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;
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) {
264                    def_start -= 1;
265                }
266
267                if let Some(caps) = REFERENCE_DEFINITION_REGEX.captures(&lines[def_start].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) {
329                continue;
330            }
331
332            // Find potential shortcut references
333            for caps in SHORTCUT_REFERENCE_REGEX.captures_iter(&line_info.content).flatten() {
334                if let Some(full_match) = caps.get(0)
335                    && let Some(ref_id_match) = caps.get(1)
336                {
337                    // Check if the match is within a code span
338                    let match_byte_offset = line_info.byte_offset + full_match.start();
339                    let in_code_span = code_spans
340                        .iter()
341                        .any(|span| match_byte_offset >= span.byte_offset && match_byte_offset < span.byte_end);
342
343                    if !in_code_span {
344                        let ref_id = ref_id_match.as_str().trim();
345
346                        if !Self::should_skip_pattern(ref_id) {
347                            let normalized_id = Self::unescape_reference(ref_id).to_lowercase();
348                            usages.insert(normalized_id);
349                        }
350                    }
351                }
352            }
353        }
354
355        // NOTE: The complex recursive loop trying to find references within definitions
356        // has been removed as it's not standard Markdown behavior for finding *usages*.
357        // Usages refer to `[text][ref]`, `![alt][ref]`, `[ref]`, etc., in the main content,
358        // not references potentially embedded within the URL or title of another definition.
359
360        usages
361    }
362
363    /// Get unused references with their line ranges.
364    ///
365    /// This method uses the cached definitions to improve performance.
366    ///
367    /// Note: References that are only used inside code blocks are still considered unused,
368    /// as code blocks are treated as examples or documentation rather than actual content.
369    fn get_unused_references(
370        &self,
371        definitions: &HashMap<String, Vec<(usize, usize)>>,
372        usages: &HashSet<String>,
373    ) -> Vec<(String, usize, usize)> {
374        let mut unused = Vec::new();
375        for (id, ranges) in definitions {
376            // If this id is not used anywhere and is not in the ignored list
377            if !usages.contains(id) && !self.is_ignored_definition(id) {
378                // Only report as unused if there's exactly one definition
379                // Multiple definitions are already reported as duplicates
380                if ranges.len() == 1 {
381                    let (start, end) = ranges[0];
382                    unused.push((id.clone(), start, end));
383                }
384                // If there are multiple definitions (duplicates), don't report them as unused
385                // They're already being reported as duplicate definitions
386            }
387        }
388        unused
389    }
390
391    /// Check if a definition should be ignored (kept even if unused)
392    fn is_ignored_definition(&self, definition_id: &str) -> bool {
393        self.config
394            .ignored_definitions
395            .iter()
396            .any(|ignored| ignored.eq_ignore_ascii_case(definition_id))
397    }
398}
399
400impl Default for MD053LinkImageReferenceDefinitions {
401    fn default() -> Self {
402        Self::new()
403    }
404}
405
406impl Rule for MD053LinkImageReferenceDefinitions {
407    fn name(&self) -> &'static str {
408        "MD053"
409    }
410
411    fn description(&self) -> &'static str {
412        "Link and image reference definitions should be needed"
413    }
414
415    /// Check the content for unused and duplicate link/image reference definitions.
416    ///
417    /// This implementation uses caching for improved performance on large documents.
418    fn check(&self, ctx: &crate::lint_context::LintContext) -> LintResult {
419        // Find definitions and usages using LintContext
420        let definitions = self.find_definitions(ctx);
421        let usages = self.find_usages(ctx);
422
423        // Get unused references by comparing definitions and usages
424        let unused_refs = self.get_unused_references(&definitions, &usages);
425
426        let mut warnings = Vec::new();
427
428        // Check for duplicate definitions (case-insensitive per CommonMark spec)
429        let mut seen_definitions: HashMap<String, (String, usize)> = HashMap::new(); // lowercase -> (original, first_line)
430
431        for (definition_id, ranges) in &definitions {
432            // Skip ignored definitions for duplicate checking
433            if self.is_ignored_definition(definition_id) {
434                continue;
435            }
436
437            if ranges.len() > 1 {
438                // Multiple definitions with exact same ID (already lowercase)
439                for (i, &(start_line, _)) in ranges.iter().enumerate() {
440                    if i > 0 {
441                        // Skip the first occurrence, report all others
442                        let line_num = start_line + 1;
443                        let line_content = ctx.lines.get(start_line).map(|l| l.content.as_str()).unwrap_or("");
444                        let (start_line_1idx, start_col, end_line, end_col) =
445                            calculate_line_range(line_num, line_content);
446
447                        warnings.push(LintWarning {
448                            rule_name: Some(self.name().to_string()),
449                            line: start_line_1idx,
450                            column: start_col,
451                            end_line,
452                            end_column: end_col,
453                            message: format!("Duplicate link or image reference definition: [{definition_id}]"),
454                            severity: Severity::Warning,
455                            fix: None,
456                        });
457                    }
458                }
459            }
460
461            // Track for case-variant duplicates
462            if let Some(&(start_line, _)) = ranges.first() {
463                // Find the original case version from the line
464                if let Some(line_info) = ctx.lines.get(start_line)
465                    && let Some(caps) = REFERENCE_DEFINITION_REGEX.captures(&line_info.content)
466                {
467                    let original_id = caps.get(1).unwrap().as_str().trim();
468                    let lower_id = original_id.to_lowercase();
469
470                    if let Some((first_original, first_line)) = seen_definitions.get(&lower_id) {
471                        // Found a case-variant duplicate
472                        if first_original != original_id {
473                            let line_num = start_line + 1;
474                            let line_content = &line_info.content;
475                            let (start_line_1idx, start_col, end_line, end_col) =
476                                calculate_line_range(line_num, line_content);
477
478                            warnings.push(LintWarning {
479                                    rule_name: Some(self.name().to_string()),
480                                    line: start_line_1idx,
481                                    column: start_col,
482                                    end_line,
483                                    end_column: end_col,
484                                    message: format!("Duplicate link or image reference definition: [{}] (conflicts with [{}] on line {})",
485                                                   original_id, first_original, first_line + 1),
486                                    severity: Severity::Warning,
487                                    fix: None,
488                                });
489                        }
490                    } else {
491                        seen_definitions.insert(lower_id, (original_id.to_string(), start_line));
492                    }
493                }
494            }
495        }
496
497        // Create warnings for unused references
498        for (definition, start, _end) in unused_refs {
499            let line_num = start + 1; // 1-indexed line numbers
500            let line_content = ctx.lines.get(start).map(|l| l.content.as_str()).unwrap_or("");
501
502            // Calculate precise character range for the entire reference definition line
503            let (start_line, start_col, end_line, end_col) = calculate_line_range(line_num, line_content);
504
505            warnings.push(LintWarning {
506                rule_name: Some(self.name().to_string()),
507                line: start_line,
508                column: start_col,
509                end_line,
510                end_column: end_col,
511                message: format!("Unused link/image reference: [{definition}]"),
512                severity: Severity::Warning,
513                fix: None, // MD053 is warning-only, no automatic fixes
514            });
515        }
516
517        Ok(warnings)
518    }
519
520    /// MD053 does not provide automatic fixes
521    fn fix(&self, ctx: &crate::lint_context::LintContext) -> Result<String, LintError> {
522        // This rule is warning-only, no automatic fixes provided
