fallow_extract/
html.rs

1//! HTML file parsing for script, stylesheet, and Angular template references.
2//!
3//! Extracts `<script src="...">` and `<link rel="stylesheet" href="...">` references
4//! from HTML files, creating graph edges so that referenced JS/CSS assets (and their
5//! transitive imports) are reachable from the HTML entry point.
6//!
7//! Also scans for Angular template syntax (`{{ }}`, `[prop]`, `(event)`, `@if`, etc.)
8//! and stores referenced identifiers as `MemberAccess` entries with a sentinel object,
9//! enabling the analysis phase to credit component class members used in external templates.
10
11use std::path::Path;
12use std::sync::LazyLock;
13
14use oxc_span::Span;
15
16use crate::asset_url::normalize_asset_url;
17use crate::sfc_template::angular::{self, ANGULAR_TPL_SENTINEL};
18use crate::{ImportInfo, ImportedName, MemberAccess, ModuleInfo};
19use fallow_types::discover::FileId;
20
21/// Regex to match HTML comments (`<!-- ... -->`) for stripping before extraction.
22static HTML_COMMENT_RE: LazyLock<regex::Regex> =
23    LazyLock::new(|| regex::Regex::new(r"(?s)<!--.*?-->").expect("valid regex"));
24
25/// Regex to extract `src` attribute from `<script>` tags.
26/// Matches both `<script src="...">` and `<script type="module" src="...">`.
27/// Uses `(?s)` so `.` matches newlines (multi-line attributes).
28static SCRIPT_SRC_RE: LazyLock<regex::Regex> = LazyLock::new(|| {
29    regex::Regex::new(r#"(?si)<script\b(?:[^>"']|"[^"]*"|'[^']*')*?\bsrc\s*=\s*["']([^"']+)["']"#)
30        .expect("valid regex")
31});
32
33/// Regex to extract `href` attribute from `<link>` tags with `rel="stylesheet"` or
34/// `rel="modulepreload"`.
35/// Handles attributes in any order (rel before or after href).
36static LINK_HREF_RE: LazyLock<regex::Regex> = LazyLock::new(|| {
37    regex::Regex::new(
38        r#"(?si)<link\b(?:[^>"']|"[^"]*"|'[^']*')*?\brel\s*=\s*["'](stylesheet|modulepreload)["'](?:[^>"']|"[^"]*"|'[^']*')*?\bhref\s*=\s*["']([^"']+)["']"#,
39    )
40    .expect("valid regex")
41});
42
43/// Regex for the reverse attribute order: href before rel.
44static LINK_HREF_REVERSE_RE: LazyLock<regex::Regex> = LazyLock::new(|| {
45    regex::Regex::new(
46        r#"(?si)<link\b(?:[^>"']|"[^"]*"|'[^']*')*?\bhref\s*=\s*["']([^"']+)["'](?:[^>"']|"[^"]*"|'[^']*')*?\brel\s*=\s*["'](stylesheet|modulepreload)["']"#,
47    )
48    .expect("valid regex")
49});
50
51/// Check if a path is an HTML file.
52// Keep in sync with fallow_core::analyze::predicates::is_html_file (crate boundary prevents sharing)
53pub(crate) fn is_html_file(path: &Path) -> bool {
54    path.extension()
55        .and_then(|e| e.to_str())
56        .is_some_and(|ext| ext == "html")
57}
58
59/// Returns true if an HTML asset reference is a remote URL that should be skipped.
60pub(crate) fn is_remote_url(src: &str) -> bool {
61    src.starts_with("http://")
62        || src.starts_with("https://")
63        || src.starts_with("//")
64        || src.starts_with("data:")
65}
66
67/// Extract local (non-remote) asset references from HTML-like markup.
68///
69/// Returns the raw `src`/`href` strings (trimmed, remote URLs filtered). Shared
70/// between the HTML file parser and the JS/TS visitor's tagged template
71/// literal override so `` html`<script src="...">` `` in Hono/lit-html/htm
72/// layouts emits the same asset edges as a real `.html` file.
