CSV export

Because grid cells are aggregates, “export” means three different things — and the engine serializes all of them so the CSV always matches what the math produced. Export the raw source rows behind a selection, the aggregated roll-up tree exactly as shown (filter + sort + top-N applied), or a pivot cross-tab flattened to columns. Values ship as raw numerics (machine-readable) by default, or formatted ($420.00, 45%) to match the display. The built-in Export button below emits the selection-or-all CSV.

from dash import dcc, html, Input, Output, callback
from dash_tensor_grid import build_grid
from dash_tensor_grid.engine import GridDataEngine

engine = GridDataEngine(sales)
payload = build_grid(sales, {"rows": ["region", "country"], "measures": {"revenue": "sum"}})

layout = html.Div([
    # ...TensorGrid(...),
    html.Button("Export", id="export"),
    dcc.Download(id="download"),
])

@callback(Output("download", "data"), Input("export", "n_clicks"), prevent_initial_call=True)
def export(_):
    csv = engine.to_csv()                                   # SOURCE rows (raw frame)
    # csv = engine.tree_to_csv(payload["row_data"],         # AGGREGATED tree as shown
    #                          ["region", "country"], ["revenue"])
    # pv = build_grid(sales, {"rows": ["region"], "columns": ["category"],
    #                         "measures": {"revenue": "sum"}})
    # csv = engine.pivot_to_csv(pv, ["region"])             # PIVOT cross-tab, flattened
    return dcc.send_string(csv, "export.csv")

<TensorGrid
  data={data}
  rows={['region', 'country']}
  measures={{ revenