Fuzzy Search Filter

Use the Fuzzy Search filter for names, titles, and other text where misspellings or transposed characters should still find a result. It applies deterministic Damerau-Levenshtein similarity and previews the best matching unique values.

Source code

grid.plugins = [AdvanceFilterPlugin];
grid.columns = [
  { prop: 'customer', name: 'Customer', filter: ['string', FILTER_FUZZY] },
];
grid.filter = true;

return [
  { customer: 'Jon Smith' },
  { customer: 'John Smith' },
  { customer: 'Jon Smythe' },
  { customer: 'Jane Stone' },
];

Quick start

1
import { AdvanceFilterPlugin, FILTER_FUZZY } from '@revolist/revogrid-pro';
2
import '@revolist/revogrid-pro/dist/revogrid-pro.css';
3

4
grid.plugins = [AdvanceFilterPlugin];
5
grid.columns = [
6
  { prop: 'customer', name: 'Customer', filter: ['string', FILTER_FUZZY] },
7
];
8
grid.filter = true;

The popup lets users enter a term and adjust its similarity threshold. Keeping 'string' also exposes exact text operators.

Operator and state

Body ID	Saved operator	Value shape
`fuzzy`	`fuzzy`	`{ term: string, threshold: number }`

The term is trimmed. The threshold is clamped to 0..1 and defaults to 0.55 when the saved value is missing or invalid. Matching is case-insensitive; exact and contained text score highly, while word-level similarity also contributes.

Restore a fuzzy filter

1
grid.filter = {
2
  multiFilterItems: {
3
    customer: [
4
      {
5
        id: 1,
6
        type: 'fuzzy',
7
        value: { term: 'Jon Smth', threshold: 0.7 },
8
        relation: 'and',
9
      },
10
    ],
11
  },
12
};

An empty or whitespace-only term removes the condition.

Key behavior

The preview ranks at most five matches by default and considers at most 2,000 unique candidates.
The predicate and preview use the same deterministic score.
null, undefined, objects, and functions do not match; other scalar values are converted to text.
The preview reads parsed unique column values. With remote or partially loaded rows, provide complete structured-filter aggregates when the preview must represent the full dataset.

The built-in has no per-column option object. Store the desired threshold in the condition value; create or replace a structured registration only when you need a different editor experience.

Common mistakes

Do not use a percentage such as 70; thresholds use the 0..1 range, so use 0.7.
Do not expect the preview to scan an unlimited high-cardinality column; its candidate work is deliberately bounded.
Do not use fuzzy search for exact identifiers. Use the Token List filter instead.