Skip to main content

provenant/utils/
language.rs

1use std::collections::HashSet;
2use std::path::Path;
3
4use file_identify::tags_from_filename;
5
6pub fn detect_language(path: &Path, content: &[u8]) -> Option<String> {
7    if let Some(language) = detect_shebang_language(content) {
8        return Some(language);
9    }
10
11    if let Some(language) = detect_file_identify_language(path) {
12        return Some(language);
13    }
14
15    if let Some(language) = detect_repo_special_file_name_language(path) {
16        return Some(language);
17    }
18
19    if let Some(language) = detect_manual_extension_language(path) {
20        return Some(language);
21    }
22
23    detect_content_hint_language(content)
24}
25
26fn detect_content_hint_language(content: &[u8]) -> Option<String> {
27    let sample_end = std::cmp::min(content.len(), 1000);
28    let text_sample = std::str::from_utf8(&content[..sample_end]).ok()?;
29
30    if text_sample.contains("<?php") {
31        Some("PHP".to_string())
32    } else if text_sample.contains("<html") || text_sample.contains("<!DOCTYPE html") {
33        Some("HTML".to_string())
34    } else if text_sample.contains("plugins {")
35        || (text_sample.contains("dependencies {") && text_sample.contains("repositories {"))
36    {
37        Some("Groovy".to_string())
38    } else if text_sample.contains("import React") || text_sample.contains("import {") {
39        Some("JavaScript/TypeScript".to_string())
40    } else if has_python_definition_line(text_sample) {
41        Some("Python".to_string())
42    } else if text_sample.contains("package ")
43        && text_sample.contains("import ")
44        && text_sample.contains('{')
45    {
46        Some("Go".to_string())
47    } else {
48        None
49    }
50}
51
52fn has_python_definition_line(text: &str) -> bool {
53    text.lines().any(|line| {
54        let trimmed = line.trim_start();
55        trimmed.starts_with("def ") && trimmed.contains(':')
56    })
57}
58
59fn detect_shebang_language(content: &[u8]) -> Option<String> {
60    if content.len() <= 2 || content[0] != b'#' || content[1] != b'!' {
61        return None;
62    }
63
64    let shebang_end = content
65        .iter()
66        .position(|&b| b == b'\n')
67        .unwrap_or(content.len());
68    let shebang = String::from_utf8_lossy(&content[0..shebang_end]).to_ascii_lowercase();
69
70    if shebang.contains("python") {
71        Some("Python".to_string())
72    } else if shebang.contains("node") || shebang.contains("deno") || shebang.contains("bun") {
73        Some("JavaScript".to_string())
74    } else if shebang.contains("bash") {
75        Some("Bash".to_string())
76    } else if shebang.contains("zsh") {
77        Some("Zsh".to_string())
78    } else if shebang.contains("fish") {
79        Some("Fish".to_string())
80    } else if shebang.contains("ksh") {
81        Some("Ksh".to_string())
82    } else if shebang.contains("ruby") {
83        Some("Ruby".to_string())
84    } else if shebang.contains("perl") {
85        Some("Perl".to_string())
86    } else if shebang.contains("php") {
87        Some("PHP".to_string())
88    } else if shebang.contains("pwsh") || shebang.contains("powershell") {
89        Some("PowerShell".to_string())
90    } else if shebang.contains("awk") {
91        Some("Awk".to_string())
92    } else if shebang.contains("/sh") {
93        Some("Shell".to_string())
94    } else {
95        None
96    }
97}
98
99fn detect_file_identify_language(path: &Path) -> Option<String> {
100    let file_name = path.file_name()?.to_str()?;
101    let tags = tags_from_filename(file_name);
102
103    map_file_identify_tags(&tags).map(str::to_string)
104}
105
106fn map_file_identify_tags(tags: &HashSet<&'static str>) -> Option<&'static str> {
107    if tags.contains("dockerfile") {
108        return Some("Dockerfile");
109    }
110    if tags.contains("makefile") {
111        return Some("Makefile");
112    }
113    if tags.contains("rust") {
114        return Some("Rust");
115    }
116    if tags.contains("python") {
