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(|| crate::static_regex(r"(?s)<!--.*?-->"));
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    crate::static_regex(r#"(?si)<script\b(?:[^>"']|"[^"]*"|'[^']*')*?\bsrc\s*=\s*["']([^"']+)["']"#)
30});
31
32/// Regex to extract `href` attribute from `<link>` tags with `rel="stylesheet"` or
33/// `rel="modulepreload"`.
34/// Handles attributes in any order (rel before or after href).
35static LINK_HREF_RE: LazyLock<regex::Regex> = LazyLock::new(|| {
36    crate::static_regex(
37        r#"(?si)<link\b(?:[^>"']|"[^"]*"|'[^']*')*?\brel\s*=\s*["'](stylesheet|modulepreload)["'](?:[^>"']|"[^"]*"|'[^']*')*?\bhref\s*=\s*["']([^"']+)["']"#,
38    )
39});
40
41/// Regex for the reverse attribute order: href before rel.
42static LINK_HREF_REVERSE_RE: LazyLock<regex::Regex> = LazyLock::new(|| {
43    crate::static_regex(
44        r#"(?si)<link\b(?:[^>"']|"[^"]*"|'[^']*')*?\bhref\s*=\s*["']([^"']+)["'](?:[^>"']|"[^"]*"|'[^']*')*?\brel\s*=\s*["'](stylesheet|modulepreload)["']"#,
45    )
46});
47
48/// Check if a path is an HTML file.
49pub(crate) fn is_html_file(path: &Path) -> bool {
50    path.extension()
51        .and_then(|e| e.to_str())
52        .is_some_and(|ext| ext == "html")
53}
54
55/// Returns true if an HTML asset reference is a remote URL that should be skipped.
56pub(crate) fn is_remote_url(src: &str) -> bool {
57    src.starts_with("http://")
58        || src.starts_with("https://")
59        || src.starts_with("//")
60        || src.starts_with("data:")
61}
62
63/// Build-time template placeholders that aren't valid import specifiers and
64/// never resolve to a real file. Skip them at extraction time so they don't
65/// enter the import graph as unresolvable specifiers.
66///
67/// - `{{ ... }}` covers Handlebars (Ember `index.html`'s `{{rootURL}}`,
68///   `{{config.assetsPath}}`), Mustache (Jekyll, Hugo), Jinja2 (Pelican /
69///   11ty plugins), and pre-compiled Vue / Angular templates whose
70///   interpolation has leaked into a checked-in HTML scaffold.
71/// - `###...###` covers ember-cli blueprint scaffold placeholders
72///   (`###APPNAME###`, `###DUMMY###`) checked in as addon-fixture templates.
73///
74/// Neither shape is a legal URL or path character outside template engines,
75/// so the skip is generic across frameworks rather than gated on a plugin.
76/// Returns `true` for any `src` / `href` value that contains either marker.
77pub(crate) fn is_template_placeholder(value: &str) -> bool {
78    value.contains("{{") || value.contains("###")
79}
80
81/// Extract local (non-remote) asset references from HTML-like markup.
82///
83/// Returns the raw `src`/`href` strings (trimmed, remote URLs filtered). Shared
84/// between the HTML file parser and the JS/TS visitor's tagged template
85/// literal override so `` html`<script src="...">` `` in Hono/lit-html/htm
86/// layouts emits the same asset edges as a real `.html` file.
87pub(crate) fn collect_asset_refs(source: &str) -> Vec<String> {
88    let stripped = HTML_COMMENT_RE.replace_all(source, "");
89    let mut refs: Vec<String> = Vec::new();
90
91    for cap in SCRIPT_SRC_RE.captures_iter(&stripped) {
92        if let Some(m) = cap.get(1) {
93            let src = m.as_str().trim();
94            if !src.is_empty() && !is_remote_url(src) && !is_template_placeholder(src) {
95                refs.push(src.to_string());
96            }
97        }
98    }
99
100    for cap in LINK_HREF_RE.captures_iter(&stripped) {
101        if let Some(m) = cap.get(2) {
102            let href = m.as_str().trim();
103            if !href.is_empty() && !is_remote_url(href) && !is_template_placeholder(href) {
104                refs.push(href.to_string());
105            }
106        }
107    }
108    for cap in LINK_HREF_REVERSE_RE.captures_iter(&stripped) {
109        if let Some(m) = cap.get(1) {
110            let href = m.as_str().trim();
111            if !href.is_empty() && !is_remote_url(href) && !is_template_placeholder(href) {
112                refs.push(href.to_string());
113            }
114        }
115    }
116
117    refs
118}
119
120/// Parse an HTML file, extracting script and stylesheet references as imports.
