rumdl_lib/rules/
md057_existing_relative_links.rs

1//!
2//! Rule MD057: Existing relative links
3//!
4//! See [docs/md057.md](../../docs/md057.md) for full documentation, configuration, and examples.
5
6use crate::rule::{LintError, LintResult, LintWarning, Rule, RuleCategory, Severity};
7use crate::utils::element_cache::ElementCache;
8use lazy_static::lazy_static;
9use regex::Regex;
10use std::collections::HashMap;
11use std::env;
12use std::path::{Path, PathBuf};
13use std::sync::{Arc, Mutex};
14
15mod md057_config;
16use md057_config::MD057Config;
17
18// Thread-safe cache for file existence checks to avoid redundant filesystem operations
19lazy_static! {
20    static ref FILE_EXISTENCE_CACHE: Arc<Mutex<HashMap<PathBuf, bool>>> = Arc::new(Mutex::new(HashMap::new()));
21}
22
23// Reset the file existence cache (typically between rule runs)
24fn reset_file_existence_cache() {
25    let mut cache = FILE_EXISTENCE_CACHE.lock().unwrap();
26    cache.clear();
27}
28
29// Check if a file exists with caching
30fn file_exists_with_cache(path: &Path) -> bool {
31    let mut cache = FILE_EXISTENCE_CACHE.lock().unwrap();
32    *cache.entry(path.to_path_buf()).or_insert_with(|| path.exists())
33}
34
35lazy_static! {
36    // Regex to match the start of a link - simplified for performance
37    static ref LINK_START_REGEX: Regex =
38        Regex::new(r"!?\[[^\]]*\]").unwrap();
39
40    /// Regex to extract the URL from a markdown link
41    /// Format: `](URL)` or `](URL "title")`
42    static ref URL_EXTRACT_REGEX: Regex =
43        Regex::new("\\]\\(\\s*<?([^>\\)\\s#]+)(#[^)\\s]*)?\\s*(?:\"[^\"]*\")?\\s*>?\\s*\\)").unwrap();
44
45    /// Regex to detect code fence blocks
46    static ref CODE_FENCE_REGEX: Regex =
47        Regex::new(r"^( {0,3})(`{3,}|~{3,})").unwrap();
48
49    /// Regex to detect protocol and domain for external links
50    static ref PROTOCOL_DOMAIN_REGEX: Regex =
51        Regex::new(r"^(https?://|ftp://|mailto:|www\.)").unwrap();
52
53    /// Regex to detect media file types
54    static ref MEDIA_FILE_REGEX: Regex =
55        Regex::new(r"\.(jpg|jpeg|png|gif|bmp|svg|webp|tiff|mp3|mp4|avi|mov|webm|wav|ogg|pdf)$").unwrap();
56
57    /// Regex to detect fragment-only links
58    static ref FRAGMENT_ONLY_REGEX: Regex =
59        Regex::new(r"^#").unwrap();
60
61    // Current working directory
62    static ref CURRENT_DIR: PathBuf = env::current_dir().unwrap_or_else(|_| PathBuf::from("."));
63}
64
65/// Rule MD057: Existing relative links should point to valid files or directories.
