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        has_angular_component_template_url: false,
184        content_hash,
185        suppressions,
186        unused_import_bindings: Vec::new(),
187        type_referenced_import_bindings: Vec::new(),
188        value_referenced_import_bindings: Vec::new(),
189        line_offsets: fallow_types::extract::compute_line_offsets(source),
190        complexity,
191        flag_uses: Vec::new(),
192        class_heritage: vec![],
193        local_type_declarations: Vec::new(),
194        public_signature_type_references: Vec::new(),
195        namespace_object_aliases: Vec::new(),
196    }
197}
198
199#[cfg(test)]
200mod tests {
201    use super::*;
202
203    // ── is_html_file ─────────────────────────────────────────────
204
205    #[test]
206    fn is_html_file_html() {
207        assert!(is_html_file(Path::new("index.html")));
208    }
209
210    #[test]
211    fn is_html_file_nested() {
212        assert!(is_html_file(Path::new("pages/about.html")));
213    }
214
215    #[test]
216    fn is_html_file_rejects_htm() {
217        assert!(!is_html_file(Path::new("index.htm")));
218    }
219
220    #[test]
221    fn is_html_file_rejects_js() {
222        assert!(!is_html_file(Path::new("app.js")));
223    }
224
225    #[test]
226    fn is_html_file_rejects_ts() {
227        assert!(!is_html_file(Path::new("app.ts")));
228    }
229
230    #[test]
231    fn is_html_file_rejects_vue() {
232        assert!(!is_html_file(Path::new("App.vue")));
233    }
234
235    // ── is_remote_url ────────────────────────────────────────────
236
237    #[test]
238    fn remote_url_http() {
239        assert!(is_remote_url("http://example.com/script.js"));
240    }
241
242    #[test]
243    fn remote_url_https() {
244        assert!(is_remote_url("https://cdn.example.com/style.css"));
245    }
246
247    #[test]
248    fn remote_url_protocol_relative() {
249        assert!(is_remote_url("//cdn.example.com/lib.js"));
250    }
251
252    #[test]
253    fn remote_url_data() {
254        assert!(is_remote_url("data:text/javascript;base64,abc"));
255    }
256
257    #[test]
258    fn local_relative_not_remote() {
259        assert!(!is_remote_url("./src/entry.js"));
260    }
261
262    #[test]
263    fn local_root_relative_not_remote() {
264        assert!(!is_remote_url("/src/entry.js"));
265    }
266
267    // ── parse_html_to_module: script src extraction ──────────────
268
269    #[test]
270    fn extracts_module_script_src() {
271        let info = parse_html_to_module(
272            FileId(0),
273            r#"<script type="module" src="./src/entry.js"></script>"#,
274            0,
275        );
276        assert_eq!(info.imports.len(), 1);
277        assert_eq!(info.imports[0].source, "./src/entry.js");
278    }
279
280    #[test]
281    fn extracts_plain_script_src() {
282        let info = parse_html_to_module(
283            FileId(0),
284            r#"<script src="./src/polyfills.js"></script>"#,
285            0,
286        );
287        assert_eq!(info.imports.len(), 1);
288        assert_eq!(info.imports[0].source, "./src/polyfills.js");
289    }
290
291    #[test]
292    fn extracts_multiple_scripts() {
293        let info = parse_html_to_module(
294            FileId(0),
295            r#"
296            <script type="module" src="./src/entry.js"></script>
297            <script src="./src/polyfills.js"></script>
298            "#,
299            0,
300        );
301        assert_eq!(info.imports.len(), 2);
302    }
303
304    #[test]
305    fn skips_inline_script() {
306        let info = parse_html_to_module(FileId(0), r#"<script>console.log("hello");</script>"#, 0);
307        assert!(info.imports.is_empty());
308    }
309
310    #[test]
311    fn skips_remote_script() {
312        let info = parse_html_to_module(
313            FileId(0),
314            r#"<script src="https://cdn.example.com/lib.js"></script>"#,
315            0,
316        );
317        assert!(info.imports.is_empty());
318    }
319
320    #[test]
321    fn skips_protocol_relative_script() {
322        let info = parse_html_to_module(
323            FileId(0),
324            r#"<script src="//cdn.example.com/lib.js"></script>"#,
325            0,
326        );
327        assert!(info.imports.is_empty());
328    }
329
330    // ── parse_html_to_module: link href extraction ───────────────
331
332    #[test]
333    fn extracts_stylesheet_link() {
334        let info = parse_html_to_module(
335            FileId(0),
336            r#"<link rel="stylesheet" href="./src/global.css" />"#,
337            0,
338        );
339        assert_eq!(info.imports.len(), 1);
340        assert_eq!(info.imports[0].source, "./src/global.css");
341    }
342
343    #[test]
344    fn extracts_modulepreload_link() {
345        let info = parse_html_to_module(
346            FileId(0),
347            r#"<link rel="modulepreload" href="./src/vendor.js" />"#,
348            0,
349        );
350        assert_eq!(info.imports.len(), 1);
351        assert_eq!(info.imports[0].source, "./src/vendor.js");
352    }
353
354    #[test]
355    fn extracts_link_with_reversed_attrs() {
356        let info = parse_html_to_module(
357            FileId(0),
358            r#"<link href="./src/global.css" rel="stylesheet" />"#,
359            0,
360        );
361        assert_eq!(info.imports.len(), 1);
362        assert_eq!(info.imports[0].source, "./src/global.css");
363    }
364
365    // ── Bare asset references normalized to relative paths ──────
366    // Regression tests for the same class of bug as #99 (Angular templateUrl).
