Squashed commit of the following:
commit ac82511ddd17d5472ae3cfea9bbad9754f5a4d62 Author: Simon Cambier <simon.cambier@protonmail.com> Date: Sat Oct 22 08:23:42 2022 +0200 Screw that cache, seriously. commit 8ba40d1be73daaaffea09e07bc56c339266db9b6 Author: Simon Cambier <simon.cambier@protonmail.com> Date: Fri Oct 21 22:36:48 2022 +0200 Stuff commit 27b8fd7dc809be9714a109d3a458eb1276a47e2e Author: Simon Cambier <simon.cambier@protonmail.com> Date: Fri Oct 21 22:22:20 2022 +0200 Moved files commit fb1349c914907e586e103ca54fb04b9ddd45ef5d Author: Simon Cambier <simon.cambier@protonmail.com> Date: Thu Oct 20 22:25:29 2022 +0200 Removed duplicate code commit e7371138e60cbe4155cfd4fb44e3ee1d2e3ee088 Author: Simon Cambier <simon.cambier@protonmail.com> Date: Thu Oct 20 21:50:09 2022 +0200 Moved a bunch of files commit 2ee1b2a0e799d4b41ab3a444d8cc44dfff5b5623 Author: Simon Cambier <simon.cambier@protonmail.com> Date: Thu Oct 20 21:32:21 2022 +0200 Removed useless code commit 76c530dfb9adbad1bbe9079de2330fe43a044249 Author: Simon Cambier <simon.cambier@protonmail.com> Date: Thu Oct 20 20:44:11 2022 +0200 Split file reading and indexing
This commit is contained in:
Simon Cambier
249
src/search/search.ts
Normal file
249
src/search/search.ts
Normal file
@@ -0,0 +1,249 @@
|
||||
import MiniSearch, {
|
||||
type AsPlainObject,
|
||||
type Options,
|
||||
type SearchResult,
|
||||
} from 'minisearch'
|
||||
import {
|
||||
chsRegex,
|
||||
type IndexedDocument,
|
||||
type ResultNote,
|
||||
type SearchMatch,
|
||||
SPACE_OR_PUNCTUATION,
|
||||
} from '../globals'
|
||||
import {
|
||||
removeDiacritics,
|
||||
stringsToRegex,
|
||||
stripMarkdownCharacters,
|
||||
} from '../tools/utils'
|
||||
import type { Query } from './query'
|
||||
import { settings } from '../settings'
|
||||
import { cacheManager } from '../cache-manager'
|
||||
|
||||
let minisearchInstance: MiniSearch<IndexedDocument>
|
||||
|
||||
const tokenize = (text: string): string[] => {
|
||||
const tokens = text.split(SPACE_OR_PUNCTUATION)
|
||||
const chsSegmenter = (app as any).plugins.plugins['cm-chs-patch']
|
||||
|
||||
if (chsSegmenter) {
|
||||
return tokens.flatMap(word =>
|
||||
chsRegex.test(word) ? chsSegmenter.cut(word) : [word]
|
||||
)
|
||||
} else return tokens
|
||||
}
|
||||
|
||||
const minisearchOptions: Options<IndexedDocument> = {
|
||||
tokenize,
|
||||
processTerm: (term: string) =>
|
||||
(settings.ignoreDiacritics ? removeDiacritics(term) : term).toLowerCase(),
|
||||
idField: 'path',
|
||||
fields: [
|
||||
'basename',
|
||||
'aliases',
|
||||
'content',
|
||||
'headings1',
|
||||
'headings2',
|
||||
'headings3',
|
||||
],
|
||||
storeFields: ['tags'],
|
||||
}
|
||||
|
||||
/**
|
||||
* Initializes the MiniSearch instance,
|
||||
* and adds all the notes to the index
|
||||
*/
|
||||
export async function initSearchEngine(): Promise<void> {
|
||||
// Default instance
|
||||
minisearchInstance = new MiniSearch(minisearchOptions)
|
||||
}
|
||||
|
||||
export async function initSearchEngineFromData(json: string): Promise<void> {
|
||||
try {
|
||||
minisearchInstance = MiniSearch.loadJSON(json, minisearchOptions)
|
||||
console.log('Omnisearch - MiniSearch index loaded from the file')
|
||||
} catch (e) {
|
||||
console.error('Omnisearch - Could not load MiniSearch index from json')
|
||||
console.error(e)
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Searches the index for the given query,
|
||||
* and returns an array of raw results
|
||||
* @param query
|
||||
* @returns
|
||||
*/
|
||||
async function search(query: Query): Promise<SearchResult[]> {
|
||||
if (!query.segmentsToStr()) return []
|
||||
|
||||
let results = minisearchInstance.search(query.segmentsToStr(), {
|
||||
prefix: true,
|
||||
fuzzy: term => (term.length > 4 ? 0.2 : false),
|
||||
combineWith: 'AND',
|
||||
boost: {
|
||||
basename: settings.weightBasename,
|
||||
aliases: settings.weightBasename,
|
||||
headings1: settings.weightH1,
|
||||
headings2: settings.weightH2,
|
||||
headings3: settings.weightH3,
|
||||
},
|
||||
})
|
||||
|
||||
// Downrank files that are in Obsidian's excluded list
|
||||
if (settings.respectExcluded) {
|
||||
results.forEach(result => {
|
||||
if (
|
||||
app.metadataCache.isUserIgnored &&
|
||||
app.metadataCache.isUserIgnored(result.id)
|
||||
) {
|
||||
result.score /= 3 // TODO: make this value configurable or toggleable?
