anil.recoil.org
Detect which human language a document uses from OCaml, from the Nu Html validator
languages
unicode
ocaml
wasm+cmd
+5
bin/dune
+5
bin/dune
+65
bin/langdetect_cli.ml
+65
bin/langdetect_cli.ml
···
1
+
(*---------------------------------------------------------------------------
2
+
Copyright (c) 2025 Anil Madhavapeddy <anil@recoil.org>
3
+
SPDX-License-Identifier: MIT
4
+
---------------------------------------------------------------------------*)
5
+
6
+
let detect_language input_text =
7
+
let detector = Langdetect.create_default () in
8
+
let results = Langdetect.detect detector input_text in
9
+
List.iter
10
+
(fun (r : Langdetect.result) -> Printf.printf "%s %.4f\n" r.lang r.prob)
11
+
results
12
+
13
+
let read_all_stdin () =
14
+
let buf = Buffer.create 4096 in
15
+
try
16
+
while true do
17
+
Buffer.add_channel buf stdin 4096
18
+
done;
19
+
Buffer.contents buf
20
+
with End_of_file -> Buffer.contents buf
21
+
22
+
let read_file path =
23
+
let ic = open_in path in
24
+
let n = in_channel_length ic in
25
+
let s = really_input_string ic n in
26
+
close_in ic;
27
+
s
28
+
29
+
let run file_opt =
30
+
let text =
31
+
match file_opt with
32
+
| Some path -> read_file path
33
+
| None -> read_all_stdin ()
34
+
in
35
+
if String.length (String.trim text) = 0 then
36
+
`Error (false, "No input text provided")
37
+
else begin
38
+
detect_language text;
39
+
`Ok ()
40
+
end
41
+
42
+
open Cmdliner
43
+
44
+
let file_arg =
45
+
let doc = "Input file to detect language from. If not provided, reads from stdin." in
46
+
Arg.(value & pos 0 (some file) None & info [] ~docv:"FILE" ~doc)
47
+
48
+
let cmd =
49
+
let doc = "Detect the language of text" in
50
+
let man =
51
+
[
52
+
`S Manpage.s_description;
53
+
`P "Detects the natural language of input text using n-gram frequency analysis.";
54
+
`P "Outputs detected language codes and their probabilities as space-separated values, one per line, sorted by probability (highest first).";
55
+
`S Manpage.s_examples;
56
+
`P "Detect language from a file:";
57
+
`Pre " langdetect document.txt";
58
+
`P "Detect language from stdin:";
59
+
`Pre " echo 'Hello world' | langdetect";
60
+
]
61
+
in
62
+
let info = Cmd.info "langdetect" ~version:"%%VERSION%%" ~doc ~man in
63
+
Cmd.v info Term.(ret (const run $ file_arg))
64
+
65
+
let () = exit (Cmd.eval cmd)
+3
-2
dune-project
+3
-2
dune-project
···
21
21
(depends
22
22
(ocaml (>= 5.1.0))
23
23
(uutf (>= 1.0.0))
24
+
(cmdliner (>= 1.2.0))
24
25
(odoc :with-doc)
25
26
(alcotest (and :with-test (>= 1.7.0)))))
26
27
···
35
36
(ocaml (>= 5.1.0))
36
37
(langdetect (= :version))
37
38
(brr (>= 0.0.6))
38
-
(js_of_ocaml (>= 5.0.0))
39
-
(js_of_ocaml-compiler (>= 5.0.0))))
39
+
(js_of_ocaml (>= 6.0.0))
40
+
(js_of_ocaml-compiler (>= 6.0.0))))
+2
-2
langdetect-js.opam
+2
-2
langdetect-js.opam
