Add the full Arabic Wiktionary, and answer in Urdu before English

Classical Persian quotes Arabic outright — Hafez opens with a whole hemistich
of it — and none of it resolved before. The full Arabic export brings 36,627
entries and 819,608 new form pairs, and the forms are the point: السّاقی,
الناس, تَلْقَ and تَهْوی are all conjugated or carry the article, so they only
reach a definition through that index. Nine of the ten Arabic words in the
sample now answer where none did.

The cost is real and worth stating: the database goes from 28 MB to 87 MB and
the release APK from 10.6 MB to 29.2 MB, with another 87 MB unpacked on first
run, so about 116 MB installed.

Results are now ordered the way a reader of this app wants them: the Urdu
definition first, because it needs no translating, then Persian, then sources
keyed on another language, with English arriving only through whatever is left.
Arabic sorts last outright — its index is larger than every other source
combined, which makes it the likeliest to match by coincidence.

No Persian-to-Urdu dictionary. Wiktionary's Persian entries carry no
translations at all; the tables live only on English pages, in a 3.3 GB export,
so it would mean pivoting through an English sense. A sample of that file
projects about 6,900 Persian words with any Urdu equivalent, mostly modern
dictionary vocabulary rather than the language of the poems — not worth the
pivot. tools/README.md records why, so the question doesn't get reopened from
scratch.

Verified on an API 36 emulator: عشق returns all five sources with the Urdu
definition at the top.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

app/src/main/java/com/ganjoor/android/data/Dictionary.kt

@@ -17,7 +17,7 @@ data class Definition(val word: String, val gloss: String, val source: String)
  */
 object Dictionary {
     /** Guards against a copy interrupted half-way leaving an unopenable file behind. */
-    private const val ASSET_BYTES = 27742208L
+    private const val ASSET_BYTES = 86663168L
 
     // Not a .gz: the build packager silently gunzips those and drops the extension, which left
     // the asset under a different name than the code was opening.
@@ -83,9 +83,28 @@ object Dictionary {
             }
     }
 
+    /**
+     * Ordered the way a reader of this app wants to be answered: a definition written in Urdu
+     * first, because it needs no translating at all, then the Persian sources, then the ones
+     * keyed on another language. English is what the rest fall back to, so it comes last by
+     * coming from the sources that sit last.
+     *
+     * Arabic is last outright: its forms index is larger than every other source combined,
+     * which makes it the likeliest to match something by coincidence.
+     */
     private fun direct(database: SQLiteDatabase, word: String): List<Definition> =
         database.rawQuery(
-            "SELECT display, gloss, source FROM entry WHERE word = ? LIMIT 12",
+            """
+            SELECT display, gloss, source FROM entry WHERE word = ?
+            ORDER BY CASE source
+                WHEN 'urwiktionary' THEN 0
+                WHEN 'wiktionary-fa' THEN 1
+                WHEN 'daneshjoo' THEN 2
+                WHEN 'wiktionary-ur' THEN 3
+                ELSE 4
+            END
+            LIMIT 12
+            """,
             arrayOf(word),
         ).use { cursor ->
             buildList {

app/src/main/java/com/ganjoor/android/ui/WordSheet.kt

@@ -55,6 +55,7 @@ private fun sourceLabel(source: String) = when (source) {
     "wiktionary-fa" -> R.string.source_wiktionary
     "wiktionary-ur" -> R.string.source_wiktionary_ur
     "urwiktionary" -> R.string.source_urwiktionary
+    "wiktionary-ar" -> R.string.source_wiktionary_ar
     else -> R.string.source_daneshjoo
 }
 

app/src/main/res/values-fa/strings.xml

@@ -68,4 +68,5 @@
     <string name="source_wiktionary_ur">ویکی‌واژه — اردو به انگلیسی (CC BY-SA 3.0)</string>
     <string name="source_urwiktionary">ویکی‌واژهٔ اردو — اردو به اردو (CC BY-SA 3.0)</string>
     <string name="did_you_mean">واژه‌های نزدیک</string>
+    <string name="source_wiktionary_ar">ویکی‌واژه — عربی به انگلیسی (CC BY-SA 3.0)</string>
 </resources>

