From 436d4ed5c857d879e8741dbdad8c69d3ce10ea68 Mon Sep 17 00:00:00 2001
From: Adrien Barbaresi <adbar@users.noreply.github.com>
Date: Thu, 14 Nov 2024 17:47:09 +0100
Subject: [PATCH] add demo rules for Estonian (#157)

* add rules for Estonian

* test_rules.py: add newline

* better rules and data

* prepare merge

* fix pickler
---
 simplemma/strategies/defaultrules/__init__.py |  2 +
 simplemma/strategies/defaultrules/et.py       | 42 +++++++++++++++++++
 tests/strategies/defaultrules/test_rules.py   | 16 ++++++-
 3 files changed, 59 insertions(+), 1 deletion(-)
 create mode 100644 simplemma/strategies/defaultrules/et.py

diff --git a/simplemma/strategies/defaultrules/__init__.py b/simplemma/strategies/defaultrules/__init__.py
index 21d8bcc..2e72cb3 100644
--- a/simplemma/strategies/defaultrules/__init__.py
+++ b/simplemma/strategies/defaultrules/__init__.py
@@ -4,6 +4,7 @@
 
 from .de import apply_de
 from .en import apply_en
+from .et import apply_et
 from .fi import apply_fi
 from .lv import apply_lv
 from .nl import apply_nl
@@ -13,6 +14,7 @@
 DEFAULT_RULES: Dict[str, Callable[[str], Optional[str]]] = {
     "de": apply_de,
     "en": apply_en,
+    "et": apply_et,
     "fi": apply_fi,
     "lv": apply_lv,
     "nl": apply_nl,
diff --git a/simplemma/strategies/defaultrules/et.py b/simplemma/strategies/defaultrules/et.py
new file mode 100644
index 0000000..5cbe96a
--- /dev/null
+++ b/simplemma/strategies/defaultrules/et.py
@@ -0,0 +1,42 @@
+import re
+from typing import Optional
+
+from .generic import apply_rules
+
+
+## Just a demo, the rules are really basic and coverage is not good
+
+
+DEFAULT_RULES = {
+    # adjectives
+    # https://en.wiktionary.org/wiki/-line
+    re.compile(
+        r"(?:lise|list|lisse|lisesse|lises|lisest|lisele|lisel|liselt|liseks|liseni|lisena|liseta|lisega|lised|liste|lisi|listesse|lisisse|listes|lisis|listest|lisist|listele|lisile|listel|lisil|listelt|lisilt|listeks|lisiks|listeni|listena|listeta|listega)$"
+    ): "line",
+    # https://en.wiktionary.org/wiki/-mine
+    re.compile(
+        r"(?:mise|mist|misse|misesse|mises|misest|misele|misel|miselt|miseks|miseni|misena|miseta|misega|mised|miste|misi|mistesse|misisse|mistes|misis|mistest|misist|mistele|misile|mistel|misil|mistelt|misilt|misteks|misiks|misteni|mistena|misteta|mistega)$"
+    ): "mine",
+    # nouns
+    # https://en.wiktionary.org/wiki/-dus
+    re.compile(
+        r"(?:duse|dust|dusse|dusesse|duses|dusest|dusele|dusel|duselt|duseks|duseni|dusena|duseta|dusega|dused|duste|dusi|dustesse|dusisse|dustes|dusis|dustest|dusist|dustele|dusile|dustel|dusil|dustelt|dusilt|dusteks|dusiks|dusteni|dustena|dusteta|dustega)$"
+    ): "dus",
+    # https://en.wiktionary.org/wiki/-lik
+    # https://en.wiktionary.org/wiki/-nik
+    re.compile(
+        r"(?:iku|ikku|ikusse|ikus|ikust|ikule|ikul|ikult|ikuks|ikuni|ikuna|ikuta|ikuga|ikud|ike|ikudde|ikke|ikusid|ikesse|ikkudesse|ikes|ikkudes|ikest|ikkudest|ikele|ikkudele|ikel|ikkudel|ikelt|ikkudelt|ikeks|ikkudeks|ikeni|ikkudeni|ikena|ikkudena|iketa|ikkudeta|ikega|ikkudega)$"
+    ): "ik",
+    # https://en.wiktionary.org/wiki/-kond
+    re.compile(
+        r"(?:konna|konda|konnasse|konnas|konnast|konnale|konnal|konnalt|konnaks|konnani|konnana|konnata|konnaga|konnad|kondade|kondi|kondasid|kondadesse|konnisse|kondades|konnis|kondadest|konnist|kondadele|konnile|kondadel|konnil|kondadelt|konnilt|kondadeks|konniks|kondadeni|kondadena|kondadeta|kondadega)$"
+    ): "kond",
+}
+
+
+def apply_et(token: str) -> Optional[str]:
+    "Apply pre-defined rules for Estonian."
+    if len(token) < 8 or token[0].isupper():
+        return None
+
+    return apply_rules(token, DEFAULT_RULES)
diff --git a/tests/strategies/defaultrules/test_rules.py b/tests/strategies/defaultrules/test_rules.py
index 6da090a..4cf0868 100644
--- a/tests/strategies/defaultrules/test_rules.py
+++ b/tests/strategies/defaultrules/test_rules.py
@@ -4,14 +4,28 @@
 def test_DEFAULT_RULES() -> None:
     """Test rules on all available languages."""
     rules_strategy = RulesStrategy()
+
     assert rules_strategy.get_lemma("Pfifferlinge", "de") == "Pfifferling"
-    assert rules_strategy.get_lemma("Pfifferlinge", "en") is None
     assert rules_strategy.get_lemma("atonements", "de") is None
+
     assert rules_strategy.get_lemma("atonements", "en") == "atonement"
+    assert rules_strategy.get_lemma("Pfifferlinge", "en") is None
+
     assert rules_strategy.get_lemma("brieven", "nl") == "brief"
+
     assert rules_strategy.get_lemma("liikenaisessa", "fi") == "liikenainen"
+
     assert rules_strategy.get_lemma("pracowaliście", "pl") == "pracować"
+
     assert rules_strategy.get_lemma("безгра́мотностью", "ru") == "безгра́мотность"
+
     assert rules_strategy.get_lemma("Rīga", "lv") is None
     assert rules_strategy.get_lemma("šķirkļiem", "lv") == "šķirklis"
     assert rules_strategy.get_lemma("mācībām", "lv") == "mācība"
+
+    assert rules_strategy.get_lemma("Läänemere", "et") is None
+    assert rules_strategy.get_lemma("tavalised", "et") == "tavaline"
+    assert rules_strategy.get_lemma("peamisteks", "et") == "peamine"
+    assert rules_strategy.get_lemma("tähendustena", "et") == "tähendus"
+    assert rules_strategy.get_lemma("kunstnikud", "et") == "kunstnik"
+    assert rules_strategy.get_lemma("keelkondade", "et") == "keelkond"