117        return Some("Python");
118    }
119    if tags.contains("javascript") || tags.contains("jsx") {
120        return Some("JavaScript");
121    }
122    if tags.contains("ts") || tags.contains("tsx") {
123        return Some("TypeScript");
124    }
125    if tags.contains("html") {
126        return Some("HTML");
127    }
128    if tags.contains("css") {
129        return Some("CSS");
130    }
131    if tags.contains("c") {
132        return Some("C");
133    }
134    if tags.contains("cpp") {
135        return Some("C++");
136    }
137    if tags.contains("java") {
138        return Some("Java");
139    }
140    if tags.contains("go") {
141        return Some("Go");
142    }
143    if tags.contains("ruby") {
144        return Some("Ruby");
145    }
146    if tags.contains("php") {
147        return Some("PHP");
148    }
149    if tags.contains("perl") {
150        return Some("Perl");
151    }
152    if tags.contains("swift") {
153        return Some("Swift");
154    }
155    if tags.contains("shell") || tags.contains("bash") || tags.contains("zsh") {
156        return Some("Shell");
157    }
158    if tags.contains("kotlin") {
159        return Some("Kotlin");
160    }
161    if tags.contains("dart") {
162        return Some("Dart");
163    }
164    if tags.contains("scala") {
165        return Some("Scala");
166    }
167    if tags.contains("csharp") {
168        return Some("C#");
169    }
170    if tags.contains("fsharp") {
171        return Some("F#");
172    }
173    if tags.contains("r") {
174        return Some("R");
175    }
176    if tags.contains("lua") {
177        return Some("Lua");
178    }
179    if tags.contains("julia") {
180        return Some("Julia");
181    }
182    if tags.contains("elixir") {
183        return Some("Elixir");
184    }
185    if tags.contains("clojure") {
186        return Some("Clojure");
187    }
188    if tags.contains("haskell") {
189        return Some("Haskell");
190    }
191    if tags.contains("erlang") {
192        return Some("Erlang");
193    }
194    if tags.contains("sql") {
195        return Some("SQL");
196    }
197    if tags.contains("tex") {
198        return Some("TeX");
199    }
200    if tags.contains("groovy") || tags.contains("gradle") {
201        return Some("Groovy");
202    }
203    if tags.contains("nix") {
204        return Some("Nix");
205    }
206    if tags.contains("zig") {
207        return Some("Zig");
208    }
209    if tags.contains("powershell") {
210        return Some("PowerShell");
211    }
212    if tags.contains("starlark") {
213        return Some("Starlark");
214    }
215    if tags.contains("awk") {
216        return Some("Awk");
217    }
218    if tags.contains("ocaml") {
219        return Some("OCaml");
220    }
221    if tags.contains("meson") {
222        return Some("Meson");
223    }
224
225    None
226}
227
228fn detect_repo_special_file_name_language(path: &Path) -> Option<String> {
229    let file_name = path
230        .file_name()
231        .and_then(|n| n.to_str())
232        .map(|s| s.to_ascii_lowercase())
233        .unwrap_or_default();
234
235    if matches!(
236        file_name.as_str(),
237        "gemfile" | "rakefile" | "podfile" | "vagrantfile" | "brewfile"
238    ) {
239        Some("Ruby".to_string())
240    } else if matches!(file_name.as_str(), "apkbuild" | "pkgbuild" | "gradlew") {
241        Some("Shell".to_string())
242    } else if matches!(file_name.as_str(), "jamfile" | "jamroot") {
243        Some("Jamfile".to_string())
244    } else if matches!(file_name.as_str(), "meson.build") {
245        Some("Meson".to_string())
246    } else if matches!(file_name.as_str(), "containerfile.core") {
247        Some("Dockerfile".to_string())
248    } else if matches!(file_name.as_str(), "build" | "workspace" | "buck") {
249        Some("Starlark".to_string())
250    } else if matches!(
251        file_name.as_str(),
252        "default.nix" | "flake.nix" | "shell.nix"
253    ) {
254        Some("Nix".to_string())
255    } else {
256        None
257    }
258}
259
