fallow_extract/
lib.rs

1//! Parsing and extraction engine for fallow codebase intelligence.
2//!
3//! This crate handles all file parsing: JS/TS via Oxc, Vue/Svelte SFC extraction,
4//! Astro frontmatter, MDX import/export extraction, CSS Module class name extraction,
5//! HTML asset reference extraction, and incremental caching of parse results.
6
7#![warn(missing_docs)]
8#![cfg_attr(not(test), deny(clippy::disallowed_methods))]
9#![cfg_attr(
10    test,
11    allow(
12        clippy::unwrap_used,
13        clippy::expect_used,
14        reason = "tests use unwrap and expect to keep fixture setup concise"
15    )
16)]
17
18mod asset_url;
19pub mod astro;
20pub mod cache;
21pub(crate) mod complexity;
22pub mod css;
23pub mod css_classes;
24pub mod css_in_js;
25pub mod css_metrics;
26pub mod flags;
27pub mod glimmer;
28pub(crate) mod graphql;
29pub(crate) mod html;
30pub(crate) mod iconify;
31pub mod inventory;
32pub mod mdx;
33mod module_info;
34mod parse;
35pub mod sfc;
36pub mod sfc_css;
37mod sfc_props;
38mod sfc_template;
39mod source_map;
40pub mod suppress;
41/// Tailwind CSS arbitrary-value detection.
42pub mod tailwind;
43pub(crate) mod template_complexity;
44mod template_usage;
45/// Visitor utilities for AST extraction.
46pub mod visitor;
47
48use std::path::Path;
49
50use rayon::prelude::*;
51
52use cache::CacheStore;
53use fallow_types::discover::{DiscoveredFile, FileId};
54
55pub use fallow_types::extract::{
56    AngularComponentFieldArrayTypeFact, AngularTemplateMemberAccessFact, AngularThisSpreadFact,
57    ClassHeritageInfo, ClassThisMemberAccessFact, ClassThisWholeObjectUseFact,
58    ComputedEnumKeyUseFact, DynamicCustomElementRenderFact, DynamicImportInfo,
59    DynamicImportPattern, ExportInfo, ExportName, FactoryCallMemberAccessFact,
60    FactoryFnMemberAccessFact, FactoryFnWholeObjectFact, FactoryReturnExport,
61    FactoryReturnObjectPropertyAccessFact, FactoryReturnObjectShapeExport,
62    FluentChainMemberAccessFact, FluentChainNewMemberAccessFact, ImportInfo, ImportedName,
63    InstanceExportBindingFact, LocalTypeDeclaration, MemberAccess, MemberInfo, MemberKind,
64    ModuleInfo, ModuleLoadMechanism, ParseResult, PlaywrightFixtureAliasFact,
65    PlaywrightFixtureDefinitionFact, PlaywrightFixtureTypeFact, PlaywrightFixtureUseFact,
66    PublicSignatureTypeReference, ReExportInfo, RequireCallInfo, RequiredTypeMemberFact,
67    SemanticFact, SourceReadFailure, StringEnumMemberValueFact, TypeAliasSurfaceTargetFact,
68    TypeMemberTypeEntry, TypedPropertyMemberAccessFact, VisibilityTag, VitestModuleMockAction,
69    VitestModuleMockOperationFact, compute_line_offsets,
70};
71
72pub use astro::{
73    extract_astro_frontmatter, extract_astro_style_regions, extract_astro_template_regions,
74};
75pub use css::{
76    ThemeScan, ThemeTokenDef, extract_apply_tokens, extract_apply_tokens_located,
77    extract_css_module_exports, extract_css_var_reads_located, scan_theme_blocks,
78};
79pub use css_classes::{
80    MarkupClassScan, MarkupClassToken, is_edit_distance_one, is_typo_edit, scan_markup_class_tokens,
81};
82pub use css_in_js::{
83    ConsumerQuery, CssInJsObjectSheets, CssInJsToken, CssInJsTokenDef, CssInJsTokenOrigin,
84    TokenConsumerHit, css_in_js_consumer_scan, css_in_js_object_sheets, css_in_js_theme_consumers,
85    css_in_js_theme_token_defs, css_in_js_token_consumers, css_in_js_token_defs,
86    css_in_js_virtual_stylesheet, panda_style_value_consumers, panda_token_call_consumers,
87};
88pub use css_metrics::{compute_css_analytics, parse_css_color_rgb};
89pub use glimmer::{is_glimmer_file, strip_glimmer_templates};
90pub use mdx::extract_mdx_statements;
91pub use sfc::{
92    SourceRegion, extract_sfc_scripts, extract_sfc_styles, extract_sfc_template_regions,
93    is_sfc_file,
94};
95pub use sfc_css::{
96    scoped_unused_classes, sfc_preprocessor_virtual_stylesheet, sfc_virtual_stylesheet,
97};
98pub use tailwind::{TailwindArbitraryUse, scan_tailwind_arbitrary_values};
99
100#[expect(
101    clippy::expect_used,
102    reason = "static regex patterns are hard-coded analyzer invariants covered by extraction tests"
103)]
104fn static_regex(pattern: &str) -> regex::Regex {
105    regex::Regex::new(pattern).expect("static regex pattern should compile")
106}
107
108pub use parse::parse_source_to_module;
109
110/// Leading UTF-8 byte order mark codepoint.