367    // Browsers resolve `src="app.js"` and `href="styles.css"` relative to the
368    // HTML file, so emitting these as bare specifiers would misclassify them
369    // as unlisted npm packages.
370
371    #[test]
372    fn bare_script_src_normalized_to_relative() {
373        let info = parse_html_to_module(FileId(0), r#"<script src="app.js"></script>"#, 0);
374        assert_eq!(info.imports.len(), 1);
375        assert_eq!(info.imports[0].source, "./app.js");
376    }
377
378    #[test]
379    fn bare_module_script_src_normalized_to_relative() {
380        let info = parse_html_to_module(
381            FileId(0),
382            r#"<script type="module" src="main.ts"></script>"#,
383            0,
384        );
385        assert_eq!(info.imports.len(), 1);
386        assert_eq!(info.imports[0].source, "./main.ts");
387    }
388
389    #[test]
390    fn bare_stylesheet_link_href_normalized_to_relative() {
391        let info = parse_html_to_module(
392            FileId(0),
393            r#"<link rel="stylesheet" href="styles.css" />"#,
394            0,
395        );
396        assert_eq!(info.imports.len(), 1);
397        assert_eq!(info.imports[0].source, "./styles.css");
398    }
399
400    #[test]
401    fn bare_link_href_reversed_attrs_normalized_to_relative() {
402        let info = parse_html_to_module(
403            FileId(0),
404            r#"<link href="styles.css" rel="stylesheet" />"#,
405            0,
406        );
407        assert_eq!(info.imports.len(), 1);
408        assert_eq!(info.imports[0].source, "./styles.css");
409    }
410
411    #[test]
412    fn bare_modulepreload_link_href_normalized_to_relative() {
413        let info = parse_html_to_module(
414            FileId(0),
415            r#"<link rel="modulepreload" href="vendor.js" />"#,
416            0,
417        );
418        assert_eq!(info.imports.len(), 1);
419        assert_eq!(info.imports[0].source, "./vendor.js");
420    }
421
422    #[test]
423    fn bare_asset_with_subdir_normalized_to_relative() {
424        let info = parse_html_to_module(FileId(0), r#"<script src="assets/app.js"></script>"#, 0);
425        assert_eq!(info.imports.len(), 1);
426        assert_eq!(info.imports[0].source, "./assets/app.js");
427    }
428
429    #[test]
430    fn root_absolute_script_src_unchanged() {
431        // `/src/main.ts` is a web convention (Vite root-relative) and must
432        // stay absolute so the resolver's HTML special case applies.
433        let info = parse_html_to_module(FileId(0), r#"<script src="/src/main.ts"></script>"#, 0);
434        assert_eq!(info.imports.len(), 1);
435        assert_eq!(info.imports[0].source, "/src/main.ts");
436    }
437
438    #[test]
439    fn parent_relative_script_src_unchanged() {
440        let info = parse_html_to_module(
441            FileId(0),
442            r#"<script src="../shared/vendor.js"></script>"#,
443            0,
444        );
445        assert_eq!(info.imports.len(), 1);
446        assert_eq!(info.imports[0].source, "../shared/vendor.js");
447    }
448
449    #[test]
450    fn skips_preload_link() {
451        let info = parse_html_to_module(
452            FileId(0),
453            r#"<link rel="preload" href="./src/font.woff2" as="font" />"#,
454            0,
455        );
456        assert!(info.imports.is_empty());
457    }
458
459    #[test]
460    fn skips_icon_link() {
461        let info =
462            parse_html_to_module(FileId(0), r#"<link rel="icon" href="./favicon.ico" />"#, 0);
463        assert!(info.imports.is_empty());
464    }
465
466    #[test]
467    fn skips_remote_stylesheet() {
468        let info = parse_html_to_module(
469            FileId(0),
470            r#"<link rel="stylesheet" href="https://fonts.googleapis.com/css" />"#,
471            0,
472        );
473        assert!(info.imports.is_empty());
474    }
475
476    // ── HTML comment stripping ───────────────────────────────────
477
478    #[test]
479    fn skips_commented_out_script() {
480        let info = parse_html_to_module(
481            FileId(0),
482            r#"<!-- <script src="./old.js"></script> -->
483            <script src="./new.js"></script>"#,
484            0,
485        );
486        assert_eq!(info.imports.len(), 1);
487        assert_eq!(info.imports[0].source, "./new.js");
488    }
489
490    #[test]
491    fn skips_commented_out_link() {
492        let info = parse_html_to_module(
493            FileId(0),
494            r#"<!-- <link rel="stylesheet" href="./old.css" /> -->
495            <link rel="stylesheet" href="./new.css" />"#,
496            0,
497        );
498        assert_eq!(info.imports.len(), 1);
499        assert_eq!(info.imports[0].source, "./new.css");
500    }
501
502    // ── Multi-line attributes ────────────────────────────────────
