1
0
mirror of https://github.com/nextapps-de/flexsearch.git synced 2025-09-26 05:19:00 +02:00

bundle pre-release

This commit is contained in:
Thomas Wilkerling
2025-03-07 17:44:10 +01:00
parent 5bcbc72dac
commit 25e4b5d712
219 changed files with 44901 additions and 16106 deletions

View File

@@ -1,4 +1,231 @@
import { create_object, concat } from "./common.js";
import { create_object, concat, sort_by_length_up, get_max_len } from "./common.js";
/*
from -> result[
res[score][id],
res[score][id],
]
to -> [id]
*/
/**
* Implementation based on Object[key] provides better suggestions
* capabilities and has less performance scaling issues on large indexes.
*
* @param arrays
* @param limit
* @param offset
* @param {boolean|Array=} suggest
* @param {boolean=} resolve
* @returns {Array}
*/
export function intersect(arrays, limit, offset, suggest) {
const length = arrays.length;
// todo remove
// if(length < 2){
// throw new Error("Not optimized intersect");
// }
let result = [],
size = 0,
check,
check_suggest,
check_new,
res_arr;
if (suggest) {
suggest = [];
}
// 1. a reversed order prioritize the order of words from a query
// because it ends with the first term.
// 2. process terms in reversed order often has advantage for
// the fast path "end reached".
// alternatively the results could be sorted by length up
//arrays.sort(sort_by_length_up);
// todo the outer loop should be the res array instead of term array,
// this isn't just simple because the intersection calculation needs to reflect this
//const maxlen = get_max_len(arrays);
for (let x = length - 1, found; 0 <= x; x--) {
//for(let x = 0, found; x < length; x++){
res_arr = arrays[x];
check_new = create_object();
found = !check;
// process relevance in forward order (direction is
// important for adding IDs during the last round)
for (let y = 0, ids; y < res_arr.length; y++) {
ids = res_arr[y];
if (!ids || !ids.length) continue;
for (let z = 0, id; z < ids.length; z++) {
id = ids[z];
// check exists starting from the 2nd slot
if (check) {
if (check[id]) {
// check if in last round
if (!x) {
//if(x === length - 1){
if (offset) {
offset--;
} else {
result[size++] = id;
if (size === limit) {
// fast path "end reached"
return result; /*resolve === false
? { result, suggest }
:*/
}
}
}
if (x || suggest) {
//if((x < length - 1) || suggest){
check_new[id] = 1;
}
found = /* tag? */ /* stringify */ /* stringify */ /* skip update: */ /* append: */ /* skip update: */ /* skip_update: */ /* skip deletion */!0 /*await rows.hasNext()*/ /*await rows.hasNext()*/ /*await rows.hasNext()*/;
}
if (suggest) {
if (!check_suggest[id]) {
check_suggest[id] = 1;
const arr = suggest[y] || (suggest[y] = []);
arr.push(id);
}
// OLD:
//
// check_idx = (check_suggest[id] || 0) + 1;
// check_suggest[id] = check_idx;
//
// // do not adding IDs which are already included in the result (saves one loop)
// // the first intersection match has the check index 2, so shift by -2
//
// if(check_idx < length){
//
// const tmp = suggest[check_idx - 2] || (suggest[check_idx - 2] = []);
// tmp[tmp.length] = id;
// }
}
} else {
// pre-fill in first round
check_new[id] = 1;
}
}
}
if (suggest) {
// re-use the first pre-filled check for suggestions
check || (check_suggest = check_new);
} else if (!found) {
return [];
}
check = check_new;
}
// return intermediate result
// if(resolve === false){
// return { result, suggest };
// }
if (suggest) {
// needs to iterate in reverse direction
for (let x = suggest.length - 1, ids, len; 0 <= x; x--) {
ids = suggest[x];
len = ids.length;
for (let y = 0, id; y < len; y++) {
