1
0
Fork 0

Brazilian Portuguese (#203)

* added Brazilian Portuguese

* more dictionary processing scripts
This commit is contained in:
Dimo Karaivanov 2023-03-13 10:21:36 +02:00 committed by GitHub
parent f146eaeae2
commit cd03dec9ff
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
11 changed files with 261722 additions and 18 deletions

View file

@ -1,6 +1,6 @@
const { basename } = require('path');
const { createReadStream, existsSync } = require('fs');
const { createInterface } = require('readline');
const GEO_NAME = /[A-Z]\w+\-[^\n]+/;
@ -40,37 +40,49 @@ function getRegularWordKey(locale, word) {
function getWordKeyPreservingCaptialization(locale, word, wordMap) {
if (typeof word !== 'string' || word.length === 0 || typeof wordMap !== 'object') {
function getLowercaseWordKey(locale, word) {
return getWordkey(word).toLocaleLowerCase(locale);
}
function getWordkey(word) {
if (typeof word !== 'string' || word.length === 0) {
return '';
}
let wordKey = word.toLocaleLowerCase(locale);
if (GEO_NAME.test(word) || word.toLocaleLowerCase(locale) !== word) {
wordKey = word;
if (wordMap[word.toLocaleLowerCase(locale)]) {
delete wordMap[word.toLocaleLowerCase(locale)];
}
}
return wordKey;
return word;
}
async function removeRepeatingWords({ fileName, locale }) {
const lineReader = require('readline').createInterface({
input: createReadStream(fileName)
});
const wordMap = {};
let lineReader = createInterface({ input: createReadStream(fileName) });
for await (const line of lineReader) {
wordMap[getWordKeyPreservingCaptialization(locale, line, wordMap)] = true;
wordMap[getLowercaseWordKey(locale, line)] = true;
}
return Object.keys(wordMap);
lineReader = createInterface({ input: createReadStream(fileName) });
for await (const line of lineReader) {
const word = getWordkey(line);
const lowercaseWord = getLowercaseWordKey(locale, line);
if (word === '') {
continue;
}
if (word !== lowercaseWord) {
delete wordMap[lowercaseWord];
wordMap[word] = true;
}
}
return Object.keys(wordMap).sort();
}

View file

@ -0,0 +1,98 @@
const { basename } = require('path');
const { createReadStream, existsSync } = require('fs');
const { createInterface } = require('readline');
function printHelp() {
console.log(`Usage ${basename(process.argv[1])} DICTIONARY_LOCALE DICTIONARY.TXT FOREIGN_WORDS_LOCALE FOREIGN-WORD-DICTIONARY.txt`);
console.log('Removes foreign words from a dictionary');
}
function validateInput() {
if (process.argv.length < 6) {
printHelp();
process.exit(1);
}
if (!existsSync(process.argv[3])) {
console.error(`Failure! Could not find words file "${process.argv[3]}."`);
process.exit(2);
}
if (!existsSync(process.argv[5])) {
console.error(`Failure! Could not find foreign words file "${process.argv[4]}."`);
process.exit(2);
}
return {
locale: process.argv[2],
fileName: process.argv[3],
foreignWordsLocale: process.argv[4],
foreignWordsFileName: process.argv[5]
};
}
function getLowercaseWordKey(locale, word) {
return getWordkey(word).toLocaleLowerCase(locale);
}
function getWordkey(word) {
if (typeof word !== 'string' || word.length === 0) {
return '';
}
return word;
}
async function removeForeignWords({ locale, foreignWordsLocale, fileName, foreignWordsFileName }) {
const foreignWords = {};
let lineReader = createInterface({ input: createReadStream(foreignWordsFileName) });
for await (const line of lineReader) {
foreignWords[getLowercaseWordKey(foreignWordsLocale, line)] = true;
}
const wordMap = {};
lineReader = createInterface({ input: createReadStream(fileName) });
for await (const line of lineReader) {
const word = getWordkey(line);
const lowercaseWord = getLowercaseWordKey(locale, line);
if (word === '') {
continue;
}
if (!foreignWords[lowercaseWord]) {
wordMap[word] = true;
}
}
return Object.keys(wordMap);
}
function printWords(wordList) {
if (!Array.isArray(wordList)) {
return;
}
wordList.forEach(w => console.log(w));
}
/** main **/
removeForeignWords(validateInput())
.then(words => printWords(words))
.catch(e => console.error(e));