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        tainted_bindings: Vec::new(),
201        sanitized_sink_args: Vec::new(),
202    }
203}
204
205#[cfg(test)]
206mod tests {
207    use super::*;
208
209    #[test]
210    fn is_html_file_html() {
211        assert!(is_html_file(Path::new("index.html")));
212    }
213
214    #[test]
215    fn is_html_file_nested() {
216        assert!(is_html_file(Path::new("pages/about.html")));
217    }
218
219    #[test]
220    fn is_html_file_rejects_htm() {
221        assert!(!is_html_file(Path::new("index.htm")));
222    }
223
224    #[test]
225    fn is_html_file_rejects_js() {
226        assert!(!is_html_file(Path::new("app.js")));
227    }
228
229    #[test]
230    fn is_html_file_rejects_ts() {
231        assert!(!is_html_file(Path::new("app.ts")));
232    }
233
234    #[test]
235    fn is_html_file_rejects_vue() {
236        assert!(!is_html_file(Path::new("App.vue")));
237    }
238
239    #[test]
240    fn remote_url_http() {
241        assert!(is_remote_url("http://example.com/script.js"));
242    }
243
244    #[test]
245    fn remote_url_https() {
246        assert!(is_remote_url("https://cdn.example.com/style.css"));
247    }
248
249    #[test]
250    fn remote_url_protocol_relative() {
251        assert!(is_remote_url("//cdn.example.com/lib.js"));
252    }
253
254    #[test]
255    fn remote_url_data() {
256        assert!(is_remote_url("data:text/javascript;base64,abc"));
257    }
258
259    #[test]
260    fn local_relative_not_remote() {
261        assert!(!is_remote_url("./src/entry.js"));
262    }
263
264    #[test]
265    fn local_root_relative_not_remote() {
266        assert!(!is_remote_url("/src/entry.js"));
267    }
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_handlebars_placeholder_in_script_src() {
312        let info = parse_html_to_module(
313            FileId(0),
314            r#"<script src="{{rootURL}}assets/app.js"></script>
315               <script src="{{config.assetsPath}}vendor.js"></script>"#,
316            0,
317        );
318        assert!(
319            info.imports.is_empty(),
320            "Handlebars-placeholder script srcs should not enter the import graph; got {:?}",
321            info.imports
322        );
323    }
324
325    #[test]
326    fn skips_handlebars_placeholder_in_link_href() {
327        let info = parse_html_to_module(
328            FileId(0),
329            r#"<link rel="stylesheet" href="{{rootURL}}assets/app.css">"#,
330            0,
331        );
332        assert!(info.imports.is_empty());
333    }
334
335    #[test]
336    fn skips_ember_cli_blueprint_placeholder() {
337        let info = parse_html_to_module(
338            FileId(0),
339            r####"<script src="###APPNAME###/app.js"></script>"####,
340            0,
341        );
342        assert!(info.imports.is_empty());
343    }
344
345    #[test]
346    fn extracts_normal_specifier_alongside_placeholders() {
347        let info = parse_html_to_module(
348            FileId(0),
349            r#"<script src="{{rootURL}}assets/app.js"></script>
350               <script src="./src/main.ts"></script>"#,
351            0,
352        );
353        assert_eq!(info.imports.len(), 1);
354        assert_eq!(info.imports[0].source, "./src/main.ts");
355    }
356
357    #[test]
358    fn skips_remote_script() {
359        let info = parse_html_to_module(
360            FileId(0),
361            r#"<script src="https://cdn.example.com/lib.js"></script>"#,
362            0,
363        );
364        assert!(info.imports.is_empty());
365    }
366
367    #[test]
368    fn skips_protocol_relative_script() {
369        let info = parse_html_to_module(
370            FileId(0),
371            r#"<script src="//cdn.example.com/lib.js"></script>"#,
372            0,
373        );
374        assert!(info.imports.is_empty());
375    }
376
377    #[test]
378    fn extracts_stylesheet_link() {
379        let info = parse_html_to_module(
380            FileId(0),
381            r#"<link rel="stylesheet" href="./src/global.css" />"#,
382            0,
383        );
384        assert_eq!(info.imports.len(), 1);
385        assert_eq!(info.imports[0].source, "./src/global.css");
386    }
387
388    #[test]
389    fn extracts_modulepreload_link() {
390        let info = parse_html_to_module(
391            FileId(0),
392            r#"<link rel="modulepreload" href="./src/vendor.js" />"#,
393            0,
394        );
395        assert_eq!(info.imports.len(), 1);