73pub(crate) fn collect_asset_refs(source: &str) -> Vec<String> {
74    let stripped = HTML_COMMENT_RE.replace_all(source, "");
75    let mut refs: Vec<String> = Vec::new();
76
77    for cap in SCRIPT_SRC_RE.captures_iter(&stripped) {
78        if let Some(m) = cap.get(1) {
79            let src = m.as_str().trim();
80            if !src.is_empty() && !is_remote_url(src) {
81                refs.push(src.to_string());
82            }
83        }
84    }
85
86    for cap in LINK_HREF_RE.captures_iter(&stripped) {
87        if let Some(m) = cap.get(2) {
88            let href = m.as_str().trim();
89            if !href.is_empty() && !is_remote_url(href) {
90                refs.push(href.to_string());
91            }
92        }
93    }
94    for cap in LINK_HREF_REVERSE_RE.captures_iter(&stripped) {
95        if let Some(m) = cap.get(1) {
96            let href = m.as_str().trim();
97            if !href.is_empty() && !is_remote_url(href) {
98                refs.push(href.to_string());
99            }
100        }
101    }
102
103    refs
104}
105
106/// Parse an HTML file, extracting script and stylesheet references as imports.
107#[cfg(test)]
108pub(crate) fn parse_html_to_module(file_id: FileId, source: &str, content_hash: u64) -> ModuleInfo {
109    parse_html_to_module_with_complexity(file_id, source, content_hash, false)
110}
111
112/// Parse an HTML file and optionally compute Angular template complexity.
113pub(crate) fn parse_html_to_module_with_complexity(
114    file_id: FileId,
115    source: &str,
116    content_hash: u64,
117    need_complexity: bool,
118) -> ModuleInfo {
119    let suppressions = crate::suppress::parse_suppressions_from_source(source);
120
121    // Bare filenames (e.g., `src="app.js"`) are normalized to `./app.js` so
122    // the resolver doesn't misclassify them as npm packages.
123    let mut imports: Vec<ImportInfo> = collect_asset_refs(source)
124        .into_iter()
125        .map(|raw| ImportInfo {
126            source: normalize_asset_url(&raw),
127            imported_name: ImportedName::SideEffect,
128            local_name: String::new(),
129            is_type_only: false,
130            from_style: false,
131            span: Span::default(),
132            source_span: Span::default(),
133        })
134        .collect();
135
136    // Deduplicate: the same asset may be referenced by both <script src> and
137    // <link rel="modulepreload" href> for the same path.
138    imports.sort_unstable_by(|a, b| a.source.cmp(&b.source));
139    imports.dedup_by(|a, b| a.source == b.source);
140
141    // Scan for Angular template syntax ({{ }}, [prop], (event), @if, etc.).
142    //
143    // Bare identifier refs (e.g. `title`, `dataService`, pipe names) are stored
144    // as `MemberAccess` entries with a sentinel object name so the analysis
145    // phase can credit them as members of the component class.
146    //
147    // Static member-access chains (`dataService.getTotal`) where `dataService`
148    // is an unresolved identifier are stored as regular (non-sentinel)
149    // `MemberAccess` entries. The analysis phase resolves these through the
150    // importing component's typed instance bindings (from
151    // `ClassHeritageInfo.instance_bindings`) to credit the target class's
152    // member as used.
