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

rumdl_lib/rules/
md053_link_image_reference_definitions.rs