|
||||
}
|
||||
})
|
||||
}
|
||||
|
||||
// If the search query contains quotes, filter out results that don't have the exact match
|
||||
const exactTerms = query.getExactTerms()
|
||||
if (exactTerms.length) {
|
||||
results = results.filter(r => {
|
||||
const title = cacheManager.getDocument(r.id)?.path.toLowerCase() ?? ''
|
||||
const content = stripMarkdownCharacters(
|
||||
cacheManager.getDocument(r.id)?.content ?? ''
|
||||
).toLowerCase()
|
||||
return exactTerms.every(q => content.includes(q) || title.includes(q))
|
||||
})
|
||||
}
|
||||
|
||||
// If the search query contains exclude terms, filter out results that have them
|
||||
const exclusions = query.exclusions
|
||||
if (exclusions.length) {
|
||||
results = results.filter(r => {
|
||||
const content = stripMarkdownCharacters(
|
||||
cacheManager.getDocument(r.id)?.content ?? ''
|
||||
).toLowerCase()
|
||||
return exclusions.every(q => !content.includes(q.value))
|
||||
})
|
||||
}
|
||||
return results
|
||||
}
|
||||
|
||||
/**
|
||||
* Parses a text against a regex, and returns the { string, offset } matches
|
||||
* @param text
|
||||
* @param reg
|
||||
* @returns
|
||||
*/
|
||||
export function getMatches(text: string, reg: RegExp): SearchMatch[] {
|
||||
let match: RegExpExecArray | null = null
|
||||
const matches: SearchMatch[] = []
|
||||
let count = 0 // TODO: FIXME: this is a hack to avoid infinite loops
|
||||
while ((match = reg.exec(text)) !== null) {
|
||||
if (++count > 100) break
|
||||
const m = match[0]
|
||||
if (m) matches.push({ match: m, offset: match.index })
|
||||
}
|
||||
return matches
|
||||
}
|
||||
|
||||
/**
|
||||
* Searches the index, and returns an array of ResultNote objects.
|
||||
* If we have the singleFile option set,
|
||||
* the array contains a single result from that file
|
||||
* @param query
|
||||
* @param options
|
||||
* @returns
|
||||
*/
|
||||
export async function getSuggestions(
|
||||
query: Query,
|
||||
options?: Partial<{ singleFilePath: string | null }>
|
||||
): Promise<ResultNote[]> {
|
||||
// Get the raw results
|
||||
let results = await search(query)
|
||||
if (!results.length) return []
|
||||
|
||||
// Extract tags from the query
|
||||
const tags = query.segments
|
||||
.filter(s => s.value.startsWith('#'))
|
||||
.map(s => s.value)
|
||||
|
||||
// Either keep the 50 first results,
|
||||
// or the one corresponding to `singleFile`
|
||||
if (options?.singleFilePath) {
|
||||
const result = results.find(r => r.id === options.singleFilePath)
|
||||
if (result) results = [result]
|
||||
else results = []
|
||||
} else {
|
||||
results = results.slice(0, 50)
|
||||
|
||||
// Put the results with tags on top
|
||||
for (const tag of tags) {
|
||||
for (const result of results) {
|
||||
if ((result.tags ?? []).includes(tag)) {
|
||||
result.score *= 100
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Map the raw results to get usable suggestions
|
||||
return results.map(result => {
|
||||
const note = cacheManager.getDocument(result.id)
|
||||
if (!note) {
|
||||
throw new Error(`Note "${result.id}" not indexed`)
|
||||
}
|
||||
|
||||
// Remove '#' from tags, for highlighting
|
||||
query.segments.forEach(s => {
|
||||
s.value = s.value.replace(/^#/, '')
|
||||
})
|
||||
// Clean search matches that match quoted expressions,
|
||||
// and inject those expressions instead
|
||||
const foundWords = [
|
||||
// Matching terms from the result,
|
||||
// do not necessarily match the query
|
||||
...Object.keys(result.match),
|
||||
|
||||
// // Matching terms from the query,
|
||||
// // but only if they stem from the result's matches
|
||||
// ...Object.keys(result.match).filter(w =>
|
||||
// query.segments.some(s => w.startsWith(s.value)),
|
||||
// ),
|
||||
|
||||
// Quoted expressions
|
||||
...query.segments.filter(s => s.exact).map(s => s.value),
|
||||
|
||||
// Tags, starting with #
|
||||
...tags,
|
||||
]
|
||||
|
||||
const matches = getMatches(note.content, stringsToRegex(foundWords))
|
||||
const resultNote: ResultNote = {
|
||||
score: result.score,
|
||||
foundWords,
|
||||
matches,
|
||||
...note,
|
||||
}
|
||||
return resultNote
|
||||
})
|
||||
}
|
||||
|
||||
// #region Read/write minisearch index
|
||||
|
||||
export function getMinisearchIndexJSON(): AsPlainObject {
|
||||
return minisearchInstance.toJSON()
|
||||
}
|
||||
|
||||
export async function addAllToMinisearch(
|
||||
documents: IndexedDocument[]
|
||||
): Promise<void> {
|
||||
await minisearchInstance.addAllAsync(documents)
|
||||
}
|
||||
|
||||
export function addSingleToMinisearch(document: IndexedDocument): void {
|
||||
minisearchInstance.add(document)
|
||||
}
|
||||
|
||||
export function removeFromMinisearch(document: IndexedDocument): void {
|
||||
minisearchInstance.remove(document)
|
||||
}
|
||||
|
||||
// #endregion
|
||||
Reference in New Issue
Block a user