523        Ok(ctx.content.to_string())
524    }
525
526    /// Check if this rule should be skipped for performance
527    fn should_skip(&self, ctx: &crate::lint_context::LintContext) -> bool {
528        // Skip if content is empty or has no links/images
529        ctx.content.is_empty() || !ctx.likely_has_links_or_images()
530    }
531
532    fn as_any(&self) -> &dyn std::any::Any {
533        self
534    }
535
536    fn default_config_section(&self) -> Option<(String, toml::Value)> {
537        let default_config = MD053Config::default();
538        let json_value = serde_json::to_value(&default_config).ok()?;
539        let toml_value = crate::rule_config_serde::json_to_toml_value(&json_value)?;
540        if let toml::Value::Table(table) = toml_value {
541            if !table.is_empty() {
542                Some((MD053Config::RULE_NAME.to_string(), toml::Value::Table(table)))
543            } else {
544                None
545            }
546        } else {
547            None
548        }
549    }
550
551    fn from_config(config: &crate::config::Config) -> Box<dyn Rule>
552    where
553        Self: Sized,
554    {
555        let rule_config = crate::rule_config_serde::load_rule_config::<MD053Config>(config);
556        Box::new(MD053LinkImageReferenceDefinitions::from_config_struct(rule_config))
557    }
558}
559
560#[cfg(test)]
561mod tests {
562    use super::*;
563    use crate::lint_context::LintContext;
564
565    #[test]
566    fn test_used_reference_link() {
567        let rule = MD053LinkImageReferenceDefinitions::new();
568        let content = "[text][ref]\n\n[ref]: https://example.com";
569        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
570        let result = rule.check(&ctx).unwrap();
571
572        assert_eq!(result.len(), 0);
573    }
574
575    #[test]
576    fn test_unused_reference_definition() {
577        let rule = MD053LinkImageReferenceDefinitions::new();
578        let content = "[unused]: https://example.com";
579        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
580        let result = rule.check(&ctx).unwrap();
581
582        assert_eq!(result.len(), 1);
583        assert!(result[0].message.contains("Unused link/image reference: [unused]"));
584    }
585
586    #[test]
587    fn test_used_reference_image() {
588        let rule = MD053LinkImageReferenceDefinitions::new();
589        let content = "![alt][img]\n\n[img]: image.jpg";
590        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
591        let result = rule.check(&ctx).unwrap();
592
593        assert_eq!(result.len(), 0);
594    }
595
596    #[test]
597    fn test_case_insensitive_matching() {
598        let rule = MD053LinkImageReferenceDefinitions::new();
599        let content = "[Text][REF]\n\n[ref]: https://example.com";
600        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
601        let result = rule.check(&ctx).unwrap();
602
603        assert_eq!(result.len(), 0);
604    }
605
606    #[test]
607    fn test_shortcut_reference() {
608        let rule = MD053LinkImageReferenceDefinitions::new();
609        let content = "[ref]\n\n[ref]: https://example.com";
610        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
611        let result = rule.check(&ctx).unwrap();
612
613        assert_eq!(result.len(), 0);
614    }
615
616    #[test]
617    fn test_collapsed_reference() {
618        let rule = MD053LinkImageReferenceDefinitions::new();
619        let content = "[ref][]\n\n[ref]: https://example.com";
620        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
621        let result = rule.check(&ctx).unwrap();
622
623        assert_eq!(result.len(), 0);
624    }
625
626    #[test]
627    fn test_multiple_unused_definitions() {
628        let rule = MD053LinkImageReferenceDefinitions::new();
629        let content = "[unused1]: url1\n[unused2]: url2\n[unused3]: url3";
630        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
631        let result = rule.check(&ctx).unwrap();
632
633        assert_eq!(result.len(), 3);
634
635        // The warnings might not be in the same order, so collect all messages
636        let messages: Vec<String> = result.iter().map(|w| w.message.clone()).collect();
637        assert!(messages.iter().any(|m| m.contains("unused1")));
638        assert!(messages.iter().any(|m| m.contains("unused2")));
639        assert!(messages.iter().any(|m| m.contains("unused3")));
640    }
641
642    #[test]
643    fn test_mixed_used_and_unused() {
644        let rule = MD053LinkImageReferenceDefinitions::new();
645        let content = "[used]\n\n[used]: url1\n[unused]: url2";
646        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
647        let result = rule.check(&ctx).unwrap();
648
649        assert_eq!(result.len(), 1);
650        assert!(result[0].message.contains("unused"));
651    }
652
653    #[test]
654    fn test_multiline_definition() {
655        let rule = MD053LinkImageReferenceDefinitions::new();
656        let content = "[ref]: https://example.com\n  \"Title on next line\"";
657        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
658        let result = rule.check(&ctx).unwrap();
659
660        assert_eq!(result.len(), 1); // Still unused
661    }
662
663    #[test]
664    fn test_reference_in_code_block() {
665        let rule = MD053LinkImageReferenceDefinitions::new();
666        let content = "```\n[ref]\n```\n\n[ref]: https://example.com";
