1
0
mirror of https://github.com/nextapps-de/flexsearch.git synced 2025-09-25 12:58:59 +02:00
Files
flexsearch/dist/module-debug/encoder.js
Thomas Wilkerling 2431b51a4f update index.d.ts
2025-05-08 22:08:07 +02:00

588 lines
19 KiB
JavaScript
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

import { create_object, merge_option } from "./common.js";
import normalize_polyfill from "./charset/polyfill.js";
import { EncoderOptions } from "./type.js";
/*
Custom Encoder
----------------
// Split a passed string into an Array of words:
function englishEncoder(string){
return string.toLowerCase().split(/[^a-z]+/)
}
// For CJK split a passed string into an Array of chars:
function chineseEncoder(string){
return string.replace(/\s+/, "").split("")
}
// Alternatively do not split the input:
function fixedEncoder(string){
return [string]
}
Built-in Encoder
----------------------------
The main workflow follows an increasing strategy,
starting from a simple .toLowerCase() to full RegExp
Pipeline:
1. apply this.normalize (charset normalization)
applied on the whole input string e.g. lowercase,
everything you put later into (filter, matcher, stemmer, mapper, etc.)
has to be normalized by definition, because it won't apply to them automatically
2. apply this.prepare (custom function, string in - string out)
3 split numerics into triplets
4. split input into terms (by one of them: split/include/exclude)
5. pre-encoded term deduplication
6. apply this.filter (stop-words)
7. apply this.stemmer (replace term endings)
8. apply this.mapper (replace chars)
9. apply this.dedupe (letter deduplication)
10. apply this.matcher (replace terms)
11. apply this.replacer (custom regex)
12. post-encoded term deduplication
13. apply this.finalize (custom function, array in - array out)
*/
const whitespace = /[^\p{L}\p{N}]+/u,
numeric_split_length = /(\d{3})/g,
numeric_split_prev_char = /(\D)(\d{3})/g,
numeric_split_next_char = /(\d{3})(\D)/g,
normalize = /*"".normalize &&*//[\u0300-\u036f]/g; // /[\p{Z}\p{S}\p{P}\p{C}]+/u;
//const numeric_split = /(\d{3})/g;
//.replace(/(\d{3})/g, "$1 ")
//.replace(/([^\d])([\d])/g, "$1 $2")
//.replace(/([\d])([^\d])/g, "$1 $2")
// '´`ʼ.,
//const normalize_mapper = SUPPORT_CHARSET && !normalize && normalize_polyfill;
/**
* @param {EncoderOptions=} options
* @constructor
*/
export default function Encoder(options = {}) {
if (!this || this.constructor !== Encoder) {
// let args = Array.prototype.slice.call(arguments);
// args.unshift(Encoder);
// return new (Encoder.bind.apply(Encoder, args));
return new Encoder(...arguments);
}
if (arguments.length) {
for (let i = 0; i < arguments.length; i++) {
this.assign(arguments[i]);
}
} else {
this.assign(options);
}
}
/**
* @param {!EncoderOptions} options
*/
Encoder.prototype.assign = function (options) {
/**
* pre-processing string input
* @type {Function|boolean}
*/
this.normalize = /** @type {Function|boolean} */merge_option(options.normalize, /* tag? */ /* stringify */ /* stringify */ /* single param */ /* skip update: */ /* append: */ /* skip update: */ /* skip_update: */ /* skip deletion */!0 /*await rows.hasNext()*/
/*await rows.hasNext()*/ /*await rows.hasNext()*/, this.normalize);
// {
// letter: true,
// number: true,
// whitespace: true,
// symbol: true,
// punctuation: true,
// control: true,
// char: ""
// }
let include = options.include,
tmp = include || options.exclude || options.split,
numeric;
if (tmp || "" === tmp) {
if ("object" == typeof tmp && tmp.constructor !== RegExp) {
let regex = "";
numeric = !include;
// split on whitespace by default
include || (regex += "\\p{Z}");
if (tmp.letter) {
regex += "\\p{L}";
}
if (tmp.number) {
regex += "\\p{N}";
numeric = !!include;
}
if (tmp.symbol) {
regex += "\\p{S}";
}
if (tmp.punctuation) {
regex += "\\p{P}";
}
if (tmp.control) {
regex += "\\p{C}";
}
if (tmp = tmp.char) {
regex += "object" == typeof tmp ? tmp.join("") : tmp;
}
try {
// https://github.com/nextapps-de/flexsearch/issues/410
