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 template_refs = angular::collect_angular_template_refs(source);
152    let mut member_accesses: Vec<MemberAccess> = template_refs
153        .identifiers
154        .into_iter()
155        .map(|name| MemberAccess {
156            object: ANGULAR_TPL_SENTINEL.to_string(),
157            member: name,
158        })
159        .collect();
160    member_accesses.extend(template_refs.member_accesses);
161
162    let complexity = if need_complexity {
163        crate::template_complexity::compute_angular_template_complexity(source)
164            .into_iter()
165            .collect()
166    } else {
167        Vec::new()
168    };
169
170    ModuleInfo {
171        file_id,
172        exports: Vec::new(),
173        imports,
174        re_exports: Vec::new(),
175        dynamic_imports: Vec::new(),
176        dynamic_import_patterns: Vec::new(),
177        require_calls: Vec::new(),
178        member_accesses,
179        whole_object_uses: Vec::new(),
180        has_cjs_exports: false,
181        has_angular_component_template_url: false,
182        content_hash,
183        suppressions: parsed_suppressions.suppressions,
184        unknown_suppression_kinds: parsed_suppressions.unknown_kinds,
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        namespace_object_aliases: Vec::new(),
195        iconify_prefixes: Vec::new(),
196        auto_import_candidates: Vec::new(),
197        directives: Vec::new(),
198        security_sinks: Vec::new(),
199        security_sinks_skipped: 0,
200    }
201}
202
203#[cfg(test)]
204mod tests {
205    use super::*;
206
207    #[test]
208    fn is_html_file_html() {
209        assert!(is_html_file(Path::new("index.html")));
210    }
211
212    #[test]
213    fn is_html_file_nested() {
214        assert!(is_html_file(Path::new("pages/about.html")));
215    }
216
217    #[test]
218    fn is_html_file_rejects_htm() {
219        assert!(!is_html_file(Path::new("index.htm")));
220    }
221
222    #[test]
223    fn is_html_file_rejects_js() {
224        assert!(!is_html_file(Path::new("app.js")));
225    }
226
227    #[test]
228    fn is_html_file_rejects_ts() {
229        assert!(!is_html_file(Path::new("app.ts")));
230    }
231
232    #[test]
233    fn is_html_file_rejects_vue() {
234        assert!(!is_html_file(Path::new("App.vue")));
235    }
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    #[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_handlebars_placeholder_in_script_src() {
310        let info = parse_html_to_module(
311            FileId(0),
312            r#"<script src="{{rootURL}}assets/app.js"></script>
313               <script src="{{config.assetsPath}}vendor.js"></script>"#,
314            0,
315        );
316        assert!(
317            info.imports.is_empty(),
318            "Handlebars-placeholder script srcs should not enter the import graph; got {:?}",
319            info.imports
320        );
321    }
322
323    #[test]
324    fn skips_handlebars_placeholder_in_link_href() {
325        let info = parse_html_to_module(
326            FileId(0),
327            r#"<link rel="stylesheet" href="{{rootURL}}assets/app.css">"#,
328            0,
329        );
330        assert!(info.imports.is_empty());
331    }
332
333    #[test]
334    fn skips_ember_cli_blueprint_placeholder() {
335        let info = parse_html_to_module(
336            FileId(0),
337            r####"<script src="###APPNAME###/app.js"></script>"####,
338            0,
339        );
340        assert!(info.imports.is_empty());
341    }
342
343    #[test]
344    fn extracts_normal_specifier_alongside_placeholders() {
345        let info = parse_html_to_module(
346            FileId(0),
347            r#"<script src="{{rootURL}}assets/app.js"></script>
348               <script src="./src/main.ts"></script>"#,
349            0,
350        );
351        assert_eq!(info.imports.len(), 1);
352        assert_eq!(info.imports[0].source, "./src/main.ts");
353    }
354
355    #[test]
356    fn skips_remote_script() {
357        let info = parse_html_to_module(
358            FileId(0),
359            r#"<script src="https://cdn.example.com/lib.js"></script>"#,
360            0,
361        );
362        assert!(info.imports.is_empty());
363    }
364
365    #[test]
366    fn skips_protocol_relative_script() {
367        let info = parse_html_to_module(
368            FileId(0),
369            r#"<script src="//cdn.example.com/lib.js"></script>"#,
370            0,
371        );
372        assert!(info.imports.is_empty());
373    }
374
375    #[test]
376    fn extracts_stylesheet_link() {
377        let info = parse_html_to_module(
378            FileId(0),
379            r#"<link rel="stylesheet" href="./src/global.css" />"#,
