From 2f448636e9e6f132cb5c9cfc92c6d4c1c59f5051 Mon Sep 17 00:00:00 2001
From: sspanak <doftor.livain@gmail.com>
Date: Tue, 19 Dec 2023 13:02:16 +0200
Subject: [PATCH] scripts for cleaning up and adding new lists of words

---
 scripts/add-new-words.sh |  39 ++++++++++++++
 scripts/injest-words.js  | 107 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 146 insertions(+)
 create mode 100755 scripts/add-new-words.sh
 create mode 100644 scripts/injest-words.js

diff --git a/scripts/add-new-words.sh b/scripts/add-new-words.sh
new file mode 100755
index 00000000..d86d5a4e
--- /dev/null
+++ b/scripts/add-new-words.sh
@@ -0,0 +1,39 @@
+#!/bin/bash
+
+if [ $# -lt 4 ]; then
+	echo "Usage: $0 LOCALE base-dictionary-file.csv new-words-file.txt frequency-file.csv"
+	echo 'Cleans up and adds new words to a dictionary file.'
+	echo 'LOCALE could be any valid JS locale, for exmaple: en, en-US, etc...'
+	exit 1
+fi
+
+if ! [[ -f $2 ]]; then
+	echo "base-dictionary-file: '$2' does not exist"
+	exit 2
+fi
+
+if ! [[ -f $3 ]]; then
+	echo "new-words-file: '$3' does not exist"
+	exit 2
+fi
+
+if ! [[ -f $4 ]]; then
+	echo "frequency-file: '$4' does not exist"
+	exit 2
+fi
+
+LOCALE=$1
+DICTIONARY_FILE=$2
+NEW_WORDS_FILE=$3
+FREQUENCY_FILE=$4
+
+sed -E 's/[\t0-9]+//g' $DICTIONARY_FILE > /tmp/_TT9_base.txt \
+	&& node scripts/injest-words.js $NEW_WORDS_FILE > /tmp/_TT9_1.txt \
+	&& node scripts/remove-foreign-words.js $LOCALE /tmp/_TT9_1.txt $LOCALE /tmp/_TT9_base.txt > /tmp/_TT9_2.txt \
+	&& cp /tmp/_TT9_base.txt /tmp/_TT9_combined.txt \
+	&& cat /tmp/_TT9_2.txt >> /tmp/_TT9_combined.txt \
+	&& node scripts/remove-dictionary-repeating-words.js $LOCALE /tmp/_TT9_combined.txt > /tmp/_TT9_clean.txt \
+	&& node scripts/inject-dictionary-frequencies.js /tmp/_TT9_clean.txt $FREQUENCY_FILE $LOCALE > /tmp/_TT9_output.txt \
+	&& cat /tmp/_TT9_output.txt
+
+rm -f /tmp/_TT9*
diff --git a/scripts/injest-words.js b/scripts/injest-words.js
new file mode 100644
index 00000000..01c63e95
--- /dev/null
+++ b/scripts/injest-words.js
@@ -0,0 +1,107 @@
+const { basename } = require('path');
+const { createReadStream, existsSync } = require('fs');
+const { createInterface } = require('readline');
+
+
+function printHelp() {
+	console.log(`Usage ${basename(process.argv[1])} word-list.txt`);
+	console.log('Breaks dashed words into separate words, puts multiple words on a line on new lines and deletes repeating new lines.');
+}
+
+
+
+function validateInput() {
+	if (process.argv.length < 3) {
+		printHelp();
+		process.exit(1);
+	}
+
+	if (!existsSync(process.argv[2])) {
+		console.error(`Failure! Could not find word list file "${process.argv[3]}."`);
+		process.exit(2);
+	}
+
+	return {
+		fileName: process.argv[2]
+	}
+}
+
+
+function printWords(wordList) {
+	if (Array.isArray(wordList)) {
+		wordList.forEach(w => console.log(w));
+	}
+}
+
+
+function cleanSpecialChars(line) {
+	const spacesOnly = /^\s+$/;
+	const digits = /\d+/;
+
+	if (!line || !line.length || spacesOnly.test(line)) {
+		return [];
+	}
+
+	return line
+		.replaceAll(/[\x01-\x20]+/g, ' ')
+		.replaceAll(/[&\s,":;\*/]+/g, ' ')
+		.replaceAll(/\s+/g, ' ')
+		.replaceAll(/[\[\]\.\?\(\)]/g, '')
+		.split(' ')
+		.filter(w => w.length > 1 && !digits.test(w));
+}
+
+
+function splitDashedWords(inputWords) {
+	if (!Array.isArray(inputWords)) {
+		return [];
+	}
+
+	const roots = {};
+	const words = {};
+
+	for (const word of inputWords) {
+		if (!word.includes('-')) {
+			words[word] = true;
+			continue;
+		}
+
+		const parts = word.split('-');
+		for (let i = 0; i < parts.length - 1; i++) {
+			const key = `${parts[i]}-`;
+			if (key in roots) {
+				words[key] = true;
+			} else {
+				roots[key] = true;
+				words[parts[i]] = true;
+			}
+		}
+
+		words[parts[parts.length - 1]] = true;
+	}
+
+	return Object.keys(words);
+}
+
+
+async function work({ fileName }) {
+	words = [];
+
+	let lineReader = createInterface({ input: createReadStream(fileName) });
+	for await (const line of lineReader) {
+		newWords = cleanSpecialChars(line);
+
+		words = [
+			...words,
+			...newWords
+		];
+	}
+
+	return splitDashedWords(words).filter(w => w.length > 1).sort();
+}
+
+
+/** main **/
+work(validateInput())
+	.then(words => printWords(words))
+	.catch(e => console.error(e));