121#[cfg(test)]
122pub(crate) fn parse_html_to_module(file_id: FileId, source: &str, content_hash: u64) -> ModuleInfo {
123    parse_html_to_module_with_complexity(file_id, source, content_hash, false)
124}
125
126/// Parse an HTML file and optionally compute Angular template complexity.
127pub(crate) fn parse_html_to_module_with_complexity(
128    file_id: FileId,
129    source: &str,
130    content_hash: u64,
131    need_complexity: bool,
132) -> ModuleInfo {
133    let parsed_suppressions = crate::suppress::parse_suppressions_from_source(source);
134
135    let mut imports: Vec<ImportInfo> = collect_asset_refs(source)
136        .into_iter()
137        .map(|raw| ImportInfo {
138            source: normalize_asset_url(&raw),
139            imported_name: ImportedName::SideEffect,
140            local_name: String::new(),
141            is_type_only: false,
142            from_style: false,
143            span: Span::default(),
144            source_span: Span::default(),
145        })
146        .collect();
147
148    imports.sort_unstable_by(|a, b| a.source.cmp(&b.source));
149    imports.dedup_by(|a, b| a.source == b.source);
150
151    let angular::AngularTemplateRefs {
152        identifiers,
153        member_accesses: template_member_accesses,
154        security_sinks,
155    } = angular::collect_angular_template_refs(source);
156    let mut member_accesses: Vec<MemberAccess> = identifiers
157        .into_iter()
158        .map(|name| MemberAccess {
159            object: ANGULAR_TPL_SENTINEL.to_string(),
160            member: name,
161        })
162        .collect();
163    member_accesses.extend(template_member_accesses);
164
165    let complexity = if need_complexity {
166        crate::template_complexity::compute_angular_template_complexity(source)
167            .into_iter()
168            .collect()
169    } else {
170        Vec::new()
171    };
172
173    ModuleInfo {
174        file_id,
175        exports: Vec::new(),
176        imports,
177        re_exports: Vec::new(),
178        dynamic_imports: Vec::new(),
179        dynamic_import_patterns: Vec::new(),
180        require_calls: Vec::new(),
181        package_path_references: Vec::new(),
182        member_accesses,
183        whole_object_uses: Vec::new(),
184        has_cjs_exports: false,
185        has_angular_component_template_url: false,
186        content_hash,
187        suppressions: parsed_suppressions.suppressions,
188        unknown_suppression_kinds: parsed_suppressions.unknown_kinds,
189        unused_import_bindings: Vec::new(),
190        type_referenced_import_bindings: Vec::new(),
191        value_referenced_import_bindings: Vec::new(),
192        line_offsets: fallow_types::extract::compute_line_offsets(source),
193        complexity,
194        flag_uses: Vec::new(),
195        class_heritage: vec![],
196        injection_tokens: vec![],
197        local_type_declarations: Vec::new(),
198        public_signature_type_references: Vec::new(),
199        namespace_object_aliases: Vec::new(),
200        iconify_prefixes: Vec::new(),
201        iconify_icon_names: Vec::new(),
202        auto_import_candidates: Vec::new(),
203        directives: Vec::new(),
204        security_sinks,
205        security_sinks_skipped: 0,
206        tainted_bindings: Vec::new(),
207        sanitized_sink_args: Vec::new(),
208    }
209}
210
211#[cfg(test)]
212mod tests {
213    use super::*;
214
215    #[test]
216    fn is_html_file_html() {
217        assert!(is_html_file(Path::new("index.html")));
218    }
219
220    #[test]
221    fn is_html_file_nested() {
222        assert!(is_html_file(Path::new("pages/about.html")));
223    }
224
225    #[test]
226    fn is_html_file_rejects_htm() {
227        assert!(!is_html_file(Path::new("index.htm")));
228    }
229
230    #[test]
231    fn is_html_file_rejects_js() {
232        assert!(!is_html_file(Path::new("app.js")));
233    }
234
235    #[test]
236    fn is_html_file_rejects_ts() {
237        assert!(!is_html_file(Path::new("app.ts")));
238    }
239
240    #[test]
241    fn is_html_file_rejects_vue() {
242        assert!(!is_html_file(Path::new("App.vue")));
243    }
244
245    #[test]
246    fn remote_url_http() {
247        assert!(is_remote_url("http://example.com/script.js"));
248    }
249
250    #[test]
251    fn remote_url_https() {
252        assert!(is_remote_url("https://cdn.example.com/style.css"));
253    }
254
255    #[test]
256    fn remote_url_protocol_relative() {
257        assert!(is_remote_url("//cdn.example.com/lib.js"));