id = ids[y];
if (!check[id]) {
if (offset) {
offset--;
} else {
result[size++] = id;
if (size === limit) {
// fast path "end reached"
return result;
}
}
check[id] = 1;
}
}
}
}
return result;
}
/**
* @param mandatory
* @param arrays
* @returns {Array}
*/
export function intersect_union(mandatory, arrays) {
const check = create_object(),
union = create_object(),
result = [];
for (let x = 0; x < mandatory.length; x++) {
check[mandatory[x]] = 1;
}
for (let x = 0, arr; x < arrays.length; x++) {
arr = arrays[x];
for (let y = 0, id; y < arr.length; y++) {
id = arr[y];
if (check[id]) {
if (!union[id]) {
union[id] = 1;
result.push(id);
}
}
}
}
return result;
}
/**
* Implementation based on Array.includes() provides better performance,
@@ -192,203 +419,4 @@ import { create_object, concat } from "./common.js";
// }
//
// return result;
// }
/**
* Implementation based on Object[key] provides better suggestions
* capabilities and has less performance scaling issues on large indexes.
*
* @param arrays
* @param limit
* @param offset
* @param {boolean|Array=} suggest
* @returns {Array}
*/
export function intersect(arrays, limit, offset, suggest) {
const length = arrays.length;
let result = [],
check,
check_suggest,
size = 0;
if (suggest) {
suggest = [];
}
// process terms in reversed order often has advantage for the fast path "end reached".
// also a reversed order prioritize the order of words from a query.
for (let x = length - 1; 0 <= x; x--) {
const word_arr = arrays[x],
word_arr_len = word_arr.length,
check_new = create_object();
let found = !check;
// process relevance in forward order (direction is
// important for adding IDs during the last round)
for (let y = 0; y < word_arr_len; y++) {
const arr = word_arr[y],
arr_len = arr.length;
if (arr_len) {
// loop through IDs
for (let z = 0, check_idx, id; z < arr_len; z++) {
id = arr[z];
if (check) {
if (check[id]) {
// check if in last round
if (!x) {
if (offset) {
offset--;
} else {
result[size++] = id;
if (size === limit) {
// fast path "end reached"
return result;
}
}
}
if (x || suggest) {
check_new[id] = 1;
}
found = /* append: */ /* skip update: */ /* skip_update: */!0;
}
if (suggest) {
check_idx = (check_suggest[id] || 0) + 1;
check_suggest[id] = check_idx;
// do not adding IDs which are already included in the result (saves one loop)
// the first intersection match has the check index 2, so shift by -2
if (check_idx < length) {
const tmp = suggest[check_idx - 2] || (suggest[check_idx - 2] = []);
tmp[tmp.length] = id;
}
}
} else {
// pre-fill in first round
check_new[id] = 1;
}
}
}
}
if (suggest) {
// re-use the first pre-filled check for suggestions
check || (check_suggest = check_new);
} else if (!found) {
return [];
}
check = check_new;
}
if (suggest) {
// needs to iterate in reverse direction
for (let x = suggest.length - 1, arr, len; 0 <= x; x--) {
arr = suggest[x];
len = arr.length;
for (let y = 0, id; y < len; y++) {
id = arr[y];
if (!check[id]) {
if (offset) {
offset--;
} else {
result[size++] = id;
if (size === limit) {
// fast path "end reached"
return result;
}
}
check[id] = 1;
}
}
}
}
return result;
}
/**
* @param mandatory
* @param arrays
* @returns {Array}
*/
export function intersect_union(mandatory, arrays) {
const check = create_object(),
union = create_object(),
result = [];
for (let x = 0; x < mandatory.length; x++) {
check[mandatory[x]] = 1;
}
for (let x = 0, arr; x < arrays.length; x++) {
arr = arrays[x];
for (let y = 0, id; y < arr.length; y++) {
id = arr[y];
if (check[id]) {
if (!union[id]) {
union[id] = 1;
result[result.length] = id;
}
}
}
}
return result;
}
// }