/**
* split string input into terms
* @type {string|RegExp|boolean|null}
*/
this.split = new RegExp("[" + (include ? "^" : "") + regex + "]+", "u");
} catch (e) {
console.error("Your split configuration:", tmp, "is not supported on this platform. It falls back to using simple whitespace splitter instead: /\s+/.");
// fallback to a simple whitespace splitter
this.split = /\s+/;
}
} else {
this.split = /** @type {string|RegExp|boolean} */tmp;
// determine numeric encoding
numeric = /* suggest */ /* append: */ /* enrich */!1 === tmp || 2 > "a1a".split(tmp).length;
}
this.numeric = merge_option(options.numeric, numeric);
} else {
try {
// https://github.com/nextapps-de/flexsearch/issues/410
this.split = /** @type {string|RegExp|boolean} */merge_option(this.split, whitespace);
} catch (e) {
console.warn("This platform does not support unicode regex. It falls back to using simple whitespace splitter instead: /\s+/.");
// fallback to a simple whitespace splitter
this.split = /\s+/;
}
this.numeric = merge_option(options.numeric, merge_option(this.numeric, !0));
}
/**
* post-processing terms
* @type {Function|null}
*/
this.prepare = /** @type {Function|null} */merge_option(options.prepare, null, this.prepare);
/**
* final processing
* @type {Function|null}
*/
this.finalize = /** @type {Function|null} */merge_option(options.finalize, null, this.finalize);
// assign the normalization fallback to the mapper
// if(SUPPORT_CHARSET && !normalize){
// this.mapper = new Map(
// /** @type {Array<Array<string, string>>} */ (
// normalize_polyfill
// )
// );
// }
tmp = options.filter;
this.filter = "function" == typeof tmp ? tmp : merge_option(tmp && new Set(tmp), null, this.filter);
this.dedupe = merge_option(options.dedupe, !0, this.dedupe);
this.matcher = merge_option((tmp = options.matcher) && new Map(tmp), null, this.matcher);
this.mapper = merge_option((tmp = options.mapper) && new Map(tmp), null, this.mapper);
this.stemmer = merge_option((tmp = options.stemmer) && new Map(tmp), null, this.stemmer);
this.replacer = merge_option(options.replacer, null, this.replacer);
this.minlength = merge_option(options.minlength, 1, this.minlength);
this.maxlength = merge_option(options.maxlength, 1024, this.maxlength);
this.rtl = merge_option(options.rtl, !1, this.rtl);
// auto-balanced cache
this.cache = tmp = merge_option(options.cache, !0, this.cache);
if (tmp) {
this.timer = null;this.cache_size = "number" == typeof tmp ? tmp : 2e5;
this.cache_enc = new Map();
this.cache_term = new Map();
this.cache_enc_length = 128;
this.cache_term_length = 128;
}
// regex temporary state
this.matcher_str = "";
this.matcher_test = null;
this.stemmer_str = "";
this.stemmer_test = null;
// prebuilt
// if(this.filter && this.split){
// for(const key of this.filter){
// const tmp = key.replace(this.split, "");
// if(key !== tmp){
// this.filter.delete(key);
// this.filter.add(tmp);
// }
// }
// }
if (this.matcher) {
for (const key of this.matcher.keys()) {
this.matcher_str += (this.matcher_str ? "|" : "") + key;
}
}
if (this.stemmer) {
for (const key of this.stemmer.keys()) {
this.stemmer_str += (this.stemmer_str ? "|" : "") + key;
}
}
// if(SUPPORT_COMPRESSION){
// this.compression = merge_option(options.compress || options.compression, 0, this.compression);
// if(this.compression && !table){
// table = new Array(radix);
// for(let i = 0; i < radix; i++) table[i] = i + 33;
// table = String.fromCharCode.apply(null, table);
// }
// }
return this;
};
Encoder.prototype.addStemmer = function (match, replace) {
this.stemmer || (this.stemmer = new Map());
this.stemmer.set(match, replace);
this.stemmer_str += (this.stemmer_str ? "|" : "") + match;
this.stemmer_test = null;
this.cache && clear(this);
return this;
};
Encoder.prototype.addFilter = function (term) {
if ("function" == typeof term) {
// does not support merge yet
this.filter = term; //merge_option(term, term, this.filter);
} else {
