -> Self {
130        Self { config }
131    }
132
133    /// Check if a pattern is likely NOT a markdown reference
134    /// Returns true if this pattern should be skipped
135    fn is_likely_not_reference(text: &str) -> bool {
136        // Don't skip pure numeric patterns - they could be footnote references like [1]
137        // Only skip numeric ranges like [1:3], [0:10], etc.
138        if text.contains(':') && text.chars().all(|c| c.is_ascii_digit() || c == ':') {
139            return true;
140        }
141
142        // Skip glob/wildcard patterns like [*], [...], [**]
143        if text == "*" || text == "..." || text == "**" {
144            return true;
145        }
146
147        // Skip patterns that are just punctuation or operators
148        if text.chars().all(|c| !c.is_alphanumeric() && c != ' ') {
149            return true;
150        }
151
152        // Skip very short non-word patterns (likely operators or syntax)
153        // But allow single digits (could be footnotes) and single letters
154        if text.len() <= 2 && !text.chars().all(|c| c.is_alphanumeric()) {
155            return true;
156        }
157
158        // Skip descriptive patterns with colon like [default: the project root]
159        // But allow simple numeric ranges which are handled above
160        if text.contains(':') && text.contains(' ') {
161            return true;
162        }
163
164        // Skip alert/admonition patterns like [!WARN], [!NOTE], etc.
165        if text.starts_with('!') {
166            return true;
167        }
168
169        // Note: We don't filter out patterns with backticks because backticks in reference names
170        // are valid markdown syntax, e.g., [`dataclasses.InitVar`] is a valid reference name
171
172        // Also don't filter out references with dots - these are legitimate reference names
173        // like [tool.ruff] or [os.path] which are valid markdown references
174
175        // Note: We don't filter based on word count anymore because legitimate references
176        // can have many words, like "python language reference for import statements"
177        // Word count filtering was causing false positives where valid references were
178        // being incorrectly flagged as unused
179
180        false
181    }
182
183    /// Unescape a reference string by removing backslashes before special characters.
184    ///
185    /// This allows matching references like `[example\-reference]` with definitions like
186    /// `[example-reference]: http://example.com`
187    ///
188    /// Returns the unescaped reference string.
189    fn unescape_reference(reference: &str) -> String {
190        // Remove backslashes before special characters
191        reference.replace("\\", "")
192    }
193
194    /// Check if a reference definition is likely a comment-style reference.
195    ///
196    /// This recognizes common community patterns for comments in markdown:
197    /// - `[//]: # (comment)` - Most popular pattern
198    /// - `[comment]: # (text)` - Semantic pattern
199    /// - `[note]: # (text)` - Documentation pattern
200    /// - `[todo]: # (text)` - Task tracking pattern
201    /// - Any reference with just `#` as the URL (fragment-only, often unused)
202    ///
203    /// While not part of any official markdown spec (CommonMark, GFM), these patterns
204    /// are widely used across 23+ markdown implementations as documented in the community.
205    ///
206    /// # Arguments
207    /// * `ref_id` - The reference ID (already normalized to lowercase)
208    /// * `url` - The URL from the reference definition
209    ///
210    /// # Returns
211    /// `true` if this looks like a comment-style reference that should be ignored
212    fn is_likely_comment_reference(ref_id: &str, url: &str) -> bool {
213        // Common comment reference labels used in the community
214        const COMMENT_LABELS: &[&str] = &[
215            "//",      // [//]: # (comment) - most popular
216            "comment", // [comment]: # (text)
217            "note",    // [note]: # (text)
218            "todo",    // [todo]: # (text)
219            "fixme",   // [fixme]: # (text)
220            "hack",    // [hack]: # (text)
221        ];
222
223        let normalized_id = ref_id.trim().to_lowercase();
224        let normalized_url = url.trim();
225
226        // Pattern 1: Known comment labels with fragment URLs
227        // e.g., [//]: # (comment), [comment]: #section
228        if COMMENT_LABELS.contains(&normalized_id.as_str()) && normalized_url.starts_with('#') {
229            return true;
230        }
231
232        // Pattern 2: Any reference with just "#" as the URL
233        // This is often used as a comment placeholder or unused anchor
234        if normalized_url == "#" {
235            return true;
236        }
237
238        false
239    }
240
241    /// Find all link and image reference definitions in the content.
242    ///
243    /// This method returns a HashMap where the key is the normalized reference ID and the value is a vector of (start_line, end_line) tuples.