260fn detect_manual_extension_language(path: &Path) -> Option<String> {
261    let extension = path.extension()?.to_str()?.to_ascii_lowercase();
262
263    match extension.as_str() {
264        "rs" => Some("Rust".to_string()),
265        "py" => Some("Python".to_string()),
266        "js" | "mjs" | "cjs" => Some("JavaScript".to_string()),
267        "ts" | "tsx" | "mts" | "cts" => Some("TypeScript".to_string()),
268        "jsx" => Some("JavaScript".to_string()),
269        "html" | "htm" => Some("HTML".to_string()),
270        "css" => Some("CSS".to_string()),
271        "c" => Some("C".to_string()),
272        "cpp" | "cc" | "cxx" | "hh" | "hxx" => Some("C++".to_string()),
273        "h" => Some("C".to_string()),
274        "hpp" => Some("C++".to_string()),
275        "m" => Some("Objective-C".to_string()),
276        "mm" => Some("Objective-C++".to_string()),
277        "s" | "asm" => Some("GAS".to_string()),
278        "java" => Some("Java".to_string()),
279        "go" => Some("Go".to_string()),
280        "rb" => Some("Ruby".to_string()),
281        "php" => Some("PHP".to_string()),
282        "pl" => Some("Perl".to_string()),
283        "swift" => Some("Swift".to_string()),
284        "sql" => Some("SQL".to_string()),
285        "sh" => Some("Shell".to_string()),
286        "bash" => Some("Bash".to_string()),
287        "zsh" => Some("Zsh".to_string()),
288        "fish" => Some("Fish".to_string()),
289        "ksh" => Some("Ksh".to_string()),
290        "bat" | "cmd" => Some("Batchfile".to_string()),
291        "kt" | "kts" => Some("Kotlin".to_string()),
292        "dart" => Some("Dart".to_string()),
293        "scala" => Some("Scala".to_string()),
294        "cs" => Some("C#".to_string()),
295        "fs" | "fsx" => Some("F#".to_string()),
296        "r" => Some("R".to_string()),
297        "lua" => Some("Lua".to_string()),
298        "jl" => Some("Julia".to_string()),
299        "ex" | "exs" => Some("Elixir".to_string()),
300        "clj" | "cljs" | "cljc" => Some("Clojure".to_string()),
301        "hs" => Some("Haskell".to_string()),
302        "erl" => Some("Erlang".to_string()),
303        "tex" => Some("TeX".to_string()),
304        "groovy" | "gradle" | "gvy" | "gy" | "gsh" => Some("Groovy".to_string()),
305        "cmake" => Some("CMake".to_string()),
306        "nix" => Some("Nix".to_string()),
307        "zig" => Some("Zig".to_string()),
308        "ps1" | "psm1" | "psd1" => Some("PowerShell".to_string()),
309        "bzl" | "bazel" | "star" | "sky" => Some("Starlark".to_string()),
310        "awk" => Some("Awk".to_string()),
311        "ml" | "mli" => Some("OCaml".to_string()),
312        _ => None,
313    }
314}
315
316#[cfg(test)]
317mod tests {
318    use super::detect_language;
319    use std::path::Path;
320
321    #[test]
322    fn detect_language_supports_containerfile_names() {
323        assert_eq!(
324            detect_language(Path::new("Containerfile"), b"FROM scratch\n"),
325            Some("Dockerfile".to_string())
326        );
327        assert_eq!(
328            detect_language(Path::new("containerfile.core"), b"FROM scratch\n"),
329            Some("Dockerfile".to_string())
330        );
331    }
332
333    #[test]
334    fn detect_language_maps_c_headers_to_c() {
335        assert_eq!(
336            detect_language(Path::new("zlib.h"), b"/* header */\n"),
337            Some("C".to_string())
338        );
339    }
340
341    #[test]
342    fn detect_language_maps_uppercase_s_to_gas() {
343        assert_eq!(
344            detect_language(Path::new("gvmat64.S"), b"; asm\n"),
345            Some("GAS".to_string())
346        );
347    }
348
349    #[test]
350    fn detect_language_handles_manifest_dsl_filenames() {
351        assert_eq!(
352            detect_language(Path::new("APKBUILD"), b"pkgname=demo\n"),
353            Some("Shell".to_string())
354        );
355        assert_eq!(
356            detect_language(Path::new("Podfile"), b"source 'https://rubygems.org'\n"),
357            Some("Ruby".to_string())
358        );
359        assert_eq!(
