Calculated measures

A calculated measure is derived after aggregation — a formula over the values a group already rolled up (margin = (revenue - cost) / revenue), not a fresh aggregation of the source rows. That distinction matters: revenue and cost are summed per group by the engine, then the ratio is computed on those subtotals — never a mean-of-means. The engine does the math; the frontend does zero.

Dash (Python)

from dash import Dash
from dash_tensor_grid import TensorGrid, build_grid

payload = build_grid(sales, {
    "rows": ["region"],
    "measures": {"revenue": "sum", "cost": "sum"},          # true aggregations
    "calculated": {"margin": "(revenue - cost) / revenue"},  # post-aggregation
    "formats": {"revenue": "currency:USD", "cost": "currency:USD", "margin": "percentage:1"},
    "grand_total": True,
})

app = Dash(__name__)
app.layout = TensorGrid(
    id="grid",
    row_data=payload["row_data"],
    column_defs=payload["column_defs"],
)

React

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

<TensorGrid
  data={data}
  rows={['region']}
  measures={{ revenue: 'sum', cost: 'sum' }}
  calculated={{ margin: '(revenue - cost) / revenue' }}
  formats={{ revenue: '$0,0', cost: '$0,0', margin: '0.0%' }}
  grandTotal
/>;

Vue

<TensorGrid
  :data="data"
  :rows="['region']"
  :measures="{ revenue: 'sum', cost: 'sum' }"
  :calculated="{ margin: '(revenue - cost) / revenue' }"
  :formats="{ revenue: '$0,0', cost: '$0,0', margin: '0.0%' }"
  grand-total
/>

Vanilla JS

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

mountTensorGrid(document.getElementById('grid'), {
  data,
  rows: ['region'],
  measures: { revenue: 'sum', cost: 'sum' },
  calculated: { margin: '(revenue - cost) / revenue' },
  formats: { revenue: '$0,0', cost: '$0,0', margin: '0.0%' },
  grandTotal: true,
});

Notes

• Calculated vs. aggregation. A measures entry aggregates raw rows (sum, mean, …); a calculated entry runs a formula over the already-aggregated measures of each group. Keep ratios, margins, and weighted averages in calculated — summing a ratio column would be wrong.
• The formula reads sibling measures by name (revenue, cost), plus context values like _total_ and _parent_ where available. Full operator / function reference: formula DSL.
• Same string, everywhere. The expression parses identically in the Polars engine (Dash) and the parity-verified @tensorgrid/core port (adapters), so a calculated measure renders the same value in every framework.
• In Dash you can also tag the form explicitly — {"dsl": "..."} (safe), {"sql": "..."} (a true aggregation), or {"python": "pkg:fn"} (runs code, needs allow_code=True). See the config schema.
• Just rolling numbers up? Use a plain aggregation instead.