244    fn find_definitions(&self, ctx: &crate::lint_context::LintContext) -> HashMap<String, Vec<(usize, usize)>> {
245        let mut definitions: HashMap<String, Vec<(usize, usize)>> = HashMap::new();
246
247        // First, add all reference definitions from context
248        for ref_def in &ctx.reference_defs {
249            // Skip comment-style references (e.g., [//]: # (comment))
250            if Self::is_likely_comment_reference(&ref_def.id, &ref_def.url) {
251                continue;
252            }
253
254            // Apply unescape to handle escaped characters in definitions
255            let normalized_id = Self::unescape_reference(&ref_def.id); // Already lowercase from context
256            definitions
257                .entry(normalized_id)
258                .or_default()
259                .push((ref_def.line - 1, ref_def.line - 1)); // Convert to 0-indexed
260        }
261
262        // Handle multi-line definitions that might not be fully captured by ctx.reference_defs
263        let lines = &ctx.lines;
264        let mut i = 0;
265        while i < lines.len() {
266            let line_info = &lines[i];
267            let line = &line_info.content;
268
269            // Skip code blocks and front matter using line info
270            if line_info.in_code_block || line_info.in_front_matter {
271                i += 1;
272                continue;
273            }
274
275            // Check for multi-line continuation of existing definitions
276            if i > 0 && CONTINUATION_REGEX.is_match(line) {
277                // Find the reference definition this continues
278                let mut def_start = i - 1;
279                while def_start > 0 && !REFERENCE_DEFINITION_REGEX.is_match(&lines[def_start].content) {
280                    def_start -= 1;
281                }
282
283                if let Some(caps) = REFERENCE_DEFINITION_REGEX.captures(&lines[def_start].content) {
284                    let ref_id = caps.get(1).unwrap().as_str().trim();
285                    let normalized_id = Self::unescape_reference(ref_id).to_lowercase();
286
287                    // Update the end line for this definition
288                    if let Some(ranges) = definitions.get_mut(&normalized_id)
289                        && let Some(last_range) = ranges.last_mut()
290                        && last_range.0 == def_start
291                    {
292                        last_range.1 = i;
293                    }
294                }
295            }
296            i += 1;
297        }
298        definitions
299    }
300
301    /// Find all link and image reference reference usages in the content.
302    ///
303    /// This method returns a HashSet of all normalized reference IDs found in usage.
304    /// It leverages cached data from LintContext for efficiency.
305    fn find_usages(&self, ctx: &crate::lint_context::LintContext) -> HashSet<String> {
306        let mut usages: HashSet<String> = HashSet::new();
307
308        // 1. Add usages from cached reference links in LintContext
309        for link in &ctx.links {
310            if link.is_reference
311                && let Some(ref_id) = &link.reference_id
312            {
313                // Ensure the link itself is not inside a code block line
314                if !ctx.line_info(link.line).is_some_and(|info| info.in_code_block) {
315                    usages.insert(Self::unescape_reference(ref_id).to_lowercase());
316                }
317            }
318        }
319
320        // 2. Add usages from cached reference images in LintContext
321        for image in &ctx.images {
322            if image.is_reference
323                && let Some(ref_id) = &image.reference_id
324            {
325                // Ensure the image itself is not inside a code block line
326                if !ctx.line_info(image.line).is_some_and(|info| info.in_code_block) {
327                    usages.insert(Self::unescape_reference(ref_id).to_lowercase());
328                }
329            }
330        }
331
332        // 3. Find shortcut references [ref] not already handled by DocumentStructure.links
333        //    and ensure they are not within code spans or code blocks.