153    let template_refs = angular::collect_angular_template_refs(source);
154    let mut member_accesses: Vec<MemberAccess> = template_refs
155        .identifiers
156        .into_iter()
157        .map(|name| MemberAccess {
158            object: ANGULAR_TPL_SENTINEL.to_string(),
159            member: name,
160        })
161        .collect();
162    member_accesses.extend(template_refs.member_accesses);
163
164    let complexity = if need_complexity {
165        crate::template_complexity::compute_angular_template_complexity(source)
166            .into_iter()
167            .collect()
168    } else {
169        Vec::new()
170    };
171
172    ModuleInfo {
173        file_id,
174        exports: Vec::new(),
175        imports,
176        re_exports: Vec::new(),
177        dynamic_imports: Vec::new(),
178        dynamic_import_patterns: Vec::new(),
179        require_calls: Vec::new(),
180        member_accesses,
181        whole_object_uses: Vec::new(),
182        has_cjs_exports: false,
183        content_hash,
184        suppressions,
185        unused_import_bindings: Vec::new(),
186        type_referenced_import_bindings: Vec::new(),
187        value_referenced_import_bindings: Vec::new(),
188        line_offsets: fallow_types::extract::compute_line_offsets(source),
189        complexity,
190        flag_uses: Vec::new(),
191        class_heritage: vec![],
192        local_type_declarations: Vec::new(),
193        public_signature_type_references: Vec::new(),
194    }
195}
196
197#[cfg(test)]
198mod tests {
199    use super::*;
200
201    // ── is_html_file ─────────────────────────────────────────────
202
203    #[test]
204    fn is_html_file_html() {
205        assert!(is_html_file(Path::new("index.html")));
206    }
207
208    #[test]
209    fn is_html_file_nested() {
210        assert!(is_html_file(Path::new("pages/about.html")));
211    }
212
213    #[test]
214    fn is_html_file_rejects_htm() {
215        assert!(!is_html_file(Path::new("index.htm")));
216    }
217
218    #[test]
219    fn is_html_file_rejects_js() {
220        assert!(!is_html_file(Path::new("app.js")));
221    }
222
223    #[test]
224    fn is_html_file_rejects_ts() {
225        assert!(!is_html_file(Path::new("app.ts")));
226    }
227
228    #[test]
229    fn is_html_file_rejects_vue() {
230        assert!(!is_html_file(Path::new("App.vue")));
231    }
232
233    // ── is_remote_url ────────────────────────────────────────────
234
235    #[test]
236    fn remote_url_http() {
237        assert!(is_remote_url("http://example.com/script.js"));
238    }
239
240    #[test]
241    fn remote_url_https() {
242        assert!(is_remote_url("https://cdn.example.com/style.css"));
243    }
244
245    #[test]
246    fn remote_url_protocol_relative() {
247        assert!(is_remote_url("//cdn.example.com/lib.js"));
248    }
249
250    #[test]
251    fn remote_url_data() {
252        assert!(is_remote_url("data:text/javascript;base64,abc"));
253    }
254
255    #[test]
256    fn local_relative_not_remote() {
257        assert!(!is_remote_url("./src/entry.js"));
258    }
259
260    #[test]
261    fn local_root_relative_not_remote() {
262        assert!(!is_remote_url("/src/entry.js"));
263    }
264
265    // ── parse_html_to_module: script src extraction ──────────────
266
267    #[test]
268    fn extracts_module_script_src() {
269        let info = parse_html_to_module(
270            FileId(0),
271            r#"<script type="module" src="./src/entry.js"></script>"#,
272            0,
273        );
274        assert_eq!(info.imports.len(), 1);
275        assert_eq!(info.imports[0].source, "./src/entry.js");
276    }
277
278    #[test]
279    fn extracts_plain_script_src() {
280        let info = parse_html_to_module(
281            FileId(0),
282            r#"<script src="./src/polyfills.js"></script>"#,
283            0,
284        );
285        assert_eq!(info.imports.len(), 1);
286        assert_eq!(info.imports[0].source, "./src/polyfills.js");
287    }
288
289    #[test]
290    fn extracts_multiple_scripts() {
291        let info = parse_html_to_module(
292            FileId(0),
293            r#"
294            <script type="module" src="./src/entry.js"></script>
295            <script src="./src/polyfills.js"></script>
296            "#,
297            0,
298        );
299        assert_eq!(info.imports.len(), 2);
300    }
301
302    #[test]
303    fn skips_inline_script() {