396        assert_eq!(info.imports[0].source, "./src/vendor.js");
397    }
398
399    #[test]
400    fn extracts_link_with_reversed_attrs() {
401        let info = parse_html_to_module(
402            FileId(0),
403            r#"<link href="./src/global.css" rel="stylesheet" />"#,
404            0,
405        );
406        assert_eq!(info.imports.len(), 1);
407        assert_eq!(info.imports[0].source, "./src/global.css");
408    }
409
410    #[test]
411    fn bare_script_src_normalized_to_relative() {
412        let info = parse_html_to_module(FileId(0), r#"<script src="app.js"></script>"#, 0);
413        assert_eq!(info.imports.len(), 1);
414        assert_eq!(info.imports[0].source, "./app.js");
415    }
416
417    #[test]
418    fn bare_module_script_src_normalized_to_relative() {
419        let info = parse_html_to_module(
420            FileId(0),
421            r#"<script type="module" src="main.ts"></script>"#,
422            0,
423        );
424        assert_eq!(info.imports.len(), 1);
425        assert_eq!(info.imports[0].source, "./main.ts");
426    }
427
428    #[test]
429    fn bare_stylesheet_link_href_normalized_to_relative() {
430        let info = parse_html_to_module(
431            FileId(0),
432            r#"<link rel="stylesheet" href="styles.css" />"#,
433            0,
434        );
435        assert_eq!(info.imports.len(), 1);
436        assert_eq!(info.imports[0].source, "./styles.css");
437    }
438
439    #[test]
440    fn bare_link_href_reversed_attrs_normalized_to_relative() {
441        let info = parse_html_to_module(
442            FileId(0),
443            r#"<link href="styles.css" rel="stylesheet" />"#,
444            0,
445        );
446        assert_eq!(info.imports.len(), 1);
447        assert_eq!(info.imports[0].source, "./styles.css");
448    }
449
450    #[test]
451    fn bare_modulepreload_link_href_normalized_to_relative() {
452        let info = parse_html_to_module(
453            FileId(0),
454            r#"<link rel="modulepreload" href="vendor.js" />"#,
455            0,
456        );
457        assert_eq!(info.imports.len(), 1);
458        assert_eq!(info.imports[0].source, "./vendor.js");
459    }
460
461    #[test]
462    fn bare_asset_with_subdir_normalized_to_relative() {
463        let info = parse_html_to_module(FileId(0), r#"<script src="assets/app.js"></script>"#, 0);
464        assert_eq!(info.imports.len(), 1);
465        assert_eq!(info.imports[0].source, "./assets/app.js");
466    }
467
468    #[test]
469    fn root_absolute_script_src_unchanged() {
470        let info = parse_html_to_module(FileId(0), r#"<script src="/src/main.ts"></script>"#, 0);
471        assert_eq!(info.imports.len(), 1);
472        assert_eq!(info.imports[0].source, "/src/main.ts");
473    }
474
475    #[test]
476    fn parent_relative_script_src_unchanged() {
477        let info = parse_html_to_module(
478            FileId(0),
479            r#"<script src="../shared/vendor.js"></script>"#,
480            0,
481        );
482        assert_eq!(info.imports.len(), 1);
483        assert_eq!(info.imports[0].source, "../shared/vendor.js");
484    }
485
486    #[test]
487    fn skips_preload_link() {
488        let info = parse_html_to_module(
489            FileId(0),
490            r#"<link rel="preload" href="./src/font.woff2" as="font" />"#,
491            0,
492        );
493        assert!(info.imports.is_empty());
494    }
495
496    #[test]
497    fn skips_icon_link() {
498        let info =
499            parse_html_to_module(FileId(0), r#"<link rel="icon" href="./favicon.ico" />"#, 0);
500        assert!(info.imports.is_empty());
501    }
502
503    #[test]
504    fn skips_remote_stylesheet() {
505        let info = parse_html_to_module(
506            FileId(0),
507            r#"<link rel="stylesheet" href="https://fonts.googleapis.com/css" />"#,
508            0,
509        );
510        assert!(info.imports.is_empty());
511    }
512
513    #[test]
514    fn skips_commented_out_script() {
515        let info = parse_html_to_module(
516            FileId(0),
517            r#"<!-- <script src="./old.js"></script> -->
518            <script src="./new.js"></script>"#,
519            0,
520        );
521        assert_eq!(info.imports.len(), 1);
522        assert_eq!(info.imports[0].source, "./new.js");
523    }
524
525    #[test]
526    fn skips_commented_out_link() {
527        let info = parse_html_to_module(
528            FileId(0),
529            r#"<!-- <link rel="stylesheet" href="./old.css" /> -->
530            <link rel="stylesheet" href="./new.css" />"#,
