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        security_control_sites: Vec::new(),
209    }
210}
211
212#[cfg(test)]
213mod tests {
214    use super::*;
215
216    #[test]
217    fn is_html_file_html() {
218        assert!(is_html_file(Path::new("index.html")));
219    }
220
221    #[test]
222    fn is_html_file_nested() {
223        assert!(is_html_file(Path::new("pages/about.html")));
224    }
225
226    #[test]
227    fn is_html_file_rejects_htm() {
228        assert!(!is_html_file(Path::new("index.htm")));
229    }
230
231    #[test]
232    fn is_html_file_rejects_js() {
233        assert!(!is_html_file(Path::new("app.js")));
234    }
235
236    #[test]
237    fn is_html_file_rejects_ts() {
238        assert!(!is_html_file(Path::new("app.ts")));
239    }
240
241    #[test]
242    fn is_html_file_rejects_vue() {
243        assert!(!is_html_file(Path::new("App.vue")));
244    }
245
246    #[test]
247    fn remote_url_http() {
248        assert!(is_remote_url("http://example.com/script.js"));
249    }
250
251    #[test]
252    fn remote_url_https() {
253        assert!(is_remote_url("https://cdn.example.com/style.css"));
254    }
255
256    #[test]
257    fn remote_url_protocol_relative() {
258        assert!(is_remote_url("//cdn.example.com/lib.js"));
259    }
260
261    #[test]
262    fn remote_url_data() {
263        assert!(is_remote_url("data:text/javascript;base64,abc"));
264    }
265
266    #[test]
267    fn local_relative_not_remote() {
268        assert!(!is_remote_url("./src/entry.js"));
269    }
270
271    #[test]
272    fn local_root_relative_not_remote() {
273        assert!(!is_remote_url("/src/entry.js"));
274    }
275
276    #[test]
277    fn extracts_module_script_src() {
278        let info = parse_html_to_module(
279            FileId(0),
280            r#"<script type="module" src="./src/entry.js"></script>"#,
281            0,
282        );
283        assert_eq!(info.imports.len(), 1);
284        assert_eq!(info.imports[0].source, "./src/entry.js");
285    }
286
287    #[test]
288    fn extracts_plain_script_src() {
289        let info = parse_html_to_module(
290            FileId(0),
291            r#"<script src="./src/polyfills.js"></script>"#,
292            0,
293        );
294        assert_eq!(info.imports.len(), 1);
295        assert_eq!(info.imports[0].source, "./src/polyfills.js");
296    }
297
298    #[test]
299    fn extracts_multiple_scripts() {
300        let info = parse_html_to_module(
301            FileId(0),
302            r#"
303            <script type="module" src="./src/entry.js"></script>
304            <script src="./src/polyfills.js"></script>
305            "#,
306            0,
307        );
308        assert_eq!(info.imports.len(), 2);
309    }
310
311    #[test]
312    fn skips_inline_script() {
313        let info = parse_html_to_module(FileId(0), r#"<script>console.log("hello");</script>"#, 0);
314        assert!(info.imports.is_empty());
315    }
316
317    #[test]
318    fn skips_handlebars_placeholder_in_script_src() {
319        let info = parse_html_to_module(
320            FileId(0),
321            r#"<script src="{{rootURL}}assets/app.js"></script>
322               <script src="{{config.assetsPath}}vendor.js"></script>"#,
323            0,
324        );
325        assert!(
326            info.imports.is_empty(),
327            "Handlebars-placeholder script srcs should not enter the import graph; got {:?}",
328            info.imports
329        );
330    }
331
332    #[test]
333    fn skips_handlebars_placeholder_in_link_href() {
334        let info = parse_html_to_module(
335            FileId(0),
336            r#"<link rel="stylesheet" href="{{rootURL}}assets/app.css">"#,
337            0,
338        );
339        assert!(info.imports.is_empty());
340    }
341
342    #[test]
343    fn skips_ember_cli_blueprint_placeholder() {
344        let info = parse_html_to_module(
345            FileId(0),
346            r####"<script src="###APPNAME###/app.js"></script>"####,
347            0,
348        );
349        assert!(info.imports.is_empty());
350    }
351
352    #[test]
353    fn extracts_normal_specifier_alongside_placeholders() {
354        let info = parse_html_to_module(
355            FileId(0),
356            r#"<script src="{{rootURL}}assets/app.js"></script>
357               <script src="./src/main.ts"></script>"#,
358            0,
359        );
360        assert_eq!(info.imports.len(), 1);
361        assert_eq!(info.imports[0].source, "./src/main.ts");
362    }
363
364    #[test]
365    fn skips_remote_script() {
366        let info = parse_html_to_module(
367            FileId(0),
368            r#"<script src="https://cdn.example.com/lib.js"></script>"#,
369            0,
370        );
371        assert!(info.imports.is_empty());
372    }
373
374    #[test]
375    fn skips_protocol_relative_script() {
376        let info = parse_html_to_module(
377            FileId(0),