66#[derive(Debug, Default, Clone)]
67pub struct MD057ExistingRelativeLinks {
68    /// Base directory for resolving relative links
69    base_path: Arc<Mutex<Option<PathBuf>>>,
70    /// Configuration
71    config: MD057Config,
72}
73
74impl MD057ExistingRelativeLinks {
75    /// Create a new instance with default settings
76    pub fn new() -> Self {
77        Self::default()
78    }
79
80    /// Set the base path for resolving relative links
81    pub fn with_path<P: AsRef<Path>>(self, path: P) -> Self {
82        let path = path.as_ref();
83        let dir_path = if path.is_file() {
84            path.parent().map(|p| p.to_path_buf())
85        } else {
86            Some(path.to_path_buf())
87        };
88
89        *self.base_path.lock().unwrap() = dir_path;
90        self
91    }
92
93    /// Configure whether to skip checking media files
94    pub fn with_skip_media_files(mut self, skip_media_files: bool) -> Self {
95        self.config.skip_media_files = skip_media_files;
96        self
97    }
98
99    pub fn from_config_struct(config: MD057Config) -> Self {
100        Self {
101            base_path: Arc::new(Mutex::new(None)),
102            config,
103        }
104    }
105
106    /// Check if a URL is external (optimized version)
107    #[inline]
108    fn is_external_url(&self, url: &str) -> bool {
109        if url.is_empty() {
110            return false;
111        }
112
113        // Quick checks for common external URL patterns
114        if PROTOCOL_DOMAIN_REGEX.is_match(url) || url.starts_with("www.") {
115            return true;
116        }
117
118        // More restrictive domain check using a simpler pattern
119        if !self.is_media_file(url) && url.ends_with(".com") {
120            return true;
121        }
122
123        // Absolute paths within the site are not external
124        if url.starts_with('/') {
125            return false;
126        }
127
128        // All other cases (relative paths, etc.) are not external
129        false
130    }
131
132    /// Check if the URL is a fragment-only link (internal document link)
133    #[inline]
134    fn is_fragment_only_link(&self, url: &str) -> bool {
135        url.starts_with('#')
136    }
137
138    /// Check if the URL has a media file extension (optimized with early returns)
139    #[inline]
140    fn is_media_file(&self, url: &str) -> bool {
141        // Quick check before using regex
142        if !url.contains('.') {
143            return false;
144        }
145        MEDIA_FILE_REGEX.is_match(url)
146    }
147
148    /// Determine if we should skip checking this media file
149    #[inline]
150    fn should_skip_media_file(&self, url: &str) -> bool {
151        self.config.skip_media_files && self.is_media_file(url)
152    }
153
154    /// Resolve a relative link against the base path
155    fn resolve_link_path(&self, link: &str) -> Option<PathBuf> {
156        self.base_path
157            .lock()
158            .unwrap()
159            .as_ref()
160            .map(|base_path| base_path.join(link))
161    }
162
163    /// Process a single link and check if it exists
164    fn process_link(&self, url: &str, line_num: usize, column: usize, warnings: &mut Vec<LintWarning>) {
165        // Skip empty URLs
166        if url.is_empty() {
167            return;
168        }
169
170        // Skip external URLs and fragment-only links (optimized order)
171        if self.is_external_url(url) || self.is_fragment_only_link(url) {
172            return;
173        }
174
175        // Skip media files if configured to do so
176        if self.should_skip_media_file(url) {
177            return;
178        }
179
180        // Resolve the relative link against the base path
181        if let Some(resolved_path) = self.resolve_link_path(url) {
182            // Check if the file exists (with caching to avoid filesystem calls)
183            if !file_exists_with_cache(&resolved_path) {
184                warnings.push(LintWarning {
185                    rule_name: Some(self.name().to_string()),
186                    line: line_num,
187                    column,
188                    end_line: line_num,
189                    end_column: column + url.len(),
190                    message: format!("Relative link '{url}' does not exist"),
191                    severity: Severity::Warning,
192                    fix: None, // No automatic fix for missing files
193                });
194            }
195        }
196    }
197}
198
199impl Rule for MD057ExistingRelativeLinks {
200    fn name(&self) -> &'static str {
201        "MD057"
202    }
203
204    fn description(&self) -> &'static str {
205        "Relative links should point to existing files"
206    }
207
208    fn category(&self) -> RuleCategory {
209        RuleCategory::Link
210    }
211
212    fn should_skip(&self, ctx: &crate::lint_context::LintContext) -> bool {
213        ctx.content.is_empty() || !ctx.likely_has_links_or_images()
214    }
215
216    fn check(&self, ctx: &crate::lint_context::LintContext) -> LintResult {
217        let content = ctx.content;