+1
langdetect.opam
+1
langdetect.opam
+17
lib/js/dune
+17
lib/js/dune
···
48
48
(targets langdetect-tests.wasm.js)
49
49
(deps langdetect_js_tests.bc.wasm.js)
50
50
(action (copy %{deps} %{targets})))
51
+
52
+
; Install web assets to share/langdetect-js/
53
+
; Includes HTML demo, JS files, WASM loaders, and WASM assets with source maps
54
+
(install
55
+
(package langdetect-js)
56
+
(section share)
57
+
(files
58
+
(langdetect.html as langdetect-js/langdetect.html)
59
+
; JS files (work standalone in browsers)
60
+
(langdetect.js as langdetect-js/langdetect.js)
61
+
(langdetect-tests.js as langdetect-js/langdetect-tests.js)
62
+
; WASM loaders (in same dir so relative asset paths work)
63
+
(langdetect_js_main.bc.wasm.js as langdetect-js/langdetect_js_main.bc.wasm.js)
64
+
(langdetect_js_tests.bc.wasm.js as langdetect-js/langdetect_js_tests.bc.wasm.js)
65
+
; WASM assets - must be in langdetect-js/ so relative paths from loaders work
66
+
(glob_files_rec (langdetect_js_main.bc.wasm.assets/* with_prefix langdetect-js/langdetect_js_main.bc.wasm.assets))
67
+
(glob_files_rec (langdetect_js_tests.bc.wasm.assets/* with_prefix langdetect-js/langdetect_js_tests.bc.wasm.assets))))
+368
lib/js/langdetect.html
+368
lib/js/langdetect.html
···
1
+
<!DOCTYPE html>
2
+
<html lang="en">
3
+
<head>
4
+
<meta charset="UTF-8">
5
+
<meta name="viewport" content="width=device-width, initial-scale=1.0">
6
+
<title>Langdetect - Language Detection Demo</title>
7
+
<style>
8
+
* {
9
+
box-sizing: border-box;
10
+
}
11
+
body {
12
+
font-family: system-ui, -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
13
+
line-height: 1.6;
14
+
max-width: 1000px;
15
+
margin: 0 auto;
16
+
padding: 2rem;
17
+
background: #f5f5f5;
18
+
color: #333;
19
+
}
20
+
h1 {
21
+
color: #2563eb;
22
+
border-bottom: 3px solid #3b82f6;
23
+
padding-bottom: 0.5rem;
24
+
margin-bottom: 0.5rem;
25
+
}
26
+
.subtitle {
27
+
color: #666;
28
+
margin-top: 0;
29
+
margin-bottom: 2rem;
30
+
}
31
+
.section {
32
+
background: white;
33
+
border: 1px solid #e0e0e0;
34
+
border-radius: 12px;
35
+
padding: 1.5rem;
36
+
margin: 1.5rem 0;
37
+
box-shadow: 0 2px 4px rgba(0,0,0,0.05);
38
+
}
39
+
.section h2 {
40
+
margin-top: 0;
41
+
color: #1e40af;
42
+
}
43
+
.demo-area {
44
+
display: flex;
45
+
flex-direction: column;
46
+
gap: 1rem;
47
+
}
48
+
textarea {
49
+
width: 100%;
50
+
padding: 1rem;
51
+
font-size: 1rem;
52
+
font-family: inherit;
53
+
border: 2px solid #e0e0e0;
54
+
border-radius: 8px;
55
+
resize: vertical;
56
+
transition: border-color 0.2s;
57
+
}
58
+
textarea:focus {
59
+
outline: none;
60
+
border-color: #3b82f6;
61
+
}
62
+
button {
63
+
padding: 0.75rem 2rem;
64
+
font-size: 1rem;
65
+
font-weight: 600;
66
+
cursor: pointer;
67
+
border: none;
68
+
border-radius: 8px;
69
+
background: #2563eb;
70
+