111///
112/// Windows editors (Notepad, older VS settings, some IDE plugins) emit a UTF-8
113/// BOM at the start of source files. fallow's contract is "UTF-8 with or
114/// without BOM; line offsets are computed against the post-BOM view; the BOM,
115/// if present on input, is preserved on output by `fallow fix`."
116const BOM_CHAR: char = '\u{FEFF}';
117// Small, cache-hot inputs are faster on one thread than through Rayon setup.
118// Larger file sets still use parallel parsing where parse work dominates.
119const PARALLEL_PARSE_FILE_THRESHOLD: usize = 32;
120
121/// Strip the leading UTF-8 BOM if present.
122///
123/// Called at every file-read entry point in this crate so the rest of the
124/// pipeline (content hash, `compute_line_offsets`, oxc parser, downstream
125/// analyses) sees a consistent post-BOM view. Mirrors the
126/// `fallow_config` layer (`config_writer.rs::BOM`) so config-shaped sources
127/// and source-code-shaped sources are processed symmetrically. See issue #475.
128#[must_use]
129fn strip_bom(source: &str) -> &str {
130    source.strip_prefix(BOM_CHAR).unwrap_or(source)
131}
132
133/// Parse all files, extracting imports and exports.
134///
135/// Small file sets use a sequential fast path to avoid parallel scheduling
136/// overhead; larger file sets use parallel extraction.
137/// Uses the cache to skip reparsing files whose content hasn't changed.
138///
139/// When `need_complexity` is true, per-function cyclomatic/cognitive complexity
140/// metrics are computed during parsing (needed by the `health` command).
141/// Pass `false` for dead-code analysis where complexity data is unused.
142pub fn parse_all_files(
143    files: &[DiscoveredFile],
144    cache: Option<&CacheStore>,
145    need_complexity: bool,
146) -> ParseResult {
147    let results: Vec<ParseFileResult> = if files.len() <= PARALLEL_PARSE_FILE_THRESHOLD {
148        files
149            .iter()
150            .map(|file| parse_single_file_cached(file, cache, need_complexity))
151            .collect()
152    } else {
153        files
154            .par_iter()
155            .map(|file| parse_single_file_cached(file, cache, need_complexity))
156            .collect()
157    };
158
159    let mut modules = Vec::with_capacity(results.len());
160    let mut read_failures = Vec::new();
161    let mut hits = 0usize;
162    let mut misses = 0usize;
163    let mut parse_cpu_nanos = 0u64;
164
165    for result in results {
166        hits += result.cache_hits;
167        misses += result.cache_misses;
168        parse_cpu_nanos = parse_cpu_nanos.saturating_add(result.parse_cpu_nanos);
169        if let Some(module) = result.module {
170            modules.push(module);
171        }
172        if let Some(failure) = result.read_failure {
173            read_failures.push(failure);
174        }
175    }
176
177    if hits > 0 || misses > 0 {
178        tracing::info!(
179            cache_hits = hits,
180            cache_misses = misses,
181            "incremental cache stats"
182        );
183    }
184
185    ParseResult {
186        modules,
187        read_failures,
188        cache_hits: hits,
189        cache_misses: misses,
190        parse_cpu_ms: parse_cpu_nanos as f64 / 1_000_000.0,
191    }
192}
193
194struct ParseFileResult {
195    module: Option<ModuleInfo>,
196    read_failure: Option<SourceReadFailure>,
197    cache_hits: usize,
198    cache_misses: usize,
199    parse_cpu_nanos: u64,
200}
201
202impl ParseFileResult {
203    fn cache_hit(module: ModuleInfo) -> Self {
204        Self {
205            module: Some(module),
206            read_failure: None,
207            cache_hits: 1,
208            cache_misses: 0,
209            parse_cpu_nanos: 0,
210        }
211    }
212
213    fn cache_miss(module: ModuleInfo, parse_cpu_nanos: u64) -> Self {
214        Self {
215            module: Some(module),
216            read_failure: None,
217            cache_hits: 0,
218            cache_misses: 1,
219            parse_cpu_nanos,
220        }
221    }
222
223    fn read_failure(file: &DiscoveredFile, error: &std::io::Error) -> Self {
224        Self {
225            module: None,
226            read_failure: Some(SourceReadFailure {
227                file_id: file.id,
228                path: file.path.clone(),
229                error: error.to_string(),
230            }),
231            cache_hits: 0,
232            cache_misses: 0,
233            parse_cpu_nanos: 0,
234        }
235    }
236}
237
238/// Parse a single file, consulting the cache first.