218
219        // Early returns for performance
220        if content.is_empty() || !content.contains('[') {
221            return Ok(Vec::new());
222        }
223
224        // Quick check for any potential links before expensive operations
225        if !content.contains("](") {
226            return Ok(Vec::new());
227        }
228
229        // Reset the file existence cache for a fresh run
230        reset_file_existence_cache();
231
232        let mut warnings = Vec::new();
233
234        // Cache base path lookup to avoid repeated mutex operations
235        let base_path = {
236            let base_path_guard = self.base_path.lock().unwrap();
237            if base_path_guard.is_some() {
238                base_path_guard.clone()
239            } else {
240                // Try to determine the base path from the file being processed (cached)
241                static CACHED_FILE_PATH: std::sync::OnceLock<Option<PathBuf>> = std::sync::OnceLock::new();
242                CACHED_FILE_PATH
243                    .get_or_init(|| {
244                        if let Ok(file_path) = env::var("RUMDL_FILE_PATH") {
245                            let path = Path::new(&file_path);
246                            if path.exists() {
247                                path.parent()
248                                    .map(|p| p.to_path_buf())
249                                    .or_else(|| Some(CURRENT_DIR.clone()))
250                            } else {
251                                Some(CURRENT_DIR.clone())
252                            }
253                        } else {
254                            Some(CURRENT_DIR.clone())
255                        }
256                    })
257                    .clone()
258            }
259        };
260
261        // If we still don't have a base path, we can't validate relative links
262        if base_path.is_none() {
263            return Ok(warnings);
264        }
265
266        // Use LintContext links instead of expensive regex parsing
267        if !ctx.links.is_empty() {
268            // Use LineIndex for correct position calculation across all line ending types
269            let line_index = &ctx.line_index;
270
271            // Create element cache once for all links
272            let element_cache = ElementCache::new(content);
273
274            // Pre-collect lines to avoid repeated line iteration
275            let lines: Vec<&str> = content.lines().collect();
276
277            for link in &ctx.links {
278                let line_idx = link.line - 1;
279                if line_idx >= lines.len() {
280                    continue;
281                }
282
283                let line = lines[line_idx];
284
285                // Quick check for link pattern in this line
286                if !line.contains("](") {
287                    continue;
288                }
289
290                // Find all links in this line using optimized regex
291                for link_match in LINK_START_REGEX.find_iter(line) {
292                    let start_pos = link_match.start();
293                    let end_pos = link_match.end();
294
295                    // Calculate absolute position using LineIndex
296                    let line_start_byte = line_index.get_line_start_byte(line_idx + 1).unwrap_or(0);
297                    let absolute_start_pos = line_start_byte + start_pos;
298
299                    // Skip if this link is in a code span
300                    if element_cache.is_in_code_span(absolute_start_pos) {
301                        continue;
302                    }
303
304                    // Find the URL part after the link text
305                    if let Some(caps) = URL_EXTRACT_REGEX.captures_at(line, end_pos - 1)
306                        && let Some(url_group) = caps.get(1)
307                    {
308                        let url = url_group.as_str().trim();
309
310                        // Calculate column position
311                        let column = start_pos + 1;
312
313                        // Process and validate the link
314                        self.process_link(url, link.line, column, &mut warnings);
315                    }
316                }
317            }
318        }
319
320        Ok(warnings)
321    }
322
323    fn fix(&self, ctx: &crate::lint_context::LintContext) -> Result<String, LintError> {
324        Ok(ctx.content.to_string())
325    }
326
327    fn as_any(&self) -> &dyn std::any::Any {
328        self
329    }
330
331    fn default_config_section(&self) -> Option<(String, toml::Value)> {
332        let json_value = serde_json::to_value(&self.config).ok()?;
333        Some((
334            self.name().to_string(),
335            crate::rule_config_serde::json_to_toml_value(&json_value)?,
336        ))
337    }
338
339    fn from_config(config: &crate::config::Config) -> Box<dyn Rule>
340    where
341        Self: Sized,
342    {
343        let rule_config = crate::rule_config_serde::load_rule_config::<MD057Config>(config);