334        // Cache code spans once before the loop
335        let code_spans = ctx.code_spans();
336
337        for line_info in ctx.lines.iter() {
338            // Skip lines in code blocks or front matter
339            if line_info.in_code_block || line_info.in_front_matter {
340                continue;
341            }
342
343            // Skip lines that are reference definitions (start with [ref]: at beginning)
344            if REFERENCE_DEFINITION_REGEX.is_match(&line_info.content) {
345                continue;
346            }
347
348            // Find potential shortcut references
349            for caps in SHORTCUT_REFERENCE_REGEX.captures_iter(&line_info.content).flatten() {
350                if let Some(full_match) = caps.get(0)
351                    && let Some(ref_id_match) = caps.get(1)
352                {
353                    // Check if the match is within a code span
354                    let match_byte_offset = line_info.byte_offset + full_match.start();
355                    let in_code_span = code_spans
356                        .iter()
357                        .any(|span| match_byte_offset >= span.byte_offset && match_byte_offset < span.byte_end);
358
359                    if !in_code_span {
360                        let ref_id = ref_id_match.as_str().trim();
361
362                        // Skip patterns that are likely not markdown references
363                        if !Self::is_likely_not_reference(ref_id) {
364                            let normalized_id = Self::unescape_reference(ref_id).to_lowercase();
365                            usages.insert(normalized_id);
366                        }
367                    }
368                }
369            }
370        }
371
372        // NOTE: The complex recursive loop trying to find references within definitions
373        // has been removed as it's not standard Markdown behavior for finding *usages*.
374        // Usages refer to `[text][ref]`, `![alt][ref]`, `[ref]`, etc., in the main content,
375        // not references potentially embedded within the URL or title of another definition.
376
377        usages
378    }
379
380    /// Get unused references with their line ranges.
381    ///
382    /// This method uses the cached definitions to improve performance.
383    ///
384    /// Note: References that are only used inside code blocks are still considered unused,
385    /// as code blocks are treated as examples or documentation rather than actual content.
386    fn get_unused_references(
387        &self,
388        definitions: &HashMap<String, Vec<(usize, usize)>>,
389        usages: &HashSet<String>,
390    ) -> Vec<(String, usize, usize)> {
391        let mut unused = Vec::new();
392        for (id, ranges) in definitions {
393            // If this id is not used anywhere and is not in the ignored list
394            if !usages.contains(id) && !self.is_ignored_definition(id) {
395                // Only report as unused if there's exactly one definition
396                // Multiple definitions are already reported as duplicates
397                if ranges.len() == 1 {
398                    let (start, end) = ranges[0];
399                    unused.push((id.clone(), start, end));
400                }
401                // If there are multiple definitions (duplicates), don't report them as unused
402                // They're already being reported as duplicate definitions
403            }
404        }
405        unused
406    }
407
408    /// Check if a definition should be ignored (kept even if unused)
409    fn is_ignored_definition(&self, definition_id: &str) -> bool {
410        self.config
411            .ignored_definitions
412            .iter()
413            .any(|ignored| ignored.eq_ignore_ascii_case(definition_id))
414    }
415}
416
417impl Default for MD053LinkImageReferenceDefinitions {
418    fn default() -> Self {
419        Self::new()
420    }
421}
422
423impl Rule for MD053LinkImageReferenceDefinitions {
424    fn name(&self) -> &'static str {
425        "MD053"
426    }
427
428    fn description(&self) -> &'static str {
429        "Link and image reference definitions should be needed"
430    }
431
432    /// Check the content for unused and duplicate link/image reference definitions.
433    ///
434    /// This implementation uses caching for improved performance on large documents.
435    fn check(&self, ctx: &crate::lint_context::LintContext) -> LintResult {
436        // Find definitions and usages using LintContext
437        let definitions = self.find_definitions(ctx);
438        let usages = self.find_usages(ctx);
439
440        // Get unused references by comparing definitions and usages
441        let unused_refs = self.get_unused_references(&definitions, &usages);
442
443        let mut warnings = Vec::new();
444
445        // Check for duplicate definitions (case-insensitive per CommonMark spec)
446        let mut seen_definitions: HashMap<String, (String, usize)> = HashMap::new(); // lowercase -> (original, first_line)
447
448        for (definition_id, ranges) in &definitions {
449            // Skip ignored definitions for duplicate checking
450            if self.is_ignored_definition(definition_id) {
451                continue;