239///
240/// Cache validation strategy (fast path -> slow path):
241/// 1. Open the file so unreadable sources cannot use stale cached analysis
242/// 2. Read mtime + size from the open handle
243/// 3. If mtime+size match the cached entry -> cache hit, return immediately
244/// 4. If mtime+size differ -> read file, compute content hash
245/// 5. If content hash matches cached entry -> cache hit (file was `touch`ed but unchanged)
246/// 6. Otherwise -> cache miss, full parse
247fn parse_single_file_cached(
248    file: &DiscoveredFile,
249    cache: Option<&CacheStore>,
250    need_complexity: bool,
251) -> ParseFileResult {
252    let cached_by_path = cache.and_then(|store| store.get_by_path_only(&file.path));
253
254    if let Some(cached) = cached_by_path
255        && cached.file_size == file.size_bytes
256    {
257        let source_file = match std::fs::File::open(&file.path) {
258            Ok(source_file) => source_file,
259            Err(error) => return ParseFileResult::read_failure(file, &error),
260        };
261        if let Ok(metadata) = source_file.metadata()
262            && metadata.len() == cached.file_size
263        {
264            let fingerprint =
265                fallow_types::source_fingerprint::SourceFingerprint::from_metadata(&metadata);
266            if cached.source_fingerprint() == fingerprint
267                && fingerprint.has_known_mtime()
268                && (!need_complexity || !cached.complexity.is_empty())
269            {
270                return ParseFileResult::cache_hit(cache::cached_to_module_opts(
271                    cached,
272                    file.id,
273                    need_complexity,
274                ));
275            }
276        }
277    }
278
279    let raw = match std::fs::read_to_string(&file.path) {
280        Ok(raw) => raw,
281        Err(error) => return ParseFileResult::read_failure(file, &error),
282    };
283    let source = strip_bom(&raw);
284    let content_hash = xxhash_rust::xxh3::xxh3_64(source.as_bytes());
285
286    if let Some(cached) = cached_by_path
287        && cached.content_hash == content_hash
288        && (!need_complexity || !cached.complexity.is_empty())
289    {
290        return ParseFileResult::cache_hit(cache::cached_to_module_opts(
291            cached,
292            file.id,
293            need_complexity,
294        ));
295    }
296
297    let parse_start = std::time::Instant::now();
298    let module = parse_source_to_module(file.id, &file.path, source, content_hash, need_complexity);
299    let parse_cpu_nanos = u64::try_from(parse_start.elapsed().as_nanos()).unwrap_or(u64::MAX);
300    ParseFileResult::cache_miss(module, parse_cpu_nanos)
301}
302
303/// Parse a single file and extract module information (without complexity).
304#[must_use]
305pub fn parse_single_file(file: &DiscoveredFile) -> Option<ModuleInfo> {
306    let raw = std::fs::read_to_string(&file.path).ok()?;
307    let source = strip_bom(&raw);
308    let content_hash = xxhash_rust::xxh3::xxh3_64(source.as_bytes());
309    Some(parse_source_to_module(
310        file.id,
311        &file.path,
312        source,
313        content_hash,
314        false,
315    ))
316}
317
318/// Parse from in-memory content (for LSP, includes complexity).
319#[must_use]
320pub fn parse_from_content(file_id: FileId, path: &Path, content: &str) -> ModuleInfo {
321    let content = strip_bom(content);
322    let content_hash = xxhash_rust::xxh3::xxh3_64(content.as_bytes());
323    parse_source_to_module(file_id, path, content, content_hash, true)
324}
325
326#[cfg(all(test, not(miri)))]
327mod tests;
fallow_extract/lib.rs

fallow_extract/
lib.rs