667        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
668        let result = rule.check(&ctx).unwrap();
669
670        // Reference used only in code block is still considered unused
671        assert_eq!(result.len(), 1);
672    }
673
674    #[test]
675    fn test_reference_in_inline_code() {
676        let rule = MD053LinkImageReferenceDefinitions::new();
677        let content = "`[ref]`\n\n[ref]: https://example.com";
678        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
679        let result = rule.check(&ctx).unwrap();
680
681        // Reference in inline code is not a usage
682        assert_eq!(result.len(), 1);
683    }
684
685    #[test]
686    fn test_escaped_reference() {
687        let rule = MD053LinkImageReferenceDefinitions::new();
688        let content = "[example\\-ref]\n\n[example-ref]: https://example.com";
689        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
690        let result = rule.check(&ctx).unwrap();
691
692        // Should match despite escaping
693        assert_eq!(result.len(), 0);
694    }
695
696    #[test]
697    fn test_duplicate_definitions() {
698        let rule = MD053LinkImageReferenceDefinitions::new();
699        let content = "[ref]: url1\n[ref]: url2\n\n[ref]";
700        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
701        let result = rule.check(&ctx).unwrap();
702
703        // Should flag the duplicate definition even though it's used (matches markdownlint)
704        assert_eq!(result.len(), 1);
705    }
706
707    #[test]
708    fn test_fix_returns_original() {
709        // MD053 is warning-only, fix should return original content
710        let rule = MD053LinkImageReferenceDefinitions::new();
711        let content = "[used]\n\n[used]: url1\n[unused]: url2\n\nMore content";
712        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
713        let fixed = rule.fix(&ctx).unwrap();
714
715        assert_eq!(fixed, content);
716    }
717
718    #[test]
719    fn test_fix_preserves_content() {
720        // MD053 is warning-only, fix should preserve all content
721        let rule = MD053LinkImageReferenceDefinitions::new();
722        let content = "Content\n\n[unused]: url\n\nMore content";
723        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
724        let fixed = rule.fix(&ctx).unwrap();
725
726        assert_eq!(fixed, content);
727    }
728
729    #[test]
730    fn test_fix_does_not_remove() {
731        // MD053 is warning-only, fix should not remove anything
732        let rule = MD053LinkImageReferenceDefinitions::new();
733        let content = "[unused1]: url1\n[unused2]: url2\n[unused3]: url3";
734        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
735        let fixed = rule.fix(&ctx).unwrap();
736
737        assert_eq!(fixed, content);
738    }
739
740    #[test]
741    fn test_special_characters_in_reference() {
742        let rule = MD053LinkImageReferenceDefinitions::new();
743        let content = "[ref-with_special.chars]\n\n[ref-with_special.chars]: url";
744        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
745        let result = rule.check(&ctx).unwrap();
746
747        assert_eq!(result.len(), 0);
748    }
749
750    #[test]
751    fn test_find_definitions() {
752        let rule = MD053LinkImageReferenceDefinitions::new();
753        let content = "[ref1]: url1\n[ref2]: url2\nSome text\n[ref3]: url3";
754        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
755        let defs = rule.find_definitions(&ctx);
756
757        assert_eq!(defs.len(), 3);
758        assert!(defs.contains_key("ref1"));
759        assert!(defs.contains_key("ref2"));
760        assert!(defs.contains_key("ref3"));
761    }
762
763    #[test]
764    fn test_find_usages() {
765        let rule = MD053LinkImageReferenceDefinitions::new();
766        let content = "[text][ref1] and [ref2] and ![img][ref3]";
767        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
768        let usages = rule.find_usages(&ctx);
769
770        assert!(usages.contains("ref1"));
771        assert!(usages.contains("ref2"));
772        assert!(usages.contains("ref3"));
773    }
774
775    #[test]
776    fn test_ignored_definitions_config() {
777        // Test with ignored definitions
778        let config = MD053Config {
779            ignored_definitions: vec!["todo".to_string(), "draft".to_string()],
780        };
781        let rule = MD053LinkImageReferenceDefinitions::from_config_struct(config);
782
783        let content = "[todo]: https://example.com/todo\n[draft]: https://example.com/draft\n[unused]: https://example.com/unused";
784        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
785        let result = rule.check(&ctx).unwrap();
786
787        // Should only flag "unused", not "todo" or "draft"
788        assert_eq!(result.len(), 1);
789        assert!(result[0].message.contains("unused"));
790        assert!(!result[0].message.contains("todo"));
791        assert!(!result[0].message.contains("draft"));
792    }
793
794    #[test]
795    fn test_ignored_definitions_case_insensitive() {
796        // Test case-insensitive matching of ignored definitions
797        let config = MD053Config {
798            ignored_definitions: vec!["TODO".to_string()],
799        };
800        let rule = MD053LinkImageReferenceDefinitions::from_config_struct(config);
801
802        let content = "[todo]: https://example.com/todo\n[unused]: https://example.com/unused";