258    }
259
260    #[test]
261    fn remote_url_data() {
262        assert!(is_remote_url("data:text/javascript;base64,abc"));
263    }
264
265    #[test]
266    fn local_relative_not_remote() {
267        assert!(!is_remote_url("./src/entry.js"));
268    }
269
270    #[test]
271    fn local_root_relative_not_remote() {
272        assert!(!is_remote_url("/src/entry.js"));
273    }
274
275    #[test]
276    fn extracts_module_script_src() {
277        let info = parse_html_to_module(
278            FileId(0),
279            r#"<script type="module" src="./src/entry.js"></script>"#,
280            0,
281        );
282        assert_eq!(info.imports.len(), 1);
283        assert_eq!(info.imports[0].source, "./src/entry.js");
284    }
285
286    #[test]
287    fn extracts_plain_script_src() {
288        let info = parse_html_to_module(
289            FileId(0),
290            r#"<script src="./src/polyfills.js"></script>"#,
291            0,
292        );
293        assert_eq!(info.imports.len(), 1);
294        assert_eq!(info.imports[0].source, "./src/polyfills.js");
295    }
296
297    #[test]
298    fn extracts_multiple_scripts() {
299        let info = parse_html_to_module(
300            FileId(0),
301            r#"
302            <script type="module" src="./src/entry.js"></script>
303            <script src="./src/polyfills.js"></script>
304            "#,
305            0,
306        );
307        assert_eq!(info.imports.len(), 2);
308    }
309
310    #[test]
311    fn skips_inline_script() {
312        let info = parse_html_to_module(FileId(0), r#"<script>console.log("hello");</script>"#, 0);
313        assert!(info.imports.is_empty());
314    }
315
316    #[test]
317    fn skips_handlebars_placeholder_in_script_src() {
318        let info = parse_html_to_module(
319            FileId(0),
320            r#"<script src="{{rootURL}}assets/app.js"></script>
321               <script src="{{config.assetsPath}}vendor.js"></script>"#,
322            0,
323        );
324        assert!(
325            info.imports.is_empty(),
326            "Handlebars-placeholder script srcs should not enter the import graph; got {:?}",
327            info.imports
328        );
329    }
330
331    #[test]
332    fn skips_handlebars_placeholder_in_link_href() {
333        let info = parse_html_to_module(
334            FileId(0),
335            r#"<link rel="stylesheet" href="{{rootURL}}assets/app.css">"#,
336            0,
337        );
338        assert!(info.imports.is_empty());
339    }
340
341    #[test]
342    fn skips_ember_cli_blueprint_placeholder() {
343        let info = parse_html_to_module(
344            FileId(0),
345            r####"<script src="###APPNAME###/app.js"></script>"####,
346            0,
347        );
348        assert!(info.imports.is_empty());
349    }
350
351    #[test]
352    fn extracts_normal_specifier_alongside_placeholders() {
353        let info = parse_html_to_module(
354            FileId(0),
355            r#"<script src="{{rootURL}}assets/app.js"></script>
356               <script src="./src/main.ts"></script>"#,
357            0,
358        );
359        assert_eq!(info.imports.len(), 1);
360        assert_eq!(info.imports[0].source, "./src/main.ts");
361    }
362
363    #[test]
364    fn skips_remote_script() {
365        let info = parse_html_to_module(
366            FileId(0),
367            r#"<script src="https://cdn.example.com/lib.js"></script>"#,
368            0,
369        );
370        assert!(info.imports.is_empty());
371    }
372
373    #[test]
374    fn skips_protocol_relative_script() {
375        let info = parse_html_to_module(
376            FileId(0),
377            r#"<script src="//cdn.example.com/lib.js"></script>"#,
378            0,
379        );
380        assert!(info.imports.is_empty());
381    }
382
383    #[test]
384    fn extracts_stylesheet_link() {
385        let info = parse_html_to_module(
386            FileId(0),
387            r#"<link rel="stylesheet" href="./src/global.css" />"#,
388            0,
389        );
390        assert_eq!(info.imports.len(), 1);
391        assert_eq!(info.imports[0].source, "./src/global.css");
392    }
393
394    #[test]
395    fn extracts_modulepreload_link() {
396        let info = parse_html_to_module(
397            FileId(0),
398            r#"<link rel="modulepreload" href="./src/vendor.js" />"#,
399            0,
400        );
401        assert_eq!(info.imports.len(), 1);
402        assert_eq!(info.imports[0].source, "./src/vendor.js");