531            0,
532        );
533        assert_eq!(info.imports.len(), 1);
534        assert_eq!(info.imports[0].source, "./new.css");
535    }
536
537    #[test]
538    fn handles_multiline_script_tag() {
539        let info = parse_html_to_module(
540            FileId(0),
541            "<script\n  type=\"module\"\n  src=\"./src/entry.js\"\n></script>",
542            0,
543        );
544        assert_eq!(info.imports.len(), 1);
545        assert_eq!(info.imports[0].source, "./src/entry.js");
546    }
547
548    #[test]
549    fn handles_multiline_link_tag() {
550        let info = parse_html_to_module(
551            FileId(0),
552            "<link\n  rel=\"stylesheet\"\n  href=\"./src/global.css\"\n/>",
553            0,
554        );
555        assert_eq!(info.imports.len(), 1);
556        assert_eq!(info.imports[0].source, "./src/global.css");
557    }
558
559    #[test]
560    fn full_vite_html() {
561        let info = parse_html_to_module(
562            FileId(0),
563            r#"<!doctype html>
564<html>
565  <head>
566    <link rel="stylesheet" href="./src/global.css" />
567    <link rel="icon" href="/favicon.ico" />
568  </head>
569  <body>
570    <div id="app"></div>
571    <script type="module" src="./src/entry.js"></script>
572  </body>
573</html>"#,
574            0,
575        );
576        assert_eq!(info.imports.len(), 2);
577        let sources: Vec<&str> = info.imports.iter().map(|i| i.source.as_str()).collect();
578        assert!(sources.contains(&"./src/global.css"));
579        assert!(sources.contains(&"./src/entry.js"));
580    }
581
582    #[test]
583    fn empty_html() {
584        let info = parse_html_to_module(FileId(0), "", 0);
585        assert!(info.imports.is_empty());
586    }
587
588    #[test]
589    fn html_with_no_assets() {
590        let info = parse_html_to_module(
591            FileId(0),
592            r"<!doctype html><html><body><h1>Hello</h1></body></html>",
593            0,
594        );
595        assert!(info.imports.is_empty());
596    }
597
598    #[test]
599    fn single_quoted_attributes() {
600        let info = parse_html_to_module(FileId(0), r"<script src='./src/entry.js'></script>", 0);
601        assert_eq!(info.imports.len(), 1);
602        assert_eq!(info.imports[0].source, "./src/entry.js");
603    }
604
605    #[test]
606    fn all_imports_are_side_effect() {
607        let info = parse_html_to_module(
608            FileId(0),
609            r#"<script src="./entry.js"></script>
610            <link rel="stylesheet" href="./style.css" />"#,
611            0,
612        );
613        for imp in &info.imports {
614            assert!(matches!(imp.imported_name, ImportedName::SideEffect));
615            assert!(imp.local_name.is_empty());
616            assert!(!imp.is_type_only);
617        }
618    }
619
620    #[test]
621    fn suppression_comments_extracted() {
622        let info = parse_html_to_module(
623            FileId(0),
624            "<!-- fallow-ignore-file -->\n<script src=\"./entry.js\"></script>",
625            0,
626        );
627        assert_eq!(info.imports.len(), 1);
628    }
629
630    #[test]
631    fn angular_template_extracts_member_refs() {
632        let info = parse_html_to_module(
633            FileId(0),
634            "<h1>{{ title() }}</h1>\n\
635             <p [class.highlighted]=\"isHighlighted\">{{ greeting() }}</p>\n\
636             <button (click)=\"onButtonClick()\">Toggle</button>",
637            0,
638        );
639        let names: rustc_hash::FxHashSet<&str> = info
640            .member_accesses
641            .iter()
642            .filter(|a| a.object == ANGULAR_TPL_SENTINEL)
643            .map(|a| a.member.as_str())
644            .collect();
645        assert!(names.contains("title"), "should contain 'title'");
646        assert!(
647            names.contains("isHighlighted"),
648            "should contain 'isHighlighted'"
649        );
650        assert!(names.contains("greeting"), "should contain 'greeting'");
651        assert!(
652            names.contains("onButtonClick"),
653            "should contain 'onButtonClick'"
654        );
655    }
656
657    #[test]
658    fn plain_html_no_angular_refs() {
659        let info = parse_html_to_module(
660            FileId(0),
661            "<!doctype html><html><body><h1>Hello</h1></body></html>",
662            0,
663        );
664        assert!(info.member_accesses.is_empty());
665    }
666}
fallow_extract/html.rs

fallow_extract/
html.rs