304        let info = parse_html_to_module(FileId(0), r#"<script>console.log("hello");</script>"#, 0);
305        assert!(info.imports.is_empty());
306    }
307
308    #[test]
309    fn skips_remote_script() {
310        let info = parse_html_to_module(
311            FileId(0),
312            r#"<script src="https://cdn.example.com/lib.js"></script>"#,
313            0,
314        );
315        assert!(info.imports.is_empty());
316    }
317
318    #[test]
319    fn skips_protocol_relative_script() {
320        let info = parse_html_to_module(
321            FileId(0),
322            r#"<script src="//cdn.example.com/lib.js"></script>"#,
323            0,
324        );
325        assert!(info.imports.is_empty());
326    }
327
328    // ── parse_html_to_module: link href extraction ───────────────
329
330    #[test]
331    fn extracts_stylesheet_link() {
332        let info = parse_html_to_module(
333            FileId(0),
334            r#"<link rel="stylesheet" href="./src/global.css" />"#,
335            0,
336        );
337        assert_eq!(info.imports.len(), 1);
338        assert_eq!(info.imports[0].source, "./src/global.css");
339    }
340
341    #[test]
342    fn extracts_modulepreload_link() {
343        let info = parse_html_to_module(
344            FileId(0),
345            r#"<link rel="modulepreload" href="./src/vendor.js" />"#,
346            0,
347        );
348        assert_eq!(info.imports.len(), 1);
349        assert_eq!(info.imports[0].source, "./src/vendor.js");
350    }
351
352    #[test]
353    fn extracts_link_with_reversed_attrs() {
354        let info = parse_html_to_module(
355            FileId(0),
356            r#"<link href="./src/global.css" rel="stylesheet" />"#,
357            0,
358        );
359        assert_eq!(info.imports.len(), 1);
360        assert_eq!(info.imports[0].source, "./src/global.css");
361    }
362
363    // ── Bare asset references normalized to relative paths ──────
364    // Regression tests for the same class of bug as #99 (Angular templateUrl).
365    // Browsers resolve `src="app.js"` and `href="styles.css"` relative to the
366    // HTML file, so emitting these as bare specifiers would misclassify them
367    // as unlisted npm packages.
368
369    #[test]
370    fn bare_script_src_normalized_to_relative() {
371        let info = parse_html_to_module(FileId(0), r#"<script src="app.js"></script>"#, 0);
372        assert_eq!(info.imports.len(), 1);
373        assert_eq!(info.imports[0].source, "./app.js");
374    }
375
376    #[test]
377    fn bare_module_script_src_normalized_to_relative() {
378        let info = parse_html_to_module(
379            FileId(0),
380            r#"<script type="module" src="main.ts"></script>"#,
381            0,
382        );
383        assert_eq!(info.imports.len(), 1);
384        assert_eq!(info.imports[0].source, "./main.ts");
385    }
386
387    #[test]
388    fn bare_stylesheet_link_href_normalized_to_relative() {
389        let info = parse_html_to_module(
390            FileId(0),
391            r#"<link rel="stylesheet" href="styles.css" />"#,
392            0,
393        );
394        assert_eq!(info.imports.len(), 1);
395        assert_eq!(info.imports[0].source, "./styles.css");
396    }
397
398    #[test]
399    fn bare_link_href_reversed_attrs_normalized_to_relative() {
400        let info = parse_html_to_module(
401            FileId(0),
402            r#"<link href="styles.css" rel="stylesheet" />"#,
403            0,
404        );
405        assert_eq!(info.imports.len(), 1);
406        assert_eq!(info.imports[0].source, "./styles.css");
407    }
408
409    #[test]
410    fn bare_modulepreload_link_href_normalized_to_relative() {
411        let info = parse_html_to_module(
412            FileId(0),
413            r#"<link rel="modulepreload" href="vendor.js" />"#,
414            0,
415        );
416        assert_eq!(info.imports.len(), 1);
417        assert_eq!(info.imports[0].source, "./vendor.js");
418    }
419
420    #[test]
421    fn bare_asset_with_subdir_normalized_to_relative() {
422        let info = parse_html_to_module(FileId(0), r#"<script src="assets/app.js"></script>"#, 0);
423        assert_eq!(info.imports.len(), 1);
424        assert_eq!(info.imports[0].source, "./assets/app.js");
425    }
426
427    #[test]
428    fn root_absolute_script_src_unchanged() {
429        // `/src/main.ts` is a web convention (Vite root-relative) and must
430        // stay absolute so the resolver's HTML special case applies.
