Build your personal knowledge base with Trilium Notes
TypeScript electronelectron-appknowledge-baseknowledge-graphknowledge-managementknowledge-management-graphlocal-firstnote-managmentnote-takernote-takingnotebooknotesnotes-apppersonal-knowledge-basepersonal-wikiscriptableself-hostedself-hostingwiki
perf(search): normalize a comparator's compared value once, not per note
Elian Doranadd0f86cbbaf
diff --git a/packages/trilium-core/src/services/search/services/build_comparator.ts b/packages/trilium-core/src/services/search/services/build_comparator.tsindex b36ec36034..5603a02c2c 100644--- a/packages/trilium-core/src/services/search/services/build_comparator.ts+++ b/packages/trilium-core/src/services/search/services/build_comparator.ts@@ -17,26 +17,36 @@ const stringComparators: Record<string, Comparator<string>> = {// ignoring case and diacritics. This is the documented label-equality semantics, so// #capital=Vienna matches "vienna" but not "Vienna Austria". Word and phrase matching lives// in the internal "word=" operator.- "=": (comparedValue) => (val) => normalizeSearchText(val) === normalizeSearchText(comparedValue),- "!=": (comparedValue) => (val) => normalizeSearchText(val) !== normalizeSearchText(comparedValue),+ "=": (comparedValue) => {+ const normalizedCompared = normalizeSearchText(comparedValue);+ return (val) => normalizeSearchText(val) === normalizedCompared;+ },+ "!=": (comparedValue) => {+ const normalizedCompared = normalizeSearchText(comparedValue);+ return (val) => normalizeSearchText(val) !== normalizedCompared;+ },// Internal operator (not user-typable): word/phrase match used by the leading-"="// fulltext title comparison. Punctuation-aware via tokenizeIntoWords/stripWordPunctuation.- "word=": (comparedValue) => (val) => {- if (!val) return false;-- const normalizedVal = normalizeSearchText(val);+ "word=": (comparedValue) => {const normalizedCompared = normalizeSearchText(comparedValue);+ // A compared value with spaces is a multi-word phrase, matched as a substring.+ const isPhrase = normalizedCompared.includes(" ");+ const comparedWord = stripWordPunctuation(normalizedCompared);++ return (val) => {+ if (!val) return false;- // If comparedValue has spaces, it's a multi-word phrase- // Check for substring match (consecutive phrase)- if (normalizedCompared.includes(" ")) {- return normalizedVal.includes(normalizedCompared);- }+ const normalizedVal = normalizeSearchText(val);- // For single word, tokenize into punctuation-stripped words and check for- // an exact word match, so a value like "(Books)" matches the token "books".- const words = tokenizeIntoWords(normalizedVal);- return words.some(word => word === stripWordPunctuation(normalizedCompared));+ if (isPhrase) {+ return normalizedVal.includes(normalizedCompared);+ }++ // For single word, tokenize into punctuation-stripped words and check for+ // an exact word match, so a value like "(Books)" matches the token "books".+ const words = tokenizeIntoWords(normalizedVal);+ return words.some(word => word === comparedWord);+ };},">": (comparedValue) => (val) => val > comparedValue,">=": (comparedValue) => (val) => val >= comparedValue,@@ -46,45 +56,53 @@ const stringComparators: Record<string, Comparator<string>> = {"=*": (comparedValue) => (val) => !!val && val.startsWith(comparedValue),"*=*": (comparedValue) => (val) => !!val && val.includes(comparedValue),"%=": (comparedValue) => (val) => !!val && !!getRegex(comparedValue).test(val),- "~=": (comparedValue) => (val) => {- if (!val || !comparedValue) return false;-+ "~=": (comparedValue) => {// Validate minimum length for fuzzy search to prevent false positives- if (comparedValue.length < FUZZY_SEARCH_CONFIG.MIN_FUZZY_TOKEN_LENGTH) {- return val.includes(comparedValue);- }-- const normalizedVal = normalizeSearchText(val);+ const tooShortForFuzzy = comparedValue.length < FUZZY_SEARCH_CONFIG.MIN_FUZZY_TOKEN_LENGTH;const normalizedCompared = normalizeSearchText(comparedValue);-- // First try exact substring match- if (normalizedVal.includes(normalizedCompared)) {- return true;- }-- // Then try fuzzy word matching over the tokenized (punctuation-stripped) value- const words = tokenizeIntoWords(normalizedVal);- return words.some(word => fuzzyMatchWord(normalizedCompared, word));++ return (val) => {+ if (!val || !comparedValue) return false;++ if (tooShortForFuzzy) {+ return val.includes(comparedValue);+ }++ const normalizedVal = normalizeSearchText(val);++ // First try exact substring match+ if (normalizedVal.includes(normalizedCompared)) {+ return true;+ }++ // Then try fuzzy word matching over the tokenized (punctuation-stripped) value+ const words = tokenizeIntoWords(normalizedVal);+ return words.some(word => fuzzyMatchWord(normalizedCompared, word));+ };},- "~*": (comparedValue) => (val) => {- if (!val || !comparedValue) return false;-+ "~*": (comparedValue) => {// Validate minimum length for fuzzy search- if (comparedValue.length < FUZZY_SEARCH_CONFIG.MIN_FUZZY_TOKEN_LENGTH) {- return val.includes(comparedValue);- }-- const normalizedVal = normalizeSearchText(val);+ const tooShortForFuzzy = comparedValue.length < FUZZY_SEARCH_CONFIG.MIN_FUZZY_TOKEN_LENGTH;const normalizedCompared = normalizeSearchText(comparedValue);- // "~*" is fuzzy CONTAINS: first try a plain substring (fragment) match, so- // a fragment like "progr" matches "programming" (mirrors the ~= fallback).- // Then fall back to fuzzy matching for typos that are not substrings.- if (normalizedVal.includes(normalizedCompared)) {- return true;- }+ return (val) => {+ if (!val || !comparedValue) return false;++ if (tooShortForFuzzy) {+ return val.includes(comparedValue);+ }++ const normalizedVal = normalizeSearchText(val);++ // "~*" is fuzzy CONTAINS: first try a plain substring (fragment) match, so+ // a fragment like "progr" matches "programming" (mirrors the ~= fallback).+ // Then fall back to fuzzy matching for typos that are not substrings.+ if (normalizedVal.includes(normalizedCompared)) {+ return true;+ }- return fuzzyMatchWord(normalizedCompared, normalizedVal);+ return fuzzyMatchWord(normalizedCompared, normalizedVal);+ };}};