803        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
804        let result = rule.check(&ctx).unwrap();
805
806        // Should only flag "unused", not "todo" (matches "TODO" case-insensitively)
807        assert_eq!(result.len(), 1);
808        assert!(result[0].message.contains("unused"));
809        assert!(!result[0].message.contains("todo"));
810    }
811
812    #[test]
813    fn test_default_config_section() {
814        let rule = MD053LinkImageReferenceDefinitions::default();
815        let config_section = rule.default_config_section();
816
817        assert!(config_section.is_some());
818        let (name, value) = config_section.unwrap();
819        assert_eq!(name, "MD053");
820
821        // Should contain the ignored_definitions option with default empty array
822        if let toml::Value::Table(table) = value {
823            assert!(table.contains_key("ignored-definitions"));
824            assert_eq!(table["ignored-definitions"], toml::Value::Array(vec![]));
825        } else {
826            panic!("Expected TOML table");
827        }
828    }
829
830    #[test]
831    fn test_fix_with_ignored_definitions() {
832        // MD053 is warning-only, fix should not remove anything even with ignored definitions
833        let config = MD053Config {
834            ignored_definitions: vec!["template".to_string()],
835        };
836        let rule = MD053LinkImageReferenceDefinitions::from_config_struct(config);
837
838        let content = "[template]: https://example.com/template\n[unused]: https://example.com/unused\n\nSome content.";
839        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
840        let fixed = rule.fix(&ctx).unwrap();
841
842        // Should keep everything since MD053 doesn't fix
843        assert_eq!(fixed, content);
844    }
845
846    #[test]
847    fn test_duplicate_definitions_exact_case() {
848        let rule = MD053LinkImageReferenceDefinitions::new();
849        let content = "[ref]: url1\n[ref]: url2\n[ref]: url3";
850        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
851        let result = rule.check(&ctx).unwrap();
852
853        // Should have 2 duplicate warnings (for the 2nd and 3rd definitions)
854        // Plus 1 unused warning
855        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
856        assert_eq!(duplicate_warnings.len(), 2);
857        assert_eq!(duplicate_warnings[0].line, 2);
858        assert_eq!(duplicate_warnings[1].line, 3);
859    }
860
861    #[test]
862    fn test_duplicate_definitions_case_variants() {
863        let rule = MD053LinkImageReferenceDefinitions::new();
864        let content =
865            "[method resolution order]: url1\n[Method Resolution Order]: url2\n[METHOD RESOLUTION ORDER]: url3";
866        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
867        let result = rule.check(&ctx).unwrap();
868
869        // Should have 2 duplicate warnings (for the 2nd and 3rd definitions)
870        // Note: These are treated as exact duplicates since they normalize to the same ID
871        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
872        assert_eq!(duplicate_warnings.len(), 2);
873
874        // The exact duplicate messages don't include "conflicts with"
875        // Only case-variant duplicates with different normalized forms would
876        assert_eq!(duplicate_warnings[0].line, 2);
877        assert_eq!(duplicate_warnings[1].line, 3);
878    }
879
880    #[test]
881    fn test_duplicate_and_unused() {
882        let rule = MD053LinkImageReferenceDefinitions::new();
883        let content = "[used]\n[used]: url1\n[used]: url2\n[unused]: url3";
884        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
885        let result = rule.check(&ctx).unwrap();
886
887        // Should have 1 duplicate warning and 1 unused warning
888        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
889        let unused_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Unused")).collect();
890
891        assert_eq!(duplicate_warnings.len(), 1);
892        assert_eq!(unused_warnings.len(), 1);
893        assert_eq!(duplicate_warnings[0].line, 3); // Second [used] definition
894        assert_eq!(unused_warnings[0].line, 4); // [unused] definition
895    }
896
897    #[test]
898    fn test_duplicate_with_usage() {
899        let rule = MD053LinkImageReferenceDefinitions::new();
900        // Even if used, duplicates should still be reported
901        let content = "[ref]\n\n[ref]: url1\n[ref]: url2";
902        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
903        let result = rule.check(&ctx).unwrap();
904
905        // Should have 1 duplicate warning (no unused since it's referenced)
906        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
907        let unused_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Unused")).collect();
908
909        assert_eq!(duplicate_warnings.len(), 1);
910        assert_eq!(unused_warnings.len(), 0);
911        assert_eq!(duplicate_warnings[0].line, 4);
912    }
913
914    #[test]
915    fn test_no_duplicate_different_ids() {
916        let rule = MD053LinkImageReferenceDefinitions::new();
917        let content = "[ref1]: url1\n[ref2]: url2\n[ref3]: url3";
918        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
919        let result = rule.check(&ctx).unwrap();
920
921        // Should have no duplicate warnings, only unused warnings