452            }
453
454            if ranges.len() > 1 {
455                // Multiple definitions with exact same ID (already lowercase)
456                for (i, &(start_line, _)) in ranges.iter().enumerate() {
457                    if i > 0 {
458                        // Skip the first occurrence, report all others
459                        let line_num = start_line + 1;
460                        let line_content = ctx.lines.get(start_line).map(|l| l.content.as_str()).unwrap_or("");
461                        let (start_line_1idx, start_col, end_line, end_col) =
462                            calculate_line_range(line_num, line_content);
463
464                        warnings.push(LintWarning {
465                            rule_name: Some(self.name().to_string()),
466                            line: start_line_1idx,
467                            column: start_col,
468                            end_line,
469                            end_column: end_col,
470                            message: format!("Duplicate link or image reference definition: [{definition_id}]"),
471                            severity: Severity::Warning,
472                            fix: None,
473                        });
474                    }
475                }
476            }
477
478            // Track for case-variant duplicates
479            if let Some(&(start_line, _)) = ranges.first() {
480                // Find the original case version from the line
481                if let Some(line_info) = ctx.lines.get(start_line)
482                    && let Some(caps) = REFERENCE_DEFINITION_REGEX.captures(&line_info.content)
483                {
484                    let original_id = caps.get(1).unwrap().as_str().trim();
485                    let lower_id = original_id.to_lowercase();
486
487                    if let Some((first_original, first_line)) = seen_definitions.get(&lower_id) {
488                        // Found a case-variant duplicate
489                        if first_original != original_id {
490                            let line_num = start_line + 1;
491                            let line_content = &line_info.content;
492                            let (start_line_1idx, start_col, end_line, end_col) =
493                                calculate_line_range(line_num, line_content);
494
495                            warnings.push(LintWarning {
496                                    rule_name: Some(self.name().to_string()),
497                                    line: start_line_1idx,
498                                    column: start_col,
499                                    end_line,
500                                    end_column: end_col,
501                                    message: format!("Duplicate link or image reference definition: [{}] (conflicts with [{}] on line {})",
502                                                   original_id, first_original, first_line + 1),
503                                    severity: Severity::Warning,
504                                    fix: None,
505                                });
506                        }
507                    } else {
508                        seen_definitions.insert(lower_id, (original_id.to_string(), start_line));
509                    }
510                }
511            }
512        }
513
514        // Create warnings for unused references
515        for (definition, start, _end) in unused_refs {
516            let line_num = start + 1; // 1-indexed line numbers
517            let line_content = ctx.lines.get(start).map(|l| l.content.as_str()).unwrap_or("");
518
519            // Calculate precise character range for the entire reference definition line
520            let (start_line, start_col, end_line, end_col) = calculate_line_range(line_num, line_content);
521
522            warnings.push(LintWarning {
523                rule_name: Some(self.name().to_string()),
524                line: start_line,
525                column: start_col,
526                end_line,
527                end_column: end_col,
528                message: format!("Unused link/image reference: [{definition}]"),
529                severity: Severity::Warning,
530                fix: None, // MD053 is warning-only, no automatic fixes
531            });
532        }
533
534        Ok(warnings)
535    }
536
537    /// MD053 does not provide automatic fixes
538    fn fix(&self, ctx: &crate::lint_context::LintContext) -> Result<String, LintError> {
539        // This rule is warning-only, no automatic fixes provided
540        Ok(ctx.content.to_string())
541    }
542
543    /// Check if this rule should be skipped for performance
544    fn should_skip(&self, ctx: &crate::lint_context::LintContext) -> bool {
545        // Skip if content is empty or has no links/images
546        ctx.content.is_empty() || !ctx.likely_has_links_or_images()
547    }
548
549    fn as_any(&self) -> &dyn std::any::Any {
550        self
551    }
552
553    fn default_config_section(&self) -> Option<(String, toml::Value)> {
554        let default_config = MD053Config::default();
555        let json_value = serde_json::to_value(&default_config).ok()?;
556        let toml_value = crate::rule_config_serde::json_to_toml_value(&json_value)?;