378            r#"<script src="//cdn.example.com/lib.js"></script>"#,
379            0,
380        );
381        assert!(info.imports.is_empty());
382    }
383
384    #[test]
385    fn extracts_stylesheet_link() {
386        let info = parse_html_to_module(
387            FileId(0),
388            r#"<link rel="stylesheet" href="./src/global.css" />"#,
389            0,
390        );
391        assert_eq!(info.imports.len(), 1);
392        assert_eq!(info.imports[0].source, "./src/global.css");
393    }
394
395    #[test]
396    fn extracts_modulepreload_link() {
397        let info = parse_html_to_module(
398            FileId(0),
399            r#"<link rel="modulepreload" href="./src/vendor.js" />"#,
400            0,
401        );
402        assert_eq!(info.imports.len(), 1);
403        assert_eq!(info.imports[0].source, "./src/vendor.js");
404    }
405
406    #[test]
407    fn extracts_link_with_reversed_attrs() {
408        let info = parse_html_to_module(
409            FileId(0),
410            r#"<link href="./src/global.css" rel="stylesheet" />"#,
411            0,
412        );
413        assert_eq!(info.imports.len(), 1);
414        assert_eq!(info.imports[0].source, "./src/global.css");
415    }
416
417    #[test]
418    fn bare_script_src_normalized_to_relative() {
419        let info = parse_html_to_module(FileId(0), r#"<script src="app.js"></script>"#, 0);
420        assert_eq!(info.imports.len(), 1);
421        assert_eq!(info.imports[0].source, "./app.js");
422    }
423
424    #[test]
425    fn bare_module_script_src_normalized_to_relative() {
426        let info = parse_html_to_module(
427            FileId(0),
428            r#"<script type="module" src="main.ts"></script>"#,
429            0,
430        );
431        assert_eq!(info.imports.len(), 1);
432        assert_eq!(info.imports[0].source, "./main.ts");
433    }
434
435    #[test]
436    fn bare_stylesheet_link_href_normalized_to_relative() {
437        let info = parse_html_to_module(
438            FileId(0),
439            r#"<link rel="stylesheet" href="styles.css" />"#,
440            0,
441        );
442        assert_eq!(info.imports.len(), 1);
443        assert_eq!(info.imports[0].source, "./styles.css");
444    }
445
446    #[test]
447    fn bare_link_href_reversed_attrs_normalized_to_relative() {
448        let info = parse_html_to_module(
449            FileId(0),
450            r#"<link href="styles.css" rel="stylesheet" />"#,
451            0,
452        );
453        assert_eq!(info.imports.len(), 1);
454        assert_eq!(info.imports[0].source, "./styles.css");
455    }
456
457    #[test]
458    fn bare_modulepreload_link_href_normalized_to_relative() {
459        let info = parse_html_to_module(
460            FileId(0),
461            r#"<link rel="modulepreload" href="vendor.js" />"#,
462            0,
463        );
464        assert_eq!(info.imports.len(), 1);
465        assert_eq!(info.imports[0].source, "./vendor.js");
466    }
467
468    #[test]
469    fn bare_asset_with_subdir_normalized_to_relative() {
470        let info = parse_html_to_module(FileId(0), r#"<script src="assets/app.js"></script>"#, 0);
471        assert_eq!(info.imports.len(), 1);
472        assert_eq!(info.imports[0].source, "./assets/app.js");
473    }
474
475    #[test]
476    fn root_absolute_script_src_unchanged() {
477        let info = parse_html_to_module(FileId(0), r#"<script src="/src/main.ts"></script>"#, 0);
478        assert_eq!(info.imports.len(), 1);
479        assert_eq!(info.imports[0].source, "/src/main.ts");
480    }
481
482    #[test]
483    fn parent_relative_script_src_unchanged() {
484        let info = parse_html_to_module(
485            FileId(0),
486            r#"<script src="../shared/vendor.js"></script>"#,
487            0,
488        );
489        assert_eq!(info.imports.len(), 1);
490        assert_eq!(info.imports[0].source, "../shared/vendor.js");
491    }
492
493    #[test]
494    fn skips_preload_link() {
495        let info = parse_html_to_module(
496            FileId(0),
497            r#"<link rel="preload" href="./src/font.woff2" as="font" />"#,
498            0,
499        );
500        assert!(info.imports.is_empty());
501    }
502
503    #[test]
504    fn skips_icon_link() {
505        let info =
506            parse_html_to_module(FileId(0), r#"<link rel="icon" href="./favicon.ico" />"#, 0);
507        assert!(info.imports.is_empty());
508    }
509
510    #[test]
511    fn skips_remote_stylesheet() {
512        let info = parse_html_to_module(
513            FileId(0),
514            r#"<link rel="stylesheet" href="https://fonts.googleapis.com/css" />"#,
515            0,
516        );
517        assert!(info.imports.is_empty());
518    }
519
520    #[test]
521    fn skips_commented_out_script() {
522        let info = parse_html_to_module(
523            FileId(0),
524            r#"<!-- <script src="./old.js"></script> -->