922        let duplicate_warnings: Vec<_> = result.iter().filter(|w| w.message.contains("Duplicate")).collect();
923        assert_eq!(duplicate_warnings.len(), 0);
924    }
925
926    #[test]
927    fn test_comment_style_reference_double_slash() {
928        let rule = MD053LinkImageReferenceDefinitions::new();
929        // Most popular comment pattern: [//]: # (comment)
930        let content = "[//]: # (This is a comment)\n\nSome regular text.";
931        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
932        let result = rule.check(&ctx).unwrap();
933
934        // Should not report as unused - it's recognized as a comment
935        assert_eq!(result.len(), 0, "Comment-style reference [//]: # should not be flagged");
936    }
937
938    #[test]
939    fn test_comment_style_reference_comment_label() {
940        let rule = MD053LinkImageReferenceDefinitions::new();
941        // Semantic comment pattern: [comment]: # (text)
942        let content = "[comment]: # (This is a semantic comment)\n\n[note]: # (This is a note)";
943        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
944        let result = rule.check(&ctx).unwrap();
945
946        // Should not report either as unused
947        assert_eq!(result.len(), 0, "Comment-style references should not be flagged");
948    }
949
950    #[test]
951    fn test_comment_style_reference_todo_fixme() {
952        let rule = MD053LinkImageReferenceDefinitions::new();
953        // Task tracking patterns: [todo]: # and [fixme]: #
954        let content = "[todo]: # (Add more examples)\n[fixme]: # (Fix this later)\n[hack]: # (Temporary workaround)";
955        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
956        let result = rule.check(&ctx).unwrap();
957
958        // Should not report any as unused
959        assert_eq!(result.len(), 0, "TODO/FIXME comment patterns should not be flagged");
960    }
961
962    #[test]
963    fn test_comment_style_reference_fragment_only() {
964        let rule = MD053LinkImageReferenceDefinitions::new();
965        // Any reference with just "#" as URL should be treated as a comment
966        let content = "[anything]: #\n[ref]: #\n\nSome text.";
967        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
968        let result = rule.check(&ctx).unwrap();
969
970        // Should not report as unused - fragment-only URLs are often comments
971        assert_eq!(result.len(), 0, "References with just '#' URL should not be flagged");
972    }
973
974    #[test]
975    fn test_comment_vs_real_reference() {
976        let rule = MD053LinkImageReferenceDefinitions::new();
977        // Mix of comment and real reference - only real one should be flagged if unused
978        let content = "[//]: # (This is a comment)\n[real-ref]: https://example.com\n\nSome text.";
979        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
980        let result = rule.check(&ctx).unwrap();
981
982        // Should only report the real reference as unused
983        assert_eq!(result.len(), 1, "Only real unused references should be flagged");
984        assert!(result[0].message.contains("real-ref"), "Should flag the real reference");
985    }
986
987    #[test]
988    fn test_comment_with_fragment_section() {
989        let rule = MD053LinkImageReferenceDefinitions::new();
990        // Comment pattern with a fragment section (still a comment)
991        let content = "[//]: #section (Comment about section)\n\nSome text.";
992        let ctx = LintContext::new(content, crate::config::MarkdownFlavor::Standard);
993        let result = rule.check(&ctx).unwrap();
994
995        // Should not report as unused - it's still a comment pattern
996        assert_eq!(result.len(), 0, "Comment with fragment section should not be flagged");
997    }
998
999    #[test]
1000    fn test_is_likely_comment_reference_helper() {
1001        // Test the helper function directly
1002        assert!(
1003            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("//", "#"),
1004            "[//]: # should be recognized as comment"
1005        );
1006        assert!(
1007            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("comment", "#section"),
1008            "[comment]: #section should be recognized as comment"
1009        );
1010        assert!(
1011            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("note", "#"),
1012            "[note]: # should be recognized as comment"
1013        );
1014        assert!(
1015            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("todo", "#"),
1016            "[todo]: # should be recognized as comment"
1017        );
1018        assert!(
1019            MD053LinkImageReferenceDefinitions::is_likely_comment_reference("anything", "#"),
1020            "Any label with just '#' should be recognized as comment"
1021        );
1022        assert!(
1023            !MD053LinkImageReferenceDefinitions::is_likely_comment_reference("ref", "https://example.com"),
1024            "Real URL should not be recognized as comment"
1025        );
1026        assert!(
1027            !MD053LinkImageReferenceDefinitions::is_likely_comment_reference("link", "http://test.com"),
1028            "Real URL should not be recognized as comment"
1029        );
1030    }
1031}
rumdl_lib/rules/md053_link_image_reference_definitions.rs

rumdl_lib/rules/
md053_link_image_reference_definitions.rs