this.filter || (this.filter = new Set());
this.filter.add(term);
}
this.cache && clear(this);
return this;
};
/**
* Replace a single char
* @param {string} char_match
* @param {string} char_replace
* @return {Encoder}
* @suppress {invalidCasts}
*/
Encoder.prototype.addMapper = function (char_match, char_replace) {
// regex:
if ("object" == typeof char_match) {
return this.addReplacer( /** @type {RegExp} */char_match, char_replace);
}
// not a char:
if (1 < char_match.length) {
return this.addMatcher(char_match, char_replace);
}
this.mapper || (this.mapper = new Map());
this.mapper.set(char_match, char_replace);
this.cache && clear(this);
return this;
};
/**
* Replace a string
* @param {string} match
* @param {string} replace
* @return {Encoder}
* @suppress {invalidCasts}
*/
Encoder.prototype.addMatcher = function (match, replace) {
// regex:
if ("object" == typeof match) {
return this.addReplacer( /** @type {RegExp} */match, replace);
}
// a single char:
// only downgrade when dedupe is on or mapper already was filled
if (2 > match.length && (this.dedupe || this.mapper)) {
return this.addMapper(match, replace);
}
this.matcher || (this.matcher = new Map());
this.matcher.set(match, replace);
this.matcher_str += (this.matcher_str ? "|" : "") + match;
this.matcher_test = null;
this.cache && clear(this);
return this;
};
/**
* @param {RegExp} regex
* @param {string} replace
* @return {Encoder}
* @suppress {invalidCasts}
*/
Encoder.prototype.addReplacer = function (regex, replace) {
if ("string" == typeof regex) {
return this.addMatcher( /** @type {string} */regex, replace);
}
this.replacer || (this.replacer = []);
this.replacer.push(regex, replace);
this.cache && clear(this);
return this;
};
/**
* @param {!string} str
* @param {boolean=} dedupe_terms Note: term deduplication will break the context chain
* @return {!Array<string>}
*/
Encoder.prototype.encode = function (str, dedupe_terms) {
if (this.cache && str.length <= this.cache_enc_length) {
if (this.timer) {
if (this.cache_enc.has(str)) {
return this.cache_enc.get(str);
}
} else {
this.timer = setTimeout(clear, 50, this);
}
}
// apply charset normalization
if (this.normalize) {
if ("function" == typeof this.normalize) {
str = this.normalize(str);
} else if (normalize) {
str = str.normalize("NFKD").replace(normalize, "").toLowerCase();
} else {
str = str.toLowerCase();
}
}
// apply custom encoder (can replace split)
if (this.prepare) {
str = this.prepare(str);
}
// split numbers into triplets
if (this.numeric && 3 < str.length) {
str = str.replace(numeric_split_prev_char, "$1 $2").replace(numeric_split_next_char, "$1 $2").replace(numeric_split_length, "$1 ");
}
// if(this.matcher && (str.length > 1)){
// this.matcher_test || (
// this.matcher_test = new RegExp("(" + this.matcher_str + ")", "g")
// );
// str = str.replace(this.matcher_test, match => this.matcher.get(match));
// }
// if(this.stemmer){
// this.stemmer_test || (
// this.stemmer_test = new RegExp("(?!\\b)(" + this.stemmer_str + ")(\\b|_)", "g")
// );
// str = str.replace(this.stemmer_test, match => this.stemmer.get(match));
// }
const skip = !(this.dedupe || this.mapper || this.filter || this.matcher || this.stemmer || this.replacer);
let final = [],
dupes = create_object(),
last_term,
last_term_enc,
words = this.split || "" === this.split ? str.split( /** @type {string|RegExp} */this.split) : [str];
// str;
for (let i = 0, word, base; i < words.length; i++) {
if (!(word = base = words[i])) {
continue;
}
if (word.length < this.minlength || word.length > this.maxlength) {
continue;
}
if (dedupe_terms) {
if (dupes[word]) {
continue;
}
dupes[word] = 1;
} else {
if (last_term === word) {
continue;
}
last_term = word;
}
if (skip) {
final.push(word);
continue;
}
if (this.filter && ("function" == typeof this.filter ? !this.filter(word) : this.filter.has(word))) {
continue;
}
if (this.cache && word.length <= this.cache_term_length) {
if (this.timer) {