344        Box::new(Self::from_config_struct(rule_config))
345    }
346}
347
348#[cfg(test)]
349mod tests {
350    use super::*;
351    use std::fs::File;
352    use std::io::Write;
353    use tempfile::tempdir;
354
355    #[test]
356    fn test_external_urls() {
357        let rule = MD057ExistingRelativeLinks::new();
358
359        assert!(rule.is_external_url("https://example.com"));
360        assert!(rule.is_external_url("http://example.com"));
361        assert!(rule.is_external_url("ftp://example.com"));
362        assert!(rule.is_external_url("www.example.com"));
363        assert!(rule.is_external_url("example.com"));
364
365        assert!(!rule.is_external_url("./relative/path.md"));
366        assert!(!rule.is_external_url("relative/path.md"));
367        assert!(!rule.is_external_url("../parent/path.md"));
368    }
369
370    #[test]
371    fn test_media_files() {
372        // Test with default settings (skip_media_files = true)
373        let rule_default = MD057ExistingRelativeLinks::new();
374
375        // Test media file identification
376        assert!(
377            rule_default.is_media_file("image.jpg"),
378            "image.jpg should be identified as a media file"
379        );
380        assert!(
381            rule_default.is_media_file("video.mp4"),
382            "video.mp4 should be identified as a media file"
383        );
384        assert!(
385            rule_default.is_media_file("document.pdf"),
386            "document.pdf should be identified as a media file"
387        );
388        assert!(
389            rule_default.is_media_file("path/to/audio.mp3"),
390            "path/to/audio.mp3 should be identified as a media file"
391        );
392
393        assert!(
394            !rule_default.is_media_file("document.md"),
395            "document.md should not be identified as a media file"
396        );
397        assert!(
398            !rule_default.is_media_file("code.rs"),
399            "code.rs should not be identified as a media file"
400        );
401
402        // Test media file skipping with default settings (skip_media_files = true)
403        assert!(
404            rule_default.should_skip_media_file("image.jpg"),
405            "image.jpg should be skipped with default settings"
406        );
407        assert!(
408            !rule_default.should_skip_media_file("document.md"),
409            "document.md should not be skipped"
410        );
411
412        // Test media file skipping with skip_media_files = false
413        let rule_no_skip = MD057ExistingRelativeLinks::new().with_skip_media_files(false);
414        assert!(
415            !rule_no_skip.should_skip_media_file("image.jpg"),
416            "image.jpg should not be skipped when skip_media_files is false"
417        );
418    }
419
420    #[test]
421    fn test_no_warnings_without_base_path() {
422        let rule = MD057ExistingRelativeLinks::new();
423        let content = "[Link](missing.md)";
424
425        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
426        let result = rule.check(&ctx).unwrap();
427        assert!(result.is_empty(), "Should have no warnings without base path");
428    }
429
430    #[test]
431    fn test_existing_and_missing_links() {
432        // Create a temporary directory for test files
433        let temp_dir = tempdir().unwrap();
434        let base_path = temp_dir.path();
435
436        // Create an existing file
437        let exists_path = base_path.join("exists.md");
438        File::create(&exists_path).unwrap().write_all(b"# Test File").unwrap();
439
440        // Verify the file exists
441        assert!(exists_path.exists(), "exists.md should exist for this test");
442
443        // Create test content with both existing and missing links
444        let content = r#"
445# Test Document
446
447[Valid Link](exists.md)
448[Invalid Link](missing.md)
449[External Link](https://example.com)
450[Media Link](image.jpg)
451        "#;
452
453        // Initialize rule with the base path
454        let rule = MD057ExistingRelativeLinks::new().with_path(base_path);
455
456        // Test the rule
457        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
458        let result = rule.check(&ctx).unwrap();
459
460        // Should have one warning for the missing.md link but not for the media file
461        assert_eq!(result.len(), 1);
462        assert!(result[0].message.contains("missing.md"));
463
464        // Test with check method
465        let result_with_structure = rule.check(&ctx).unwrap();
466
467        // Results should be the same
468        assert_eq!(result.len(), result_with_structure.len());
469        assert!(result_with_structure[0].message.contains("missing.md"));
470    }
471
472    #[test]
473    fn test_angle_bracket_links() {
474        // Create a temporary directory for test files
475        let temp_dir = tempdir().unwrap();
476        let base_path = temp_dir.path();
477
478        // Create an existing file
479        let exists_path = base_path.join("exists.md");
480        File::create(&exists_path).unwrap().write_all(b"# Test File").unwrap();