color: white;
71
+
transition: all 0.2s;
72
+
align-self: flex-start;
73
+
}
74
+
button:hover {
75
+
background: #1d4ed8;
76
+
transform: translateY(-1px);
77
+
}
78
+
button:active {
79
+
transform: translateY(0);
80
+
}
81
+
#demo-result {
82
+
padding: 1rem;
83
+
background: #f8fafc;
84
+
border-radius: 8px;
85
+
border: 1px solid #e2e8f0;
86
+
}
87
+
#demo-result ul {
88
+
margin: 0.5rem 0;
89
+
padding-left: 1.5rem;
90
+
}
91
+
#demo-result li {
92
+
margin: 0.25rem 0;
93
+
}
94
+
.summary {
95
+
background: #eff6ff;
96
+
padding: 1rem 1.5rem;
97
+
border-radius: 8px;
98
+
margin-bottom: 1rem;
99
+
}
100
+
.summary h2 {
101
+
margin: 0 0 0.5rem 0;
102
+
}
103
+
.summary p {
104
+
margin: 0.25rem 0;
105
+
}
106
+
.results-table {
107
+
width: 100%;
108
+
border-collapse: collapse;
109
+
font-size: 0.9rem;
110
+
}
111
+
.results-table th,
112
+
.results-table td {
113
+
padding: 0.75rem 1rem;
114
+
text-align: left;
115
+
border-bottom: 1px solid #e0e0e0;
116
+
}
117
+
.results-table th {
118
+
background: #f8fafc;
119
+
font-weight: 600;
120
+
color: #475569;
121
+
}
122
+
.results-table tr:hover {
123
+
background: #f8fafc;
124
+
}
125
+
.results-table .code {
126
+
font-family: 'SF Mono', Monaco, 'Cascadia Code', monospace;
127
+
font-size: 0.85rem;
128
+
background: #f1f5f9;
129
+
padding: 0.2rem 0.4rem;
130
+
border-radius: 4px;
131
+
}
132
+
.results-table .corpus-text {
133
+
font-size: 0.8rem;
134
+
max-width: 300px;
135
+
overflow: hidden;
136
+
text-overflow: ellipsis;
137
+
white-space: nowrap;
138
+
color: #666;
139
+
}
140
+
.results-table .pass {
141
+
color: #16a34a;
142
+
font-weight: bold;
143
+
font-size: 1.1rem;
144
+
}
145
+
.results-table .fail {
146
+
color: #dc2626;
147
+
font-weight: bold;
148
+
font-size: 1.1rem;
149
+
}
150
+
.loading {
151
+
text-align: center;
152
+
padding: 2rem;
153
+
color: #666;
154
+
}
155
+
.api-docs {
156
+
background: #1e293b;
157
+
color: #e2e8f0;
158
+
padding: 1rem;
159
+
border-radius: 8px;
160
+
overflow-x: auto;
161
+
}
162
+
.api-docs code {
163
+
color: #7dd3fc;
164
+
}
165
+
.api-docs .comment {
166
+
color: #94a3b8;
167
+
}
168
+
.sample-texts {
169
+
display: grid;
170
+
grid-template-columns: repeat(auto-fill, minmax(200px, 1fr));
171
+
gap: 0.5rem;
172
+
margin-top: 1rem;
173
+
}
174
+
.sample-text {
175
+
padding: 0.5rem 1rem;
176
+
font-size: 0.85rem;
177
+
background: #f1f5f9;
178
+
border: 1px solid #e2e8f0;
179
+
border-radius: 6px;
180
+
cursor: pointer;
181
+
transition: all 0.2s;
182
+
}
183
+
.sample-text:hover {
184
+
background: #e2e8f0;
185
+
border-color: #cbd5e1;
186
+
}
187
+
.sample-text .lang {
188
+
font-weight: 600;
189
+
color: #1e40af;
190
+
}
191
+
</style>
192
+
</head>
193
+
<body>
194
+
195
+
<h1>🌍 Langdetect</h1>
196
+
<p class="subtitle">Language detection for the browser using n-gram frequency analysis</p>
