Simplify diacratic removal; modify Latin & Greek preprocessors (#724)

* Simplified diacratic removal and added preprocessors to LA and GRC * linted * Clarified the name of removeAlphabeticDiacritics * Add comment to removeAlphabeticDiacritics Signed-off-by: Darius Jahandarie <[email protected]> * Change to NFD Signed-off-by: Matttttt <[email protected]> * Remove trailing spaces in comment Signed-off-by: Darius Jahandarie <[email protected]> * Remove latin preprocessors .eslintrc.json Signed-off-by: Matttttt <[email protected]> * fix tests --------- Signed-off-by: Darius Jahandarie <[email protected]> Signed-off-by: Matttttt <[email protected]> Co-authored-by: martholomew <[email protected]> Co-authored-by: Darius Jahandarie <[email protected]> Co-authored-by: Stefan Vukovic <[email protected]>
yomidevs · Apr 8, 2024 · fe516c4 · fe516c4
1 parent 2c5af21
commit fe516c4
Show file tree

Hide file tree

Showing 5 changed files with 24 additions and 63 deletions.
diff --git a/.eslintrc.json b/.eslintrc.json
@@ -647,7 +647,6 @@
                 "ext/js/language/ja/japanese-transforms.js",
                 "ext/js/language/ja/japanese-wanakana.js",
                 "ext/js/language/ja/japanese.js",
-                "ext/js/language/la/latin-text-preprocessors.js",
                 "ext/js/language/language-descriptors.js",
                 "ext/js/language/language-transformer.js",
                 "ext/js/language/language-transforms.js",

diff --git a/ext/js/language/la/latin-text-preprocessors.js b/ext/js/language/la/latin-text-preprocessors.js
diff --git a/ext/js/language/language-descriptors.js b/ext/js/language/language-descriptors.js
@@ -21,10 +21,9 @@ import {englishTransforms} from './en/english-transforms.js';
 import {collapseEmphaticSequences, convertAlphabeticCharacters, convertHalfWidthCharacters, convertHiraganaToKatakana, convertNumericCharacters} from './ja/japanese-text-preprocessors.js';
 import {japaneseTransforms} from './ja/japanese-transforms.js';
 import {isStringPartiallyJapanese} from './ja/japanese.js';
-import {removeLatinDiacritics} from './la/latin-text-preprocessors.js';
 import {removeRussianDiacritics, yoToE} from './ru/russian-text-preprocessors.js';
 import {albanianTransforms} from './sq/albanian-transforms.js';
-import {capitalizeFirstLetter, decapitalize} from './text-preprocessors.js';
+import {capitalizeFirstLetter, decapitalize, removeAlphabeticDiacritics} from './text-preprocessors.js';
 
 const capitalizationPreprocessors = {
     decapitalize,
@@ -87,7 +86,10 @@ const languageDescriptors = [
         iso: 'grc',
         name: 'Ancient Greek',
         exampleText: 'γράφω',
-        textPreprocessors: capitalizationPreprocessors
+        textPreprocessors: {
+            ...capitalizationPreprocessors,
+            removeAlphabeticDiacritics
+        }
     },
     {
         iso: 'hu',
@@ -113,7 +115,7 @@ const languageDescriptors = [
         exampleText: 'legere',
         textPreprocessors: {
             ...capitalizationPreprocessors,
-            removeLatinDiacritics
+            removeAlphabeticDiacritics
         }
     },
     {

diff --git a/ext/js/language/text-preprocessors.js b/ext/js/language/text-preprocessors.js
@@ -33,3 +33,17 @@ export const capitalizeFirstLetter = {
     options: basicTextPreprocessorOptions,
     process: (str, setting) => (setting ? str.charAt(0).toUpperCase() + str.slice(1) : str)
 };
+
+/**
+ * WARNING: This should NOT be used with languages that use Han characters,
+ *          as it can result in undesirable normalization:
+ *            - '\u9038'.normalize('NFD') => '\u9038' (逸)
+ *            - '\ufa67'.normalize('NFD') => '\u9038' (逸 => 逸)
+ * @type {import('language').TextPreprocessor<boolean>}
+ */
+export const removeAlphabeticDiacritics = {
+    name: 'Remove Alphabetic Diacritics',
+    description: 'ἄήé -> αηe',
+    options: basicTextPreprocessorOptions,
+    process: (str, setting) => (setting ? str.normalize('NFD').replace(/[\u0300-\u036f]/g, '') : str)
+};
diff --git a/types/ext/language-descriptors.d.ts b/types/ext/language-descriptors.d.ts
@@ -69,12 +69,14 @@ type AllTextPreprocessors = {
         removeArabicScriptDiacritics: TextPreprocessor<boolean>;
     };
     fr: CapitalizationPreprocessors;
-    grc: CapitalizationPreprocessors;
+    grc: CapitalizationPreprocessors & {
+        removeAlphabeticDiacritics: TextPreprocessor<boolean>;
+    };
     hu: CapitalizationPreprocessors;
     id: CapitalizationPreprocessors;
     it: CapitalizationPreprocessors;
     la: CapitalizationPreprocessors & {
-        removeLatinDiacritics: TextPreprocessor<boolean>;
+        removeAlphabeticDiacritics: TextPreprocessor<boolean>;
     };
     ja: {
         convertHalfWidthCharacters: TextPreprocessor<boolean>;