481
482        // Create test content with angle bracket links
483        let content = r#"
484# Test Document
485
486[Valid Link](<exists.md>)
487[Invalid Link](<missing.md>)
488[External Link](<https://example.com>)
489    "#;
490
491        // Test with default settings
492        let rule = MD057ExistingRelativeLinks::new().with_path(base_path);
493
494        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
495        let result = rule.check(&ctx).unwrap();
496
497        // Should have one warning for missing.md
498        assert_eq!(result.len(), 1, "Should have exactly one warning");
499        assert!(
500            result[0].message.contains("missing.md"),
501            "Warning should mention missing.md"
502        );
503    }
504
505    #[test]
506    fn test_media_file_handling() {
507        // Create a temporary directory for test files
508        let temp_dir = tempdir().unwrap();
509        let base_path = temp_dir.path();
510
511        // Explicitly check that image.jpg doesn't exist in the test directory
512        let image_path = base_path.join("image.jpg");
513        assert!(
514            !image_path.exists(),
515            "Test precondition failed: image.jpg should not exist"
516        );
517
518        // Create a test content with a media link - make sure it's very explicit
519        let content = "[Media Link](image.jpg)";
520
521        // Test with skip_media_files = true (default)
522        let rule_skip_media = MD057ExistingRelativeLinks::new().with_path(base_path);
523
524        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
525        let result_skip = rule_skip_media.check(&ctx).unwrap();
526
527        // Should have no warnings when media files are skipped
528        assert_eq!(
529            result_skip.len(),
530            0,
531            "Should have no warnings when skip_media_files is true"
532        );
533
534        // Test with skip_media_files = false
535        let rule_check_all = MD057ExistingRelativeLinks::new()
536            .with_path(base_path)
537            .with_skip_media_files(false);
538
539        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
540        let result_all = rule_check_all.check(&ctx).unwrap();
541
542        // Should warn about the missing media file
543        assert_eq!(
544            result_all.len(),
545            1,
546            "Should have one warning when skip_media_files is false"
547        );
548        assert!(
549            result_all[0].message.contains("image.jpg"),
550            "Warning should mention image.jpg"
551        );
552    }
553
554    #[test]
555    fn test_code_span_detection() {
556        let rule = MD057ExistingRelativeLinks::new();
557
558        // Create a temporary directory for test files
559        let temp_dir = tempdir().unwrap();
560        let base_path = temp_dir.path();
561
562        let rule = rule.with_path(base_path);
563
564        // Test with document structure
565        let content = "This is a [link](nonexistent.md) and `[not a link](not-checked.md)` in code.";
566
567        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
568        let result = rule.check(&ctx).unwrap();
569
570        // Should only find the real link, not the one in code
571        assert_eq!(result.len(), 1, "Should only flag the real link");
572        assert!(result[0].message.contains("nonexistent.md"));
573    }
574
575    #[test]
576    fn test_inline_code_spans() {
577        // Create a temporary directory for test files
578        let temp_dir = tempdir().unwrap();
579        let base_path = temp_dir.path();
580
581        // Create test content with links in inline code spans
582        let content = r#"
583# Test Document
584
585This is a normal link: [Link](missing.md)
586
587This is a code span with a link: `[Link](another-missing.md)`
588
589Some more text with `inline code [Link](yet-another-missing.md) embedded`.
590
591    "#;
592
593        // Initialize rule with the base path
594        let rule = MD057ExistingRelativeLinks::new().with_path(base_path);
595
596        // Test the rule
597        let ctx = crate::lint_context::LintContext::new(content, crate::config::MarkdownFlavor::Standard);
598        let result = rule.check(&ctx).unwrap();
599
600        // Should only have warning for the normal link, not for links in code spans
601        assert_eq!(result.len(), 1, "Should have exactly one warning");
602        assert!(
603            result[0].message.contains("missing.md"),
604            "Warning should be for missing.md"
605        );
606        assert!(
607            !result.iter().any(|w| w.message.contains("another-missing.md")),
608            "Should not warn about link in code span"
609        );
610        assert!(
611            !result.iter().any(|w| w.message.contains("yet-another-missing.md")),
612            "Should not warn about link in inline code"
613        );
614    }
615}
rumdl_lib/rules/md057_existing_relative_links.rs

rumdl_lib/rules/
md057_existing_relative_links.rs