197
+
198
+
<!-- Mode Selector -->
199
+
<div class="section" style="padding: 1rem 1.5rem; display: flex; align-items: center; gap: 1rem; flex-wrap: wrap;">
200
+
<label for="mode-select" style="font-weight: 600;">Runtime:</label>
201
+
<select id="mode-select" style="padding: 0.5rem 1rem; border-radius: 6px; border: 2px solid #e0e0e0; font-size: 1rem;">
202
+
<option value="js">JavaScript (js_of_ocaml)</option>
203
+
<option value="wasm">WebAssembly (wasm_of_ocaml)</option>
204
+
</select>
205
+
<button onclick="reloadWithMode()" style="padding: 0.5rem 1.5rem;">Reload</button>
206
+
<span id="mode-status" style="color: #666; font-size: 0.9rem;"></span>
207
+
</div>
208
+
209
+
<!-- Interactive Demo (created by OCaml code, but we add sample texts) -->
210
+
<div class="section">
211
+
<div id="demo">
212
+
<!-- Demo UI created by langdetect_js_tests.ml -->
213
+
<p class="loading">Loading demo...</p>
214
+
</div>
215
+
216
+
<p style="margin-top: 1rem; margin-bottom: 0.5rem; color: #666; font-size: 0.9rem;">
217
+
Click a sample to try:
218
+
</p>
219
+
<div class="sample-texts" id="sample-texts">
220
+
<div class="sample-text" data-text="The quick brown fox jumps over the lazy dog.">
221
+
<span class="lang">🇬🇧 English</span>
222
+
</div>
223
+
<div class="sample-text" data-text="Bonjour le monde! Comment allez-vous aujourd'hui?">
224
+
<span class="lang">🇫🇷 French</span>
225
+
</div>
226
+
<div class="sample-text" data-text="Guten Tag! Wie geht es Ihnen heute?">
227
+
<span class="lang">🇩🇪 German</span>
228
+
</div>
229
+
<div class="sample-text" data-text="¡Hola mundo! ¿Cómo estás hoy?">
230
+
<span class="lang">🇪🇸 Spanish</span>
231
+
</div>
232
+
<div class="sample-text" data-text="你好世界！今天你好吗？">
233
+
<span class="lang">🇨🇳 Chinese</span>
234
+
</div>
235
+
<div class="sample-text" data-text="こんにちは世界！今日はお元気ですか？">
236
+
<span class="lang">🇯🇵 Japanese</span>
237
+
</div>
238
+
<div class="sample-text" data-text="مرحبا بالعالم! كيف حالك اليوم؟">
239
+
<span class="lang">🇸🇦 Arabic</span>
240
+
</div>
241
+
<div class="sample-text" data-text="Привет мир! Как дела сегодня?">
242
+
<span class="lang">🇷🇺 Russian</span>
243
+
</div>
244
+
</div>
245
+
</div>
246
+
247
+
<!-- Test Results -->
248
+
<div class="section">
249
+
<div id="test-results">
250
+
<p class="loading">Loading tests...</p>
251
+
</div>
252
+
</div>
253
+
254
+
<!-- API Documentation -->
255
+
<div class="section">
256
+
<h2>JavaScript API</h2>
257
+
<div class="api-docs">
258
+
<pre><span class="comment">// Detect the most likely language</span>
259
+
<code>langdetect.detect</code>("Hello, world!")
260
+
<span class="comment">// → "en"</span>
261
+
262
+
<span class="comment">// Get detection with confidence score</span>
263
+
<code>langdetect.detectWithProb</code>("Bonjour le monde!")
