Having (group filter)

having filters on the aggregated value, not the source rows — the SQL HAVING to a column filter’s WHERE. After the engine rolls up the tree, having keeps only the groups whose measures satisfy a safe formula-DSL predicate (e.g. "revenue > 1000"); nodes that fail are pruned, but the ancestor path to any survivor is kept so the hierarchy stays intact. It runs in the engine — the frontend does no math — and applies to the grid (row-tree) flow only.

This is best seen in a running app; the config is the whole feature:

Dash (Python)

from dash_tensor_grid import TensorGrid, build_grid

payload = build_grid(sales, {
    "rows": ["region", "country"],
    "measures": {"revenue": "sum", "qty": "sum"},
    "formats": {"revenue": "currency:USD"},
    "having": "revenue > 1000",     # keep only groups whose aggregated revenue exceeds 1000
})
TensorGrid(id="grid", row_data=payload["row_data"], column_defs=payload["column_defs"])

React

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

<TensorGrid
  data={data}
  rows={['region', 'country']}
  measures={{ revenue: 'sum', qty: 'sum' }}
  having="revenue > 1000"
/>;

Vue

<TensorGrid
  :data="data"
  :rows="['region', 'country']"
  :measures="{ revenue: 'sum', qty: 'sum' }"
  having="revenue > 1000"
/>

Vanilla JS

import { mountTensorGrid } from '@tensorgrid/vanilla';

mountTensorGrid(document.getElementById('grid'), {
  data,
  rows: ['region', 'country'],
  measures: { revenue: 'sum', qty: 'sum' },
  having: 'revenue > 1000',
});

Notes

• Post-aggregation. The predicate is evaluated against each node’s rolled-up measures — the same safe DSL as a calculated measure, so "revenue > cost * 1.5" works too. It never runs arbitrary code.
• Ancestors kept. A failing parent that still has a passing descendant is retained as a path; only genuinely empty branches disappear.
• Grid flow only. having is refused on a pivot (columns) config — the cross-tab has a different shape. In lazy / server mode it prunes the shipped root level until children are fetched.
• Contrast with a column range filter, which tests the source rows before aggregation.