503
504    #[test]
505    fn handles_multiline_script_tag() {
506        let info = parse_html_to_module(
507            FileId(0),
508            "<script\n  type=\"module\"\n  src=\"./src/entry.js\"\n></script>",
509            0,
510        );
511        assert_eq!(info.imports.len(), 1);
512        assert_eq!(info.imports[0].source, "./src/entry.js");
513    }
514
515    #[test]
516    fn handles_multiline_link_tag() {
517        let info = parse_html_to_module(
518            FileId(0),
519            "<link\n  rel=\"stylesheet\"\n  href=\"./src/global.css\"\n/>",
520            0,
521        );
522        assert_eq!(info.imports.len(), 1);
523        assert_eq!(info.imports[0].source, "./src/global.css");
524    }
525
526    // ── Full HTML document ───────────────────────────────────────
527
528    #[test]
529    fn full_vite_html() {
530        let info = parse_html_to_module(
531            FileId(0),
532            r#"<!doctype html>
533<html>
534  <head>
535    <link rel="stylesheet" href="./src/global.css" />
536    <link rel="icon" href="/favicon.ico" />
537  </head>
538  <body>
539    <div id="app"></div>
540    <script type="module" src="./src/entry.js"></script>
541  </body>
542</html>"#,
543            0,
544        );
545        assert_eq!(info.imports.len(), 2);
546        let sources: Vec<&str> = info.imports.iter().map(|i| i.source.as_str()).collect();
547        assert!(sources.contains(&"./src/global.css"));
548        assert!(sources.contains(&"./src/entry.js"));
549    }
550
551    // ── Edge cases ───────────────────────────────────────────────
552
553    #[test]
554    fn empty_html() {
555        let info = parse_html_to_module(FileId(0), "", 0);
556        assert!(info.imports.is_empty());
557    }
558
559    #[test]
560    fn html_with_no_assets() {
561        let info = parse_html_to_module(
562            FileId(0),
563            r"<!doctype html><html><body><h1>Hello</h1></body></html>",
564            0,
565        );
566        assert!(info.imports.is_empty());
567    }
568
569    #[test]
570    fn single_quoted_attributes() {
571        let info = parse_html_to_module(FileId(0), r"<script src='./src/entry.js'></script>", 0);
572        assert_eq!(info.imports.len(), 1);
573        assert_eq!(info.imports[0].source, "./src/entry.js");
574    }
575
576    #[test]
577    fn all_imports_are_side_effect() {
578        let info = parse_html_to_module(
579            FileId(0),
580            r#"<script src="./entry.js"></script>
581            <link rel="stylesheet" href="./style.css" />"#,
582            0,
583        );
584        for imp in &info.imports {
585            assert!(matches!(imp.imported_name, ImportedName::SideEffect));
586            assert!(imp.local_name.is_empty());
587            assert!(!imp.is_type_only);
588        }
589    }
590
591    #[test]
592    fn suppression_comments_extracted() {
593        let info = parse_html_to_module(
594            FileId(0),
595            "<!-- fallow-ignore-file -->\n<script src=\"./entry.js\"></script>",
596            0,
597        );
598        // HTML comments use <!-- --> not //, so suppression parsing
599        // from source text won't find standard JS-style comments.
600        // This is expected — HTML suppression is not supported.
601        assert_eq!(info.imports.len(), 1);
602    }
603
604    // ── Angular template scanning ──────────────────────────────
605
606    #[test]
607    fn angular_template_extracts_member_refs() {
608        let info = parse_html_to_module(
609            FileId(0),
610            "<h1>{{ title() }}</h1>\n\
611             <p [class.highlighted]=\"isHighlighted\">{{ greeting() }}</p>\n\
612             <button (click)=\"onButtonClick()\">Toggle</button>",
613            0,
614        );
615        let names: rustc_hash::FxHashSet<&str> = info
616            .member_accesses
617            .iter()
618            .filter(|a| a.object == ANGULAR_TPL_SENTINEL)
619            .map(|a| a.member.as_str())
620            .collect();
621        assert!(names.contains("title"), "should contain 'title'");
622        assert!(
623            names.contains("isHighlighted"),
624            "should contain 'isHighlighted'"
625        );
626        assert!(names.contains("greeting"), "should contain 'greeting'");
627        assert!(
628            names.contains("onButtonClick"),
629            "should contain 'onButtonClick'"
630        );
631    }
632
633    #[test]
634    fn plain_html_no_angular_refs() {
635        let info = parse_html_to_module(
636            FileId(0),
637            "<!doctype html><html><body><h1>Hello</h1></body></html>",
638            0,
639        );
640        assert!(info.member_accesses.is_empty());
641    }
642}
fallow_extract/html.rs

fallow_extract/
html.rs