264
+
<span class="comment">// → { lang: "fr", prob: 0.9999 }</span>
265
+
266
+
<span class="comment">// Get all matching languages</span>
267
+
<code>langdetect.detectAll</code>("Hello world")
268
+
<span class="comment">// → [{ lang: "en", prob: 0.85 }, { lang: "de", prob: 0.10 }, ...]</span>
269
+
270
+
<span class="comment">// List supported languages</span>
271
+
<code>langdetect.languages</code>()
272
+
<span class="comment">// → ["ar", "bg", "bn", "ca", "cs", "da", "de", "el", ...]</span></pre>
273
+
</div>
274
+
</div>
275
+
276
+
<!-- Supported Languages -->
277
+
<div class="section">
278
+
<h2>Supported Languages (47)</h2>
279
+
<p>
280
+
Arabic, Bengali, Bulgarian, Catalan, Chinese (Simplified), Chinese (Traditional),
281
+
Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek,
282
+
Gujarati, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Latvian,
283
+
Lithuanian, Macedonian, Malayalam, Norwegian, Panjabi, Persian, Polish, Portuguese,
284
+
Romanian, Russian, Sinhala, Slovak, Spanish, Swedish, Tagalog, Tamil, Telugu, Thai,
285
+
Turkish, Ukrainian, Urdu, Vietnamese
286
+
</p>
287
+
</div>
288
+
289
+
<script>
290
+
// Get mode from URL param or localStorage
291
+
function getMode() {
292
+
const params = new URLSearchParams(window.location.search);
293
+
return params.get('mode') || localStorage.getItem('langdetect-mode') || 'js';
294
+
}
295
+
296
+
// Set mode and reload
297
+
function reloadWithMode() {
298
+
const mode = document.getElementById('mode-select').value;
299
+
localStorage.setItem('langdetect-mode', mode);
300
+
const url = new URL(window.location);
301
+
url.searchParams.set('mode', mode);
302
+
window.location.href = url.toString();
303
+
}
304
+
305
+
// Initialize mode selector
306
+
const currentMode = getMode();
307
+
document.getElementById('mode-select').value = currentMode;
308
+
309
+
// Load the appropriate script (relative paths for installed version)
310
+
const scriptName = currentMode === 'wasm' ? 'langdetect_js_tests.bc.wasm.js' : 'langdetect-tests.js';
311
+
document.getElementById('mode-status').textContent = `Loading ${currentMode.toUpperCase()}...`;
312
+
313
+
const script = document.createElement('script');
314
+
script.src = scriptName;
315
+
script.onload = function() {
316
+
document.getElementById('mode-status').textContent = `Loaded: ${currentMode.toUpperCase()}`;
317
+
document.getElementById('mode-status').style.color = '#16a34a';
318
+
setTimeout(setupSampleTexts, 100);
319
+
};
320
+
script.onerror = function() {
321
+
document.getElementById('mode-status').textContent = `Failed to load ${scriptName}`;
322
+
document.getElementById('mode-status').style.color = '#dc2626';
323
+
showLoadError();
324
+
};
325
+
document.head.appendChild(script);
326
+
327
+
// Wait for langdetect to be ready, then set up sample text handlers
328
+
function setupSampleTexts() {
329
+
document.querySelectorAll('.sample-text').forEach(el => {
330
+
el.addEventListener('click', () => {
331
+
const text = el.getAttribute('data-text');
332
+
// Find the OCaml-created input and button
333
+
const input = document.getElementById('demo-input');
334
+
const button = document.getElementById('demo-button');
335
+
if (input && button) {
336
+
input.value = text;
337
+
button.click();
338
+
}
339
+
});
340
+
});
341
+
}
342
+
343
+
function showLoadError() {
344
+
const mode = getMode();
345
+
const scriptName = mode === 'wasm' ? 'langdetect_js_tests.bc.wasm.js' : 'langdetect-tests.js';
346
+
347
+
const demo = document.getElementById('demo');
348
+
if (demo) {
349
+
demo.innerHTML =
350
+
`<p style="color: #dc2626;"><strong>Library not loaded</strong><br>` +
351
+
`Could not load <code>${scriptName}</code>.</p>`;
352
+
}
353
+
document.getElementById('test-results').innerHTML =
354
+
`<p style="color: #dc2626;"><strong>Tests cannot run:</strong> ${scriptName} not found.</p>`;
355
+
}
356
+
357
+
// Check if library loaded after timeout
358
+
window.addEventListener('load', () => {
359
+
setTimeout(() => {
360
+
if (typeof langdetect === 'undefined') {
361
+
showLoadError();
362
+
}
363
+
}, 500);
364
+
});
365
+
</script>
366
+
367
+
</body>
368
+
</html>