557        if let toml::Value::Table(table) = toml_value {
558            if !table.is_empty() {
559                Some((MD053Config::RULE_NAME.to_string(), toml::Value::Table(table)))
560            } else {
561                None
562            }
563        } else {
564            None
565        }
566    }
567
568    fn from_config(config: &crate::config::Config) -> Box<dyn Rule>
569    where
570        Self: Sized,
571    {
572        let rule_config = crate::rule_config_serde::load_rule_config::<MD053Config>(config);
573        Box::new(MD053LinkImageReferenceDefinitions::from_config_struct(rule_config))
574    }
575}
576
577#[cfg(test)]
578mod tests {
579    use super::*;
580    use crate::lint_context::LintContext;
581
582    #[test]
583    fn test_used_reference_link() {
584        let rule = MD053LinkImageReferenceDefinitions::new();
585        let content = "[text][ref]\n\n[ref]: https://example.com";
586        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
587        let result = rule.check(&ctx).unwrap();
588
589        assert_eq!(result.len(), 0);
590    }
591
592    #[test]
593    fn test_unused_reference_definition() {
594        let rule = MD053LinkImageReferenceDefinitions::new();
595        let content = "[unused]: https://example.com";
596        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
597        let result = rule.check(&ctx).unwrap();
598
599        assert_eq!(result.len(), 1);
600        assert!(result[0].message.contains("Unused link/image reference: [unused]"));
601    }
602
603    #[test]
604    fn test_used_reference_image() {
605        let rule = MD053LinkImageReferenceDefinitions::new();
606        let content = "![alt][img]\n\n[img]: image.jpg";
607        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
608        let result = rule.check(&ctx).unwrap();
609
610        assert_eq!(result.len(), 0);
611    }
612
613    #[test]
614    fn test_case_insensitive_matching() {
615        let rule = MD053LinkImageReferenceDefinitions::new();
616        let content = "[Text][REF]\n\n[ref]: https://example.com";
617        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
618        let result = rule.check(&ctx).unwrap();
619
620        assert_eq!(result.len(), 0);
621    }
622
623    #[test]
624    fn test_shortcut_reference() {
625        let rule = MD053LinkImageReferenceDefinitions::new();
626        let content = "[ref]\n\n[ref]: https://example.com";
627        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
628        let result = rule.check(&ctx).unwrap();
629
630        assert_eq!(result.len(), 0);
631    }
632
633    #[test]
634    fn test_collapsed_reference() {
635        let rule = MD053LinkImageReferenceDefinitions::new();
636        let content = "[ref][]\n\n[ref]: https://example.com";
637        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
638        let result = rule.check(&ctx).unwrap();
639
640        assert_eq!(result.len(), 0);
641    }
642
643    #[test]
644    fn test_multiple_unused_definitions() {
645        let rule = MD053LinkImageReferenceDefinitions::new();
646        let content = "[unused1]: url1\n[unused2]: url2\n[unused3]: url3";
647        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
648        let result = rule.check(&ctx).unwrap();
649
650        assert_eq!(result.len(), 3);
651
652        // The warnings might not be in the same order, so collect all messages
653        let messages: Vec<String> = result.iter().map(|w| w.message.clone()).collect();
654        assert!(messages.iter().any(|m| m.contains("unused1")));
655        assert!(messages.iter().any(|m| m.contains("unused2")));
656        assert!(messages.iter().any(|m| m.contains("unused3")));
657    }
658
659    #[test]
660    fn test_mixed_used_and_unused() {
661        let rule = MD053LinkImageReferenceDefinitions::new();
662        let content = "[used]\n\n[used]: url1\n[unused]: url2";
663        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
664        let result = rule.check(&ctx).unwrap();
665
666        assert_eq!(result.len(), 1);
667        assert!(result[0].message.contains("unused"));
668    }
669
670    #[test]
671    fn test_multiline_definition() {
672        let rule = MD053LinkImageReferenceDefinitions::new();
673        let content = "[ref]: https://example.com\n  \"Title on next line\"";
674        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
675        let result = rule.check(&ctx).unwrap();
676
677        assert_eq!(result.len(), 1); // Still unused
678    }
679
680    #[test]
681    fn test_reference_in_code_block() {
682        let rule = MD053LinkImageReferenceDefinitions::new();
683        let content = "```\n[ref]\n```\n\n[ref]: https://example.com";
684        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
685        let result = rule.check(&ctx).unwrap();
686
687        // Reference used only in code block is still considered unused
688        assert_eq!(result.len(), 1);
689    }
690
691    #[test]
692    fn test_reference_in_inline_code() {
693        let rule = MD053LinkImageReferenceDefinitions::new();
694        let content = "`[ref]`\n\n[ref]: https://example.com";
695        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