app/src/main/res/values-ur/strings.xml

@@ -68,4 +68,5 @@
     <string name="source_wiktionary_ur">ویکی لغت — اردو سے انگریزی (CC BY-SA 3.0)</string>
     <string name="source_urwiktionary">اردو ویکی لغت — اردو سے اردو (CC BY-SA 3.0)</string>
     <string name="did_you_mean">ملتے جلتے الفاظ</string>
+    <string name="source_wiktionary_ar">ویکی لغت — عربی سے انگریزی (CC BY-SA 3.0)</string>
 </resources>

app/src/main/res/values/strings.xml

@@ -73,4 +73,5 @@
     <string name="source_wiktionary_ur">Wiktionary — Urdu to English (CC BY-SA 3.0)</string>
     <string name="source_urwiktionary">Urdu Wiktionary — Urdu to Urdu (CC BY-SA 3.0)</string>
     <string name="did_you_mean">Similar words</string>
+    <string name="source_wiktionary_ar">Wiktionary — Arabic to English (CC BY-SA 3.0)</string>
 </resources>

tools/README.md

@@ -44,6 +44,25 @@ source here whose definitions are written **in Urdu** — thin (around 3,100 usa
 31,000 pages, many being stubs), but for a word it does carry an Urdu reader is better served by
 it than by a translation into English.
 
+## Arabic
+
+```sh
+curl -L -o ar.jsonl https://kaikki.org/dictionary/Arabic/kaikki.org-dictionary-Arabic.jsonl
+```
+
+521 MB, almost all of it the inflection index, and that index is the point: the Arabic quoted
+inside Persian verse is conjugated, so السّاقی, الناس, تَلْقَ and تَهْوی only reach a definition
+through it. 36,627 entries and 819,608 new form pairs for about 59 MB of database.
+
+## Why there is no Persian-to-Urdu
+
+Wiktionary's Persian entries carry no translations at all — the translation tables live only on
+English pages, in a 3.3 GB export. Going Persian to Urdu would mean pivoting through an English
+sense, and a sample of that file projects only about 6,900 Persian words with any Urdu
+equivalent, most of them modern dictionary vocabulary rather than the language of the poems. The
+definitions written in Urdu therefore come from Urdu Wiktionary directly, and are preferred over
+the English ones wherever they exist.
+
 ## Licences
 
 Each row carries its `source`, so attribution stays accurate and either source can be dropped

tools/build_dictionary.py

@@ -69,6 +69,25 @@ for line in open('ur.jsonl', encoding='utf-8'):
             forms.add((normalise(t), normalise(word)))
 print(f"wiktionary-ur: {len(entries) - n_fa} entries, {len(forms)} forms total")
 
+# Arabic, for the lines classical Persian quotes outright — Hafez opens with one.
+if os.path.exists('ar.jsonl'):
+    n_ar, f_ar = len(entries), len(forms)
+    for line in open('ar.jsonl', encoding='utf-8'):
+        try: e = json.loads(line)
+        except Exception: continue
+        word = e.get('word')
+        if not word: continue
+        gs = [g.strip() for s in e.get('senses', []) for g in (s.get('glosses') or []) if g.strip()]
+        if gs:
+            pos = e.get('pos') or ''
+            gloss = '; '.join(dict.fromkeys(gs))[:600]
+            entries.append((normalise(word), word, f"({pos}) {gloss}" if pos else gloss, 'wiktionary-ar'))
+        for f in e.get('forms', []):
+            t = f.get('form')
+            if t and t != word and not t.startswith('-') and len(t) > 1:
+                forms.add((normalise(t), normalise(word)))
+    print(f"wiktionary-ar: {len(entries) - n_ar} entries, {len(forms) - f_ar} new forms")
+
 tag = re.compile(r'<[^>]+>')
 n0 = len(entries)
 for k, v in MDX('daneshjoo.mdx').items():