Faceted filters

A set filter lists each distinct value with a live count and narrows as you pick — and the facets are faceted: selecting a value in one column re-narrows the counts in the others, but a column never narrows its own list. In client mode the counts are derived from the loaded rows; in server / infinite mode the engine supplies them via facet_counts (keyed into column_facets), because the browser only holds a window and can’t count the whole frame. Either way the frontend does zero math over them.

This one is best seen in a running app. The code below wires it in each framework.

Dash (Python)

from dash_tensor_grid import GridDataEngine, TensorGrid

engine = GridDataEngine(df)             # a large flat frame
FACET_COLS = ["region", "status", "amount"]

def facets(filters):
    # Each column self-excluded (faceted search); recompute only when the filter model changes.
    return {c: engine.facet_counts(c, filters=filters) for c in FACET_COLS}

grid = TensorGrid(
    id="grid",
    column_defs=[
        {"accessorKey": "region", "header": "Region", "type": "string", "filterType": "set"},
        {"accessorKey": "status", "header": "Status", "type": "string", "filterType": "select"},
        {"accessorKey": "amount", "header": "Amount", "type": "number", "filterType": "range"},
    ],
    column_facets=facets({}),           # seed the initial (unfiltered) facets
    row_model="infinite",
    row_count=len(df),
)

# In the rows_request callback, return refreshed column_facets when `filters` changes.

React

import { TensorGrid } from '@tensorgrid/react';

// `rootFilter="set"` derives the value list + counts client-side (core facetCounts).
<TensorGrid
  data={data}
  rows={['region', 'country']}
  measures={{ revenue: 'sum' }}
  enableColumnFilters
  rootFilter="set"
/>;

Vue

<TensorGrid
  :data="data"
  :rows="['region', 'country']"
  :measures="{ revenue: 'sum' }"
  enable-column-filters
  root-filter="set"
/>

Vanilla JS

import { facetCounts } from '@tensorgrid/core';

// The set-filter value list + counts for a column, faceted by the other active filters.
const region = facetCounts(data, 'region', { kind: 'set' });
// -> { type: 'set', column: 'region', values: [{ value: 'EMEA', count: 4 }, …], total, truncated }

// Render your own checkbox list from `region.values`, then drive the grid with setFilter.

Notes

• Faceted search. The facet for a column reflects every other active filter but not its own, so picking a value narrows the sibling columns while the column keeps every value still reachable.
• facet_counts / facetCounts return either a set result (values: [{value, count}], total, truncated) or a range result ({min, max, count}); a set list is capped (top ~200) with truncated flagging that more distinct values exist.
• Server mode. Feed column_facets from the engine when row_model="infinite" — recompute only when the filter model changes, not on scroll. See filtering for the control types and infinite row model for the windowed data path.