380            0,
381        );
382        assert_eq!(info.imports.len(), 1);
383        assert_eq!(info.imports[0].source, "./src/global.css");
384    }
385
386    #[test]
387    fn extracts_modulepreload_link() {
388        let info = parse_html_to_module(
389            FileId(0),
390            r#"<link rel="modulepreload" href="./src/vendor.js" />"#,
391            0,
392        );
393        assert_eq!(info.imports.len(), 1);
394        assert_eq!(info.imports[0].source, "./src/vendor.js");
395    }
396
397    #[test]
398    fn extracts_link_with_reversed_attrs() {
399        let info = parse_html_to_module(
400            FileId(0),
401            r#"<link href="./src/global.css" rel="stylesheet" />"#,
402            0,
403        );
404        assert_eq!(info.imports.len(), 1);
405        assert_eq!(info.imports[0].source, "./src/global.css");
406    }
407
408    #[test]
409    fn bare_script_src_normalized_to_relative() {
410        let info = parse_html_to_module(FileId(0), r#"<script src="app.js"></script>"#, 0);
411        assert_eq!(info.imports.len(), 1);
412        assert_eq!(info.imports[0].source, "./app.js");
413    }
414
415    #[test]
416    fn bare_module_script_src_normalized_to_relative() {
417        let info = parse_html_to_module(
418            FileId(0),
419            r#"<script type="module" src="main.ts"></script>"#,
420            0,
421        );
422        assert_eq!(info.imports.len(), 1);
423        assert_eq!(info.imports[0].source, "./main.ts");
424    }
425
426    #[test]
427    fn bare_stylesheet_link_href_normalized_to_relative() {
428        let info = parse_html_to_module(
429            FileId(0),
430            r#"<link rel="stylesheet" href="styles.css" />"#,
431            0,
432        );
433        assert_eq!(info.imports.len(), 1);
434        assert_eq!(info.imports[0].source, "./styles.css");
435    }
436
437    #[test]
438    fn bare_link_href_reversed_attrs_normalized_to_relative() {
439        let info = parse_html_to_module(
440            FileId(0),
441            r#"<link href="styles.css" rel="stylesheet" />"#,
442            0,
443        );
444        assert_eq!(info.imports.len(), 1);
445        assert_eq!(info.imports[0].source, "./styles.css");
446    }
447
448    #[test]
449    fn bare_modulepreload_link_href_normalized_to_relative() {
450        let info = parse_html_to_module(
451            FileId(0),
452            r#"<link rel="modulepreload" href="vendor.js" />"#,
453            0,
454        );
455        assert_eq!(info.imports.len(), 1);
456        assert_eq!(info.imports[0].source, "./vendor.js");
457    }
458
459    #[test]
460    fn bare_asset_with_subdir_normalized_to_relative() {
461        let info = parse_html_to_module(FileId(0), r#"<script src="assets/app.js"></script>"#, 0);
462        assert_eq!(info.imports.len(), 1);
463        assert_eq!(info.imports[0].source, "./assets/app.js");
464    }
465
466    #[test]
467    fn root_absolute_script_src_unchanged() {
468        let info = parse_html_to_module(FileId(0), r#"<script src="/src/main.ts"></script>"#, 0);
469        assert_eq!(info.imports.len(), 1);
470        assert_eq!(info.imports[0].source, "/src/main.ts");
471    }
472
473    #[test]
474    fn parent_relative_script_src_unchanged() {
475        let info = parse_html_to_module(
476            FileId(0),
477            r#"<script src="../shared/vendor.js"></script>"#,
478            0,
479        );
480        assert_eq!(info.imports.len(), 1);
481        assert_eq!(info.imports[0].source, "../shared/vendor.js");
482    }
483
484    #[test]
485    fn skips_preload_link() {
486        let info = parse_html_to_module(
487            FileId(0),
488            r#"<link rel="preload" href="./src/font.woff2" as="font" />"#,
489            0,
490        );
491        assert!(info.imports.is_empty());
492    }
493
494    #[test]
495    fn skips_icon_link() {
496        let info =
497            parse_html_to_module(FileId(0), r#"<link rel="icon" href="./favicon.ico" />"#, 0);
498        assert!(info.imports.is_empty());
499    }
500
501    #[test]
502    fn skips_remote_stylesheet() {
503        let info = parse_html_to_module(
504            FileId(0),
505            r#"<link rel="stylesheet" href="https://fonts.googleapis.com/css" />"#,
506            0,
507        );
508        assert!(info.imports.is_empty());
509    }
510
511    #[test]
512    fn skips_commented_out_script() {
513        let info = parse_html_to_module(
514            FileId(0),
515            r#"<!-- <script src="./old.js"></script> -->
516            <script src="./new.js"></script>"#,
517            0,
518        );
519        assert_eq!(info.imports.len(), 1);
520        assert_eq!(info.imports[0].source, "./new.js");