const tmp = this.cache_term.get(word);
if (tmp || "" === tmp) {
tmp && final.push(tmp);
continue;
}
} else {
this.timer = setTimeout(clear, 50, this);
}
}
// from here minlength should not apply again
// when the input string is further shrinking
// it needs to apply stemmer before bigger transformations
// it needs to apply stemmer after filter (user -> us -> filter out)
if (this.stemmer) {
// for(const item of this.stemmer){
// const key = item[0];
// const value = item[1];
// if(word.length > key.length && word.endsWith(key)){
// word = word.substring(0, word.length - key.length) + value;
// break;
// }
// // const position = word.length - key.length;
// // if(position > 0 && word.substring(position) === key){
// // word = word.substring(0, position) + value;
// // break;
// // }
// }
// todo compare advantages when filter/stemmer are also encoded
this.stemmer_test || (this.stemmer_test = new RegExp("(?!^)(" + this.stemmer_str + ")$"));
let old;
// loop stemmer as long as anything has matched
// just terms with length > 2 should need a stemmer (its -> it)
// the minlength also prevents stemmer looping to cut off everything
while (old !== word && 2 < word.length) {
old = word;
word = word.replace(this.stemmer_test, match => this.stemmer.get(match));
}
}
// apply mapper and collapsing
if (word && (this.mapper || this.dedupe && 1 < word.length)) {
let final = "";
for (let i = 0, prev = "", char, tmp; i < word.length; i++) {
char = word.charAt(i);
if (char !== prev || !this.dedupe) {
tmp = this.mapper && this.mapper.get(char);
if (!tmp && "" !== tmp) final += prev = char;else if ((tmp !== prev || !this.dedupe) && (prev = tmp)) final += tmp;
}
}
word = final;
}
// apply matcher
if (this.matcher && 1 < word.length) {
this.matcher_test || (this.matcher_test = new RegExp("(" + this.matcher_str + ")", "g"));
word = word.replace(this.matcher_test, match => this.matcher.get(match));
}
// apply custom regex
if (word && this.replacer) {
for (let i = 0; word && i < this.replacer.length; i += 2) {
word = word.replace(this.replacer[i], this.replacer[i + 1]);
}
}
// slower variants for removing same chars in a row:
//word = word.replace(/([^0-9])\1+/g, "$1");
//word = word.replace(/(.)\1+/g, "$1");
//word = word.replace(/(?<=(.))\1+/g, "");
if (this.cache && base.length <= this.cache_term_length) {
this.cache_term.set(base, word);
if (this.cache_term.size > this.cache_size) {
this.cache_term.clear();
this.cache_term_length = 0 | this.cache_term_length / 1.1;
}
}
if (word) {
if (word !== base) {
if (dedupe_terms) {
if (dupes[word]) {
continue;
}
dupes[word] = 1;
} else {
if (last_term_enc === word) {
continue;
}
last_term_enc = word;
}
}
final.push(word);
}
}
if (this.finalize) {
final = this.finalize(final) || final;
}
if (this.cache && str.length <= this.cache_enc_length) {
this.cache_enc.set(str, final);
if (this.cache_enc.size > this.cache_size) {
this.cache_enc.clear();
this.cache_enc_length = 0 | this.cache_enc_length / 1.1;
}
}
return final;
};
export function fallback_encoder(str) {
return str.normalize("NFKD").replace(normalize, "").toLowerCase().trim().split(/\s+/);
}
// Encoder.prototype.compress = function(str) {
//
// //return str;
// //if(!str) return str;
//
// if(SUPPORT_CACHE && this.cache && str.length <= this.cache_term_length){
// if(this.timer){
// if(this.cache_cmp.has(str)){
// return this.cache_cmp.get(str);
// }
// }
// else{
// this.timer = setTimeout(clear, 0, this);
// }
// }
//
// const result = typeof this.compression === "function"
// ? this.compression(str)
// : hash(str); //window.hash(str);
//
// if(SUPPORT_CACHE && this.cache && str.length <= this.cache_term_length){
// this.cache_cmp.set(str, result);
// this.cache_cmp.size > this.cache_size &&
// this.cache_cmp.clear();
// }
//
// return result;
// };
// function hash(str){
// return str;
// }
/**
* @param {Encoder} self
*/
function clear(self) {
self.timer = null;
self.cache_enc.clear();
self.cache_term.clear();
}