360            detect_language(Path::new("meson.build"), b"project('demo')\n"),
361            Some("Meson".to_string())
362        );
363        assert_eq!(
364            detect_language(Path::new("BUILD"), b"cc_library(name = 'demo')\n"),
365            Some("Starlark".to_string())
366        );
367        assert_eq!(
368            detect_language(Path::new("flake.nix"), b"{ inputs, ... }: {}\n"),
369            Some("Nix".to_string())
370        );
371    }
372
373    #[test]
374    fn detect_language_handles_common_build_extensions() {
375        assert_eq!(
376            detect_language(Path::new("build.gradle"), b"plugins { id 'java' }\n"),
377            Some("Groovy".to_string())
378        );
379        assert_eq!(
380            detect_language(
381                Path::new("toolchain.cmake"),
382                b"set(CMAKE_CXX_STANDARD 20)\n"
383            ),
384            Some("CMake".to_string())
385        );
386        assert_eq!(
387            detect_language(Path::new("main.nix"), b"{ pkgs }: pkgs.hello\n"),
388            Some("Nix".to_string())
389        );
390        assert_eq!(
391            detect_language(Path::new("rules.bzl"), b"def _impl(ctx):\n    pass\n"),
392            Some("Starlark".to_string())
393        );
394        assert_eq!(
395            detect_language(Path::new("script.ps1"), b"Write-Host 'hello'\n"),
396            Some("PowerShell".to_string())
397        );
398    }
399
400    #[test]
401    fn detect_language_maps_batch_and_ipp_extensions() {
402        assert_eq!(
403            detect_language(Path::new("build.cmd"), b"@echo off\r\n"),
404            Some("Batchfile".to_string())
405        );
406        assert_eq!(
407            detect_language(
408                Path::new("from_chars.ipp"),
409                b"template <class T> void parse();\n"
410            ),
411            None
412        );
413    }
414
415    #[test]
416    fn detect_language_handles_jamfile_names() {
417        assert_eq!(
418            detect_language(Path::new("Jamfile"), b"lib boost_json ;\n"),
419            Some("Jamfile".to_string())
420        );
421    }
422
423    #[test]
424    fn detect_language_omits_generic_text_fallbacks() {
425        assert_eq!(
426            detect_language(Path::new("README.txt"), b"plain text\n"),
427            None
428        );
429        assert_eq!(
430            detect_language(Path::new("data.bin"), &[0, 159, 146, 150]),
431            None
432        );
433    }
434
435    #[test]
436    fn detect_language_ignores_yaml_as_programming_language() {
437        assert_eq!(
438            detect_language(Path::new("config.yaml"), b"key: value\n"),
439            None
440        );
441    }
442
443    #[test]
444    fn detect_language_keeps_extension_detection_for_non_utf8_python() {
445        let latin1_python = b"# coding: latin-1\nprint(\"caf\xe9\")\n# comment padding\n";
446
447        assert_eq!(
448            detect_language(Path::new("script.py"), latin1_python),
449            Some("Python".to_string())
450        );
451    }
452
453    #[test]
454    fn detect_language_uses_utf8_content_hints_for_extensionless_files() {
455        assert_eq!(
456            detect_language(
457                Path::new("index"),
458                b"<!DOCTYPE html><html><body></body></html>"
459            ),
460            Some("HTML".to_string())
461        );
462    }
463
464    #[test]
465    fn detect_language_does_not_infer_python_from_default_labels() {
466        assert_eq!(
467            detect_language(
468                Path::new("from_chars.ipp"),
469                b"switch (value) {\n  default: return parse();\n}\n"
470            ),
471            None
472        );
473    }
474
475    #[test]
476    fn detect_language_does_not_use_content_hints_for_invalid_utf8() {
477        assert_eq!(
478            detect_language(
479                Path::new("index"),
480                &[0xff, b'<', b'h', b't', b'm', b'l', b'>']
481            ),
482            None
483        );
484    }
485}