525            <script src="./new.js"></script>"#,
526            0,
527        );
528        assert_eq!(info.imports.len(), 1);
529        assert_eq!(info.imports[0].source, "./new.js");
530    }
531
532    #[test]
533    fn skips_commented_out_link() {
534        let info = parse_html_to_module(
535            FileId(0),
536            r#"<!-- <link rel="stylesheet" href="./old.css" /> -->
537            <link rel="stylesheet" href="./new.css" />"#,
538            0,
539        );
540        assert_eq!(info.imports.len(), 1);
541        assert_eq!(info.imports[0].source, "./new.css");
542    }
543
544    #[test]
545    fn handles_multiline_script_tag() {
546        let info = parse_html_to_module(
547            FileId(0),
548            "<script\n  type=\"module\"\n  src=\"./src/entry.js\"\n></script>",
549            0,
550        );
551        assert_eq!(info.imports.len(), 1);
552        assert_eq!(info.imports[0].source, "./src/entry.js");
553    }
554
555    #[test]
556    fn handles_multiline_link_tag() {
557        let info = parse_html_to_module(
558            FileId(0),
559            "<link\n  rel=\"stylesheet\"\n  href=\"./src/global.css\"\n/>",
560            0,
561        );
562        assert_eq!(info.imports.len(), 1);
563        assert_eq!(info.imports[0].source, "./src/global.css");
564    }
565
566    #[test]
567    fn full_vite_html() {
568        let info = parse_html_to_module(
569            FileId(0),
570            r#"<!doctype html>
571<html>
572  <head>
573    <link rel="stylesheet" href="./src/global.css" />
574    <link rel="icon" href="/favicon.ico" />
575  </head>
576  <body>
577    <div id="app"></div>
578    <script type="module" src="./src/entry.js"></script>
579  </body>
580</html>"#,
581            0,
582        );
583        assert_eq!(info.imports.len(), 2);
584        let sources: Vec<&str> = info.imports.iter().map(|i| i.source.as_str()).collect();
585        assert!(sources.contains(&"./src/global.css"));
586        assert!(sources.contains(&"./src/entry.js"));
587    }
588
589    #[test]
590    fn empty_html() {
591        let info = parse_html_to_module(FileId(0), "", 0);
592        assert!(info.imports.is_empty());
593    }
594
595    #[test]
596    fn html_with_no_assets() {
597        let info = parse_html_to_module(
598            FileId(0),
599            r"<!doctype html><html><body><h1>Hello</h1></body></html>",
600            0,
601        );
602        assert!(info.imports.is_empty());
603    }
604
605    #[test]
606    fn single_quoted_attributes() {
607        let info = parse_html_to_module(FileId(0), r"<script src='./src/entry.js'></script>", 0);
608        assert_eq!(info.imports.len(), 1);
609        assert_eq!(info.imports[0].source, "./src/entry.js");
610    }
611
612    #[test]
613    fn all_imports_are_side_effect() {
614        let info = parse_html_to_module(
615            FileId(0),
616            r#"<script src="./entry.js"></script>
617            <link rel="stylesheet" href="./style.css" />"#,
618            0,
619        );
620        for imp in &info.imports {
621            assert!(matches!(imp.imported_name, ImportedName::SideEffect));
622            assert!(imp.local_name.is_empty());
623            assert!(!imp.is_type_only);
624        }
625    }
626
627    #[test]
628    fn suppression_comments_extracted() {
629        let info = parse_html_to_module(
630            FileId(0),
631            "<!-- fallow-ignore-file -->\n<script src=\"./entry.js\"></script>",
632            0,
633        );
634        assert_eq!(info.imports.len(), 1);
635    }
636
637    #[test]
638    fn angular_template_extracts_member_refs() {
639        let info = parse_html_to_module(
640            FileId(0),
641            "<h1>{{ title() }}</h1>\n\
642             <p [class.highlighted]=\"isHighlighted\">{{ greeting() }}</p>\n\
643             <button (click)=\"onButtonClick()\">Toggle</button>",
644            0,
645        );
646        let names: rustc_hash::FxHashSet<&str> = info
647            .member_accesses
648            .iter()
649            .filter(|a| a.object == ANGULAR_TPL_SENTINEL)
650            .map(|a| a.member.as_str())
651            .collect();
652        assert!(names.contains("title"), "should contain 'title'");
653        assert!(
654            names.contains("isHighlighted"),
655            "should contain 'isHighlighted'"
656        );
657        assert!(names.contains("greeting"), "should contain 'greeting'");
658        assert!(
659            names.contains("onButtonClick"),
660            "should contain 'onButtonClick'"
661        );
662    }
663
664    #[test]
665    fn plain_html_no_angular_refs() {
666        let info = parse_html_to_module(
667            FileId(0),
668            "<!doctype html><html><body><h1>Hello</h1></body></html>",
669            0,
670        );
671        assert!(info.member_accesses.is_empty());
672    }
673}
fallow_extract/html.rs

fallow_extract/
html.rs
