Fuzzy-Net is a lexical search index that stores each string as a document bit and finds a query by projecting hashed tokens onto those bits, so a misspelled word can still match when enough of its bytes remain.
Deno:
deno add npm:@neabyte/fuzzy-netnpm:
npm install @neabyte/fuzzy-netCDN (jsDelivr/esm.sh):
<script type="module">
import FuzzyNet from 'https://cdn.jsdelivr.net/npm/@neabyte/fuzzy-net/dist/index.mjs'
</script>Or via esm.sh:
<script type="module">
import FuzzyNet from 'https://esm.sh/@neabyte/fuzzy-net'
</script>import FuzzyNet from '@neabyte/fuzzy-net'
// Build index from five product names
const index = new FuzzyNet([
'red ceramic coffee mug',
'blue ceramic tea cup',
'linen kitchen towel',
'oak dining chair',
'wool winter scarf'
])
console.log(index.search('ceramic mug', 2))
// [
// { value: 'red ceramic coffee mug', score: 16 },
// { value: 'blue ceramic tea cup', score: 12 }
// ]
console.log(index.search('cermaic', 1))
// [{ value: 'red ceramic coffee mug', score: 8 }]
console.log(index.search('xoffee', 1))
// [{ value: 'red ceramic coffee mug', score: 12 }]limit defaults to 10, a limit of 0 or below returns [], and a query with no byte features also returns [].
| Parameter | Type | Description |
|---|---|---|
values |
readonly string[] |
Strings to index. Position in the array is the document id. |
The table is built once, and search does not change the indexed strings.
| Parameter | Type | Description |
|---|---|---|
query |
string |
Query text. A token ends on a byte whose top two bits are 00. |
limit |
number |
Optional. How many results to return. Default 10. |
Returns: FuzzyResult[], highest score first.
type FuzzyIndex = {
search: (query: string, limit?: number) => FuzzyResult[]
}
type FuzzyResult = {
value: string
score: number
}value is the original indexed string and score is the overlap stage mask, so when two scores are equal the earlier one in the scan stays ahead.
npm run builddeno task checkdeno task testThis repository is licensed under Apache 2.0.