696        let result = rule.check(&ctx).unwrap();
697
698        // Reference in inline code is not a usage
699        assert_eq!(result.len(), 1);
700    }
701
702    #[test]
703    fn test_escaped_reference() {
704        let rule = MD053LinkImageReferenceDefinitions::new();
705        let content = "[example\\-ref]\n\n[example-ref]: https://example.com";
706        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
707        let result = rule.check(&ctx).unwrap();
708
709        // Should match despite escaping
710        assert_eq!(result.len(), 0);
711    }
712
713    #[test]
714    fn test_duplicate_definitions() {
715        let rule = MD053LinkImageReferenceDefinitions::new();
716        let content = "[ref]: url1\n[ref]: url2\n\n[ref]";
717        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
718        let result = rule.check(&ctx).unwrap();
719
720        // Should flag the duplicate definition even though it's used (matches markdownlint)
721        assert_eq!(result.len(), 1);
722    }
723
724    #[test]
725    fn test_fix_returns_original() {
726        // MD053 is warning-only, fix should return original content
727        let rule = MD053LinkImageReferenceDefinitions::new();
728        let content = "[used]\n\n[used]: url1\n[unused]: url2\n\nMore content";
729        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
730        let fixed = rule.fix(&ctx).unwrap();
731
732        assert_eq!(fixed, content);
733    }
734
735    #[test]
736    fn test_fix_preserves_content() {
737        // MD053 is warning-only, fix should preserve all content
738        let rule = MD053LinkImageReferenceDefinitions::new();
739        let content = "Content\n\n[unused]: url\n\nMore content";
740        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
741        let fixed = rule.fix(&ctx).unwrap();
742
743        assert_eq!(fixed, content);
744    }
745
746    #[test]
747    fn test_fix_does_not_remove() {
748        // MD053 is warning-only, fix should not remove anything
749        let rule = MD053LinkImageReferenceDefinitions::new();
750        let content = "[unused1]: url1\n[unused2]: url2\n[unused3]: url3";
751        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
752        let fixed = rule.fix(&ctx).unwrap();
753
754        assert_eq!(fixed, content);
755    }
756
757    #[test]
758    fn test_special_characters_in_reference() {
759        let rule = MD053LinkImageReferenceDefinitions::new();
760        let content = "[ref-with_special.chars]\n\n[ref-with_special.chars]: url";
761        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
762        let result = rule.check(&ctx).unwrap();
763
764        assert_eq!(result.len(), 0);
765    }
766
767    #[test]
768    fn test_find_definitions() {
769        let rule = MD053LinkImageReferenceDefinitions::new();
770        let content = "[ref1]: url1\n[ref2]: url2\nSome text\n[ref3]: url3";
771        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
772        let defs = rule.find_definitions(&ctx);
773
774        assert_eq!(defs.len(), 3);
775        assert!(defs.contains_key("ref1"));
776        assert!(defs.contains_key("ref2"));
777        assert!(defs.contains_key("ref3"));
778    }
779
780    #[test]
781    fn test_find_usages() {
782        let rule = MD053LinkImageReferenceDefinitions::new();
783        let content = "[text][ref1] and [ref2] and ![img][ref3]";
784        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
785        let usages = rule.find_usages(&ctx);
786
787        assert!(usages.contains("ref1"));
788        assert!(usages.contains("ref2"));
789        assert!(usages.contains("ref3"));
790    }
791
792    #[test]
793    fn test_ignored_definitions_config() {
794        // Test with ignored definitions
795        let config = MD053Config {
796            ignored_definitions: vec!["todo".to_string(), "draft".to_string()],
797        };
798        let rule = MD053LinkImageReferenceDefinitions::from_config_struct(config);
799
800        let content = "[todo]: https://example.com/todo\n[draft]: https://example.com/draft\n[unused]: https://example.com/unused";
801        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
802        let result = rule.check(&ctx).unwrap();
803
804        // Should only flag "unused", not "todo" or "draft"
805        assert_eq!(result.len(), 1);
806        assert!(result[0].message.contains("unused"));
807        assert!(!result[0].message.contains("todo"));
808        assert!(!result[0].message.contains("draft"));
809    }
810
811    #[test]
812    fn test_ignored_definitions_case_insensitive() {
813        // Test case-insensitive matching of ignored definitions
814        let config = MD053Config {
815            ignored_definitions: vec!["TODO".to_string()],
816        };
817        let rule = MD053LinkImageReferenceDefinitions::from_config_struct(config);
818
819        let content = "[todo]: https://example.com/todo\n[unused]: https://example.com/unused";
820        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
821        let result = rule.check(&ctx).unwrap();
822
823        // Should only flag "unused", not "todo" (matches "TODO" case-insensitively)
824        assert_eq!(result.len(), 1);