403    }
404
405    #[test]
406    fn extracts_link_with_reversed_attrs() {
407        let info = parse_html_to_module(
408            FileId(0),
409            r#"<link href="./src/global.css" rel="stylesheet" />"#,
410            0,
411        );
412        assert_eq!(info.imports.len(), 1);
413        assert_eq!(info.imports[0].source, "./src/global.css");
414    }
415
416    #[test]
417    fn bare_script_src_normalized_to_relative() {
418        let info = parse_html_to_module(FileId(0), r#"<script src="app.js"></script>"#, 0);
419        assert_eq!(info.imports.len(), 1);
420        assert_eq!(info.imports[0].source, "./app.js");
421    }
422
423    #[test]
424    fn bare_module_script_src_normalized_to_relative() {
425        let info = parse_html_to_module(
426            FileId(0),
427            r#"<script type="module" src="main.ts"></script>"#,
428            0,
429        );
430        assert_eq!(info.imports.len(), 1);
431        assert_eq!(info.imports[0].source, "./main.ts");
432    }
433
434    #[test]
435    fn bare_stylesheet_link_href_normalized_to_relative() {
436        let info = parse_html_to_module(
437            FileId(0),
438            r#"<link rel="stylesheet" href="styles.css" />"#,
439            0,
440        );
441        assert_eq!(info.imports.len(), 1);
442        assert_eq!(info.imports[0].source, "./styles.css");
443    }
444
445    #[test]
446    fn bare_link_href_reversed_attrs_normalized_to_relative() {
447        let info = parse_html_to_module(
448            FileId(0),
449            r#"<link href="styles.css" rel="stylesheet" />"#,
450            0,
451        );
452        assert_eq!(info.imports.len(), 1);
453        assert_eq!(info.imports[0].source, "./styles.css");
454    }
455
456    #[test]
457    fn bare_modulepreload_link_href_normalized_to_relative() {
458        let info = parse_html_to_module(
459            FileId(0),
460            r#"<link rel="modulepreload" href="vendor.js" />"#,
461            0,
462        );
463        assert_eq!(info.imports.len(), 1);
464        assert_eq!(info.imports[0].source, "./vendor.js");
465    }
466
467    #[test]
468    fn bare_asset_with_subdir_normalized_to_relative() {
469        let info = parse_html_to_module(FileId(0), r#"<script src="assets/app.js"></script>"#, 0);
470        assert_eq!(info.imports.len(), 1);
471        assert_eq!(info.imports[0].source, "./assets/app.js");
472    }
473
474    #[test]
475    fn root_absolute_script_src_unchanged() {
476        let info = parse_html_to_module(FileId(0), r#"<script src="/src/main.ts"></script>"#, 0);
477        assert_eq!(info.imports.len(), 1);
478        assert_eq!(info.imports[0].source, "/src/main.ts");
479    }
480
481    #[test]
482    fn parent_relative_script_src_unchanged() {
483        let info = parse_html_to_module(
484            FileId(0),
485            r#"<script src="../shared/vendor.js"></script>"#,
486            0,
487        );
488        assert_eq!(info.imports.len(), 1);
489        assert_eq!(info.imports[0].source, "../shared/vendor.js");
490    }
491
492    #[test]
493    fn skips_preload_link() {
494        let info = parse_html_to_module(
495            FileId(0),
496            r#"<link rel="preload" href="./src/font.woff2" as="font" />"#,
497            0,
498        );
499        assert!(info.imports.is_empty());
500    }
501
502    #[test]
503    fn skips_icon_link() {
504        let info =
505            parse_html_to_module(FileId(0), r#"<link rel="icon" href="./favicon.ico" />"#, 0);
506        assert!(info.imports.is_empty());
507    }
508
509    #[test]
510    fn skips_remote_stylesheet() {
511        let info = parse_html_to_module(
512            FileId(0),
513            r#"<link rel="stylesheet" href="https://fonts.googleapis.com/css" />"#,
514            0,
515        );
516        assert!(info.imports.is_empty());
517    }
518
519    #[test]
520    fn skips_commented_out_script() {
521        let info = parse_html_to_module(
522            FileId(0),
523            r#"<!-- <script src="./old.js"></script> -->
524            <script src="./new.js"></script>"#,
525            0,
526        );
527        assert_eq!(info.imports.len(), 1);
528        assert_eq!(info.imports[0].source, "./new.js");
529    }
530
531    #[test]
532    fn skips_commented_out_link() {
533        let info = parse_html_to_module(
534            FileId(0),
535            r#"<!-- <link rel="stylesheet" href="./old.css" /> -->
536            <link rel="stylesheet" href="./new.css" />"#,