431        let info = parse_html_to_module(FileId(0), r#"<script src="/src/main.ts"></script>"#, 0);
432        assert_eq!(info.imports.len(), 1);
433        assert_eq!(info.imports[0].source, "/src/main.ts");
434    }
435
436    #[test]
437    fn parent_relative_script_src_unchanged() {
438        let info = parse_html_to_module(
439            FileId(0),
440            r#"<script src="../shared/vendor.js"></script>"#,
441            0,
442        );
443        assert_eq!(info.imports.len(), 1);
444        assert_eq!(info.imports[0].source, "../shared/vendor.js");
445    }
446
447    #[test]
448    fn skips_preload_link() {
449        let info = parse_html_to_module(
450            FileId(0),
451            r#"<link rel="preload" href="./src/font.woff2" as="font" />"#,
452            0,
453        );
454        assert!(info.imports.is_empty());
455    }
456
457    #[test]
458    fn skips_icon_link() {
459        let info =
460            parse_html_to_module(FileId(0), r#"<link rel="icon" href="./favicon.ico" />"#, 0);
461        assert!(info.imports.is_empty());
462    }
463
464    #[test]
465    fn skips_remote_stylesheet() {
466        let info = parse_html_to_module(
467            FileId(0),
468            r#"<link rel="stylesheet" href="https://fonts.googleapis.com/css" />"#,
469            0,
470        );
471        assert!(info.imports.is_empty());
472    }
473
474    // ── HTML comment stripping ───────────────────────────────────
475
476    #[test]
477    fn skips_commented_out_script() {
478        let info = parse_html_to_module(
479            FileId(0),
480            r#"<!-- <script src="./old.js"></script> -->
481            <script src="./new.js"></script>"#,
482            0,
483        );
484        assert_eq!(info.imports.len(), 1);
485        assert_eq!(info.imports[0].source, "./new.js");
486    }
487
488    #[test]
489    fn skips_commented_out_link() {
490        let info = parse_html_to_module(
491            FileId(0),
492            r#"<!-- <link rel="stylesheet" href="./old.css" /> -->
493            <link rel="stylesheet" href="./new.css" />"#,
494            0,
495        );
496        assert_eq!(info.imports.len(), 1);
497        assert_eq!(info.imports[0].source, "./new.css");
498    }
499
500    // ── Multi-line attributes ────────────────────────────────────
501
502    #[test]
503    fn handles_multiline_script_tag() {
504        let info = parse_html_to_module(
505            FileId(0),
506            "<script\n  type=\"module\"\n  src=\"./src/entry.js\"\n></script>",
507            0,
508        );
509        assert_eq!(info.imports.len(), 1);
510        assert_eq!(info.imports[0].source, "./src/entry.js");
511    }
512
513    #[test]
514    fn handles_multiline_link_tag() {
515        let info = parse_html_to_module(
516            FileId(0),
517            "<link\n  rel=\"stylesheet\"\n  href=\"./src/global.css\"\n/>",
518            0,
519        );
520        assert_eq!(info.imports.len(), 1);
521        assert_eq!(info.imports[0].source, "./src/global.css");
522    }
523
524    // ── Full HTML document ───────────────────────────────────────
525
526    #[test]
527    fn full_vite_html() {
528        let info = parse_html_to_module(
529            FileId(0),
530            r#"<!doctype html>
531<html>
532  <head>
533    <link rel="stylesheet" href="./src/global.css" />
534    <link rel="icon" href="/favicon.ico" />
535  </head>
536  <body>
537    <div id="app"></div>
538    <script type="module" src="./src/entry.js"></script>
539  </body>
540</html>"#,