825        assert!(result[0].message.contains("unused"));
826        assert!(!result[0].message.contains("todo"));
827    }
828
829    #[test]
830    fn test_default_config_section() {
831        let rule = MD053LinkImageReferenceDefinitions::default();
832        let config_section = rule.default_config_section();
833
834        assert!(config_section.is_some());
835        let (name, value) = config_section.unwrap();
836        assert_eq!(name, "MD053");
837
838        // Should contain the ignored_definitions option with default empty array
839        if let toml::Value::Table(table) = value {
840            assert!(table.contains_key("ignored-definitions"));
841            assert_eq!(table["ignored-definitions"], toml::Value::Array(vec![]));
842        } else {
843            panic!("Expected TOML table");
844        }
845    }
846
847    #[test]
848    fn test_fix_with_ignored_definitions() {
849        // MD053 is warning-only, fix should not remove anything even with ignored definitions
850        let config = MD053Config {
851            ignored_definitions: vec!["template".to_string()],
852        };
853        let rule = MD053LinkImageReferenceDefinitions::from_config_struct(config);
854
855        let content = "[template]: https://example.com/template\n[unused]: https://example.com/unused\n\nSome content.";
856        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
857        let fixed = rule.fix(&ctx).unwrap();
858
859        // Should keep everything since MD053 doesn't fix
860        assert_eq!(fixed, content);
861    }
862
863    #[test]
864    fn test_duplicate_definitions_exact_case() {
865        let rule = MD053LinkImageReferenceDefinitions::new();
866        let content = "[ref]: url1\n[ref]: url2\n[ref]: url3";
867        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
868        let result = rule.check(&ctx).unwrap();
869
870        // Should have 2 duplicate warnings (for the 2nd and 3rd definitions)
871        // Plus 1 unused warning
872        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
873        assert_eq!(duplicate_warnings.len(), 2);
874        assert_eq!(duplicate_warnings[0].line, 2);
875        assert_eq!(duplicate_warnings[1].line, 3);
876    }
877
878    #[test]
879    fn test_duplicate_definitions_case_variants() {
880        let rule = MD053LinkImageReferenceDefinitions::new();
881        let content =
882            "[method resolution order]: url1\n[Method Resolution Order]: url2\n[METHOD RESOLUTION ORDER]: url3";
883        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
884        let result = rule.check(&ctx).unwrap();
885
886        // Should have 2 duplicate warnings (for the 2nd and 3rd definitions)
887        // Note: These are treated as exact duplicates since they normalize to the same ID
888        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
889        assert_eq!(duplicate_warnings.len(), 2);
890
891        // The exact duplicate messages don't include "conflicts with"
892        // Only case-variant duplicates with different normalized forms would
893        assert_eq!(duplicate_warnings[0].line, 2);
894        assert_eq!(duplicate_warnings[1].line, 3);
895    }
896
897    #[test]
898    fn test_duplicate_and_unused() {
899        let rule = MD053LinkImageReferenceDefinitions::new();
900        let content = "[used]\n[used]: url1\n[used]: url2\n[unused]: url3";
901        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
902        let result = rule.check(&ctx).unwrap();
903
904        // Should have 1 duplicate warning and 1 unused warning
905        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
906        let unused_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Unused")).collect();
907
908        assert_eq!(duplicate_warnings.len(), 1);
909        assert_eq!(unused_warnings.len(), 1);
910        assert_eq!(duplicate_warnings[0].line, 3); // Second [used] definition
911        assert_eq!(unused_warnings[0].line, 4); // [unused] definition
912    }
913
914    #[test]
915    fn test_duplicate_with_usage() {
916        let rule = MD053LinkImageReferenceDefinitions::new();
917        // Even if used, duplicates should still be reported
918        let content = "[ref]\n\n[ref]: url1\n[ref]: url2";
919        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
920        let result = rule.check(&ctx).unwrap();
921
922        // Should have 1 duplicate warning (no unused since it's referenced)
923        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
924        let unused_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Unused")).collect();
925
926        assert_eq!(duplicate_warnings.len(), 1);
927        assert_eq!(unused_warnings.len(), 0);
928        assert_eq!(duplicate_warnings[0].line, 4);
929    }
930
931    #[test]
932    fn test_no_duplicate_different_ids() {
933        let rule = MD053LinkImageReferenceDefinitions::new();
934        let content = "[ref1]: url1\n[ref2]: url2\n[ref3]: url3";
935        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
936        let result = rule.check(&ctx).unwrap();
937
938        // Should have no duplicate warnings, only unused warnings
939        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
940        assert_eq!(duplicate_warnings.len(), 0);