537            0,
538        );
539        assert_eq!(info.imports.len(), 1);
540        assert_eq!(info.imports[0].source, "./new.css");
541    }
542
543    #[test]
544    fn handles_multiline_script_tag() {
545        let info = parse_html_to_module(
546            FileId(0),
547            "<script\n  type=\"module\"\n  src=\"./src/entry.js\"\n></script>",
548            0,
549        );
550        assert_eq!(info.imports.len(), 1);
551        assert_eq!(info.imports[0].source, "./src/entry.js");
552    }
553
554    #[test]
555    fn handles_multiline_link_tag() {
556        let info = parse_html_to_module(
557            FileId(0),
558            "<link\n  rel=\"stylesheet\"\n  href=\"./src/global.css\"\n/>",
559            0,
560        );
561        assert_eq!(info.imports.len(), 1);
562        assert_eq!(info.imports[0].source, "./src/global.css");
563    }
564
565    #[test]
566    fn full_vite_html() {
567        let info = parse_html_to_module(
568            FileId(0),
569            r#"<!doctype html>
570<html>
571  <head>
572    <link rel="stylesheet" href="./src/global.css" />
573    <link rel="icon" href="/favicon.ico" />
574  </head>
575  <body>
576    <div id="app"></div>
577    <script type="module" src="./src/entry.js"></script>
578  </body>
579</html>"#,
580            0,
581        );
582        assert_eq!(info.imports.len(), 2);
583        let sources: Vec<&str> = info.imports.iter().map(|i| i.source.as_str()).collect();
584        assert!(sources.contains(&"./src/global.css"));
585        assert!(sources.contains(&"./src/entry.js"));
586    }
587
588    #[test]
589    fn empty_html() {
590        let info = parse_html_to_module(FileId(0), "", 0);
591        assert!(info.imports.is_empty());
592    }
593
594    #[test]
595    fn html_with_no_assets() {
596        let info = parse_html_to_module(
597            FileId(0),
598            r"<!doctype html><html><body><h1>Hello</h1></body></html>",
599            0,
600        );
601        assert!(info.imports.is_empty());
602    }
603
604    #[test]
605    fn single_quoted_attributes() {
606        let info = parse_html_to_module(FileId(0), r"<script src='./src/entry.js'></script>", 0);
607        assert_eq!(info.imports.len(), 1);
608        assert_eq!(info.imports[0].source, "./src/entry.js");
609    }
610
611    #[test]
612    fn all_imports_are_side_effect() {
613        let info = parse_html_to_module(
614            FileId(0),
615            r#"<script src="./entry.js"></script>
616            <link rel="stylesheet" href="./style.css" />"#,
617            0,
618        );
619        for imp in &info.imports {
620            assert!(matches!(imp.imported_name, ImportedName::SideEffect));
621            assert!(imp.local_name.is_empty());
622            assert!(!imp.is_type_only);
623        }
624    }
625
626    #[test]
627    fn suppression_comments_extracted() {
628        let info = parse_html_to_module(
629            FileId(0),
630            "<!-- fallow-ignore-file -->\n<script src=\"./entry.js\"></script>",
631            0,
632        );
633        assert_eq!(info.imports.len(), 1);
634    }
635
636    #[test]
637    fn angular_template_extracts_member_refs() {
638        let info = parse_html_to_module(
639            FileId(0),
640            "<h1>{{ title() }}</h1>\n\
641             <p [class.highlighted]=\"isHighlighted\">{{ greeting() }}</p>\n\
642             <button (click)=\"onButtonClick()\">Toggle</button>",
643            0,
644        );
645        let names: rustc_hash::FxHashSet<&str> = info
646            .member_accesses
647            .iter()
648            .filter(|a| a.object == ANGULAR_TPL_SENTINEL)
649            .map(|a| a.member.as_str())
650            .collect();
651        assert!(names.contains("title"), "should contain 'title'");
652        assert!(
653            names.contains("isHighlighted"),
654            "should contain 'isHighlighted'"
655        );
656        assert!(names.contains("greeting"), "should contain 'greeting'");
657        assert!(
658            names.contains("onButtonClick"),
659            "should contain 'onButtonClick'"
660        );
661    }
662
663    #[test]
664    fn plain_html_no_angular_refs() {
665        let info = parse_html_to_module(
666            FileId(0),
667            "<!doctype html><html><body><h1>Hello</h1></body></html>",
668            0,
669        );
670        assert!(info.member_accesses.is_empty());
671    }
672}
fallow_extract/html.rs

fallow_extract/
html.rs