541            0,
542        );
543        assert_eq!(info.imports.len(), 2);
544        let sources: Vec<&str> = info.imports.iter().map(|i| i.source.as_str()).collect();
545        assert!(sources.contains(&"./src/global.css"));
546        assert!(sources.contains(&"./src/entry.js"));
547    }
548
549    // ── Edge cases ───────────────────────────────────────────────
550
551    #[test]
552    fn empty_html() {
553        let info = parse_html_to_module(FileId(0), "", 0);
554        assert!(info.imports.is_empty());
555    }
556
557    #[test]
558    fn html_with_no_assets() {
559        let info = parse_html_to_module(
560            FileId(0),
561            r"<!doctype html><html><body><h1>Hello</h1></body></html>",
562            0,
563        );
564        assert!(info.imports.is_empty());
565    }
566
567    #[test]
568    fn single_quoted_attributes() {
569        let info = parse_html_to_module(FileId(0), r"<script src='./src/entry.js'></script>", 0);
570        assert_eq!(info.imports.len(), 1);
571        assert_eq!(info.imports[0].source, "./src/entry.js");
572    }
573
574    #[test]
575    fn all_imports_are_side_effect() {
576        let info = parse_html_to_module(
577            FileId(0),
578            r#"<script src="./entry.js"></script>
579            <link rel="stylesheet" href="./style.css" />"#,
580            0,
581        );
582        for imp in &info.imports {
583            assert!(matches!(imp.imported_name, ImportedName::SideEffect));
584            assert!(imp.local_name.is_empty());
585            assert!(!imp.is_type_only);
586        }
587    }
588
589    #[test]
590    fn suppression_comments_extracted() {
591        let info = parse_html_to_module(
592            FileId(0),
593            "<!-- fallow-ignore-file -->\n<script src=\"./entry.js\"></script>",
594            0,
595        );
596        // HTML comments use <!-- --> not //, so suppression parsing
597        // from source text won't find standard JS-style comments.
598        // This is expected — HTML suppression is not supported.
599        assert_eq!(info.imports.len(), 1);
600    }
601
602    // ── Angular template scanning ──────────────────────────────
603
604    #[test]
605    fn angular_template_extracts_member_refs() {
606        let info = parse_html_to_module(
607            FileId(0),
608            "<h1>{{ title() }}</h1>\n\
609             <p [class.highlighted]=\"isHighlighted\">{{ greeting() }}</p>\n\
610             <button (click)=\"onButtonClick()\">Toggle</button>",
611            0,
612        );
613        let names: rustc_hash::FxHashSet<&str> = info
614            .member_accesses
615            .iter()
616            .filter(|a| a.object == ANGULAR_TPL_SENTINEL)
617            .map(|a| a.member.as_str())
618            .collect();
619        assert!(names.contains("title"), "should contain 'title'");
620        assert!(
621            names.contains("isHighlighted"),
622            "should contain 'isHighlighted'"
623        );
624        assert!(names.contains("greeting"), "should contain 'greeting'");
625        assert!(
626            names.contains("onButtonClick"),
627            "should contain 'onButtonClick'"
628        );
629    }
630
631    #[test]
632    fn plain_html_no_angular_refs() {
633        let info = parse_html_to_module(
634            FileId(0),
635            "<!doctype html><html><body><h1>Hello</h1></body></html>",
636            0,
637        );
638        assert!(info.member_accesses.is_empty());
639    }
640}
fallow_extract/html.rs

fallow_extract/
html.rs