941    }
942
943    #[test]
944    fn test_comment_style_reference_double_slash() {
945        let rule = MD053LinkImageReferenceDefinitions::new();
946        // Most popular comment pattern: [//]: # (comment)
947        let content = "[//]: # (This is a comment)\n\nSome regular text.";
948        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
949        let result = rule.check(&ctx).unwrap();
950
951        // Should not report as unused - it's recognized as a comment
952        assert_eq!(result.len(), 0, "Comment-style reference [//]: # should not be flagged");
953    }
954
955    #[test]
956    fn test_comment_style_reference_comment_label() {
957        let rule = MD053LinkImageReferenceDefinitions::new();
958        // Semantic comment pattern: [comment]: # (text)
959        let content = "[comment]: # (This is a semantic comment)\n\n[note]: # (This is a note)";
960        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
961        let result = rule.check(&ctx).unwrap();
962
963        // Should not report either as unused
964        assert_eq!(result.len(), 0, "Comment-style references should not be flagged");
965    }
966
967    #[test]
968    fn test_comment_style_reference_todo_fixme() {
969        let rule = MD053LinkImageReferenceDefinitions::new();
970        // Task tracking patterns: [todo]: # and [fixme]: #
971        let content = "[todo]: # (Add more examples)\n[fixme]: # (Fix this later)\n[hack]: # (Temporary workaround)";
972        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
973        let result = rule.check(&ctx).unwrap();
974
975        // Should not report any as unused
976        assert_eq!(result.len(), 0, "TODO/FIXME comment patterns should not be flagged");
977    }
978
979    #[test]
980    fn test_comment_style_reference_fragment_only() {
981        let rule = MD053LinkImageReferenceDefinitions::new();
982        // Any reference with just "#" as URL should be treated as a comment
983        let content = "[anything]: #\n[ref]: #\n\nSome text.";
984        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
985        let result = rule.check(&ctx).unwrap();
986
987        // Should not report as unused - fragment-only URLs are often comments
988        assert_eq!(result.len(), 0, "References with just '#' URL should not be flagged");
989    }
990
991    #[test]
992    fn test_comment_vs_real_reference() {
993        let rule = MD053LinkImageReferenceDefinitions::new();
994        // Mix of comment and real reference - only real one should be flagged if unused
995        let content = "[//]: # (This is a comment)\n[real-ref]: https://example.com\n\nSome text.";
996        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
997        let result = rule.check(&ctx).unwrap();
998
999        // Should only report the real reference as unused
1000        assert_eq!(result.len(), 1, "Only real unused references should be flagged");
1001        assert!(result[0].message.contains("real-ref"), "Should flag the real reference");
1002    }
1003
1004    #[test]
1005    fn test_comment_with_fragment_section() {
1006        let rule = MD053LinkImageReferenceDefinitions::new();
1007        // Comment pattern with a fragment section (still a comment)
1008        let content = "[//]: #section (Comment about section)\n\nSome text.";
1009        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
1010        let result = rule.check(&ctx).unwrap();
1011
1012        // Should not report as unused - it's still a comment pattern
1013        assert_eq!(result.len(), 0, "Comment with fragment section should not be flagged");
1014    }
1015
1016    #[test]
1017    fn test_is_likely_comment_reference_helper() {
1018        // Test the helper function directly
1019        assert!(
1020            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("//", "#"),
1021            "[//]: # should be recognized as comment"
1022        );
1023        assert!(
1024            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("comment", "#section"),
1025            "[comment]: #section should be recognized as comment"
1026        );
1027        assert!(
1028            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("note", "#"),
1029            "[note]: # should be recognized as comment"
1030        );
1031        assert!(
1032            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("todo", "#"),
1033            "[todo]: # should be recognized as comment"
1034        );
1035        assert!(
1036            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("anything", "#"),
1037            "Any label with just '#' should be recognized as comment"
1038        );
1039        assert!(
1040            !MD053LinkImageReferenceDefinitions::is_likely_comment_reference("ref", "https://example.com"),
1041            "Real URL should not be recognized as comment"
1042        );
1043        assert!(
1044            !MD053LinkImageReferenceDefinitions::is_likely_comment_reference("link", "http://test.com"),
1045            "Real URL should not be recognized as comment"
1046        );
1047    }
1048}
rumdl_lib/rules/md053_link_image_reference_definitions.rs

rumdl_lib/rules/
md053_link_image_reference_definitions.rs