521    }
522
523    #[test]
524    fn skips_commented_out_link() {
525        let info = parse_html_to_module(
526            FileId(0),
527            r#"<!-- <link rel="stylesheet" href="./old.css" /> -->
528            <link rel="stylesheet" href="./new.css" />"#,
529            0,
530        );
531        assert_eq!(info.imports.len(), 1);
532        assert_eq!(info.imports[0].source, "./new.css");
533    }
534
535    #[test]
536    fn handles_multiline_script_tag() {
537        let info = parse_html_to_module(
538            FileId(0),
539            "<script\n  type=\"module\"\n  src=\"./src/entry.js\"\n></script>",
540            0,
541        );
542        assert_eq!(info.imports.len(), 1);
543        assert_eq!(info.imports[0].source, "./src/entry.js");
544    }
545
546    #[test]
547    fn handles_multiline_link_tag() {
548        let info = parse_html_to_module(
549            FileId(0),
550            "<link\n  rel=\"stylesheet\"\n  href=\"./src/global.css\"\n/>",
551            0,
552        );
553        assert_eq!(info.imports.len(), 1);
554        assert_eq!(info.imports[0].source, "./src/global.css");
555    }
556
557    #[test]
558    fn full_vite_html() {
559        let info = parse_html_to_module(
560            FileId(0),
561            r#"<!doctype html>
562<html>
563  <head>
564    <link rel="stylesheet" href="./src/global.css" />
565    <link rel="icon" href="/favicon.ico" />
566  </head>
567  <body>
568    <div id="app"></div>
569    <script type="module" src="./src/entry.js"></script>
570  </body>
571</html>"#,
572            0,
573        );
574        assert_eq!(info.imports.len(), 2);
575        let sources: Vec<&str> = info.imports.iter().map(|i| i.source.as_str()).collect();
576        assert!(sources.contains(&"./src/global.css"));
577        assert!(sources.contains(&"./src/entry.js"));
578    }
579
580    #[test]
581    fn empty_html() {
582        let info = parse_html_to_module(FileId(0), "", 0);
583        assert!(info.imports.is_empty());
584    }
585
586    #[test]
587    fn html_with_no_assets() {
588        let info = parse_html_to_module(
589            FileId(0),
590            r"<!doctype html><html><body><h1>Hello</h1></body></html>",
591            0,
592        );
593        assert!(info.imports.is_empty());
594    }
595
596    #[test]
597    fn single_quoted_attributes() {
598        let info = parse_html_to_module(FileId(0), r"<script src='./src/entry.js'></script>", 0);
599        assert_eq!(info.imports.len(), 1);
600        assert_eq!(info.imports[0].source, "./src/entry.js");
601    }
602
603    #[test]
604    fn all_imports_are_side_effect() {
605        let info = parse_html_to_module(
606            FileId(0),
607            r#"<script src="./entry.js"></script>
608            <link rel="stylesheet" href="./style.css" />"#,
609            0,
610        );
611        for imp in &info.imports {
612            assert!(matches!(imp.imported_name, ImportedName::SideEffect));
613            assert!(imp.local_name.is_empty());
614            assert!(!imp.is_type_only);
615        }
616    }
617
618    #[test]
619    fn suppression_comments_extracted() {
620        let info = parse_html_to_module(
621            FileId(0),
622            "<!-- fallow-ignore-file -->\n<script src=\"./entry.js\"></script>",
623            0,
624        );
625        assert_eq!(info.imports.len(), 1);
626    }
627
628    #[test]
629    fn angular_template_extracts_member_refs() {
630        let info = parse_html_to_module(
631            FileId(0),
632            "<h1>{{ title() }}</h1>\n\
633             <p [class.highlighted]=\"isHighlighted\">{{ greeting() }}</p>\n\
634             <button (click)=\"onButtonClick()\">Toggle</button>",
635            0,
636        );
637        let names: rustc_hash::FxHashSet<&str> = info
638            .member_accesses
639            .iter()
640            .filter(|a| a.object == ANGULAR_TPL_SENTINEL)
641            .map(|a| a.member.as_str())
642            .collect();
643        assert!(names.contains("title"), "should contain 'title'");
644        assert!(
645            names.contains("isHighlighted"),
646            "should contain 'isHighlighted'"
647        );
648        assert!(names.contains("greeting"), "should contain 'greeting'");
649        assert!(
650            names.contains("onButtonClick"),
651            "should contain 'onButtonClick'"
652        );
653    }
654
655    #[test]
656    fn plain_html_no_angular_refs() {
657        let info = parse_html_to_module(
658            FileId(0),
659            "<!doctype html><html><body><h1>Hello</h1></body></html>",
660            0,
661        );
662        assert!(info.member_accesses.is_empty());
663    }
664}
fallow_extract/html.rs

fallow_extract/
html.rs
