From a1a1c45b5a0784626c0c0a00b6b0ff55e4fd2447 Mon Sep 17 00:00:00 2001
From: Levi Gruspe <mail.levig@gmail.com>
Date: Fri, 21 Apr 2023 17:14:36 +0800
Subject: [PATCH] Label inventories using Glottocodes instead of ISO639-3

---
 simphones/inventories.py | 24 +++++++++++++++++-------
 test/test_inventories.py | 26 +++++++++++++++++++++++---
 2 files changed, 40 insertions(+), 10 deletions(-)

diff --git a/simphones/inventories.py b/simphones/inventories.py
index 2e9fcb1..40880b2 100644
--- a/simphones/inventories.py
+++ b/simphones/inventories.py
@@ -39,8 +39,12 @@ def substitute(phone: Phone) -> Phone:
 def get_phonological_inventories() -> InventoryDataset:
     """Get phonological inventories from the PHOIBLE dataset.
 
-    The keys of the returned dictionary are ISO 639-3 language codes or "*".
-    The "*" inventory combines the inventories of every language.
+    The result is a dictionary with Glottocodes as keys.
+    In addition, there are three special keys:
+
+    - "*" (combination of the inventories of every language)
+    - "Djindewal" (doesn't have a Glottocode)
+    - "ModernAramaic" (doesn't have a Glottocode)
     """
     phoible = Path(__file__).with_name("phoible.csv")
 
@@ -50,7 +54,7 @@ def get_phonological_inventories() -> InventoryDataset:
         next(rows)  # Drop the header.
 
         for row in rows:
-            code = row[2]
+            glottocode = row[1]
             raw_phoneme = substitute(row[6])
             allophones = parse_allophones(row[7])
 
@@ -65,8 +69,12 @@ def get_phonological_inventories() -> InventoryDataset:
             update_inventory(combined_inventory, phoneme, allophones)
 
             # Update language inventory.
-            # NA inventories are not skipped, but they are mushed together into
-            # one.
+            # If the language has no Glottocode, use the language name as a key
+            # instead.
+            code = glottocode
+            if code == "NA":
+                language_name = row[3]
+                code = language_name.replace(" ", "")
             language_inventory = inventories.setdefault(code, {})
             update_inventory(language_inventory, phoneme, allophones)
     return inventories
@@ -117,8 +125,10 @@ def update_inventory(
 def get_sounds(language: LanguageCode = "*") -> set[Phone]:
     """Return set of sounds in the given language.
 
-    If no ISO639-3 language code is given, the return value is the combined
-    inventories of all languages in the PHOIBLE data.
+    If no Glottocode is given, the return value is the combined inventories of
+    all languages in the PHOIBLE data.
+
+    Simply returns an empty set if the language code is not in PHOIBLE.
     """
     inventories = get_phonological_inventories()
     inventory = inventories.get(language, {})
diff --git a/test/test_inventories.py b/test/test_inventories.py
index 46e6cfa..506f473 100644
--- a/test/test_inventories.py
+++ b/test/test_inventories.py
@@ -3,6 +3,7 @@
 # See https://www.gnu.org/licenses/gpl-3.0.en.html
 """Test simphones.inventories."""
 from argparse import Namespace
+import re
 
 import pytest
 
@@ -32,9 +33,28 @@ def test_get_phonological_inventories() -> None:
 
 
 def test_get_phonological_inventories_na() -> None:
-    """`get_phonological_inventories` should include an NA inventory."""
+    """`get_phonological_inventories` shouldn't contain an NA inventory.
+
+    If the Glottocode of a language isn't available, it should use the name of
+    the language as key instead.
+    """
+    inventories = get_phonological_inventories()
+    assert "NA" not in inventories
+
+    assert "Djindewal" in inventories
+    assert "ModernAramaic" in inventories
+
+
+def test_get_phonological_inventories_glottocode_keys() -> None:
+    """The keys of the return value of `get_phonological_inventories` should be
+    Glottocodes, except for special keys.
+    """
+    special = {"*", "Djindewal", "ModernAramaic"}
     inventories = get_phonological_inventories()
-    assert "NA" in inventories
+    for glottocode in inventories:
+        is_glottocode = re.match("[a-z]{4}[0-9]{4}", glottocode)
+        if not is_glottocode:
+            assert glottocode in special
 
 
 def test_get_sounds() -> None:
@@ -42,7 +62,7 @@ def test_get_sounds() -> None:
     inventory of the language.
     """
     # Test on a few languages, because it's slow.
-    languages = ["eng", "tgl"]
+    languages = ["*", "stan1293", "taga1270"]
     inventories = get_phonological_inventories()
     for language in languages:
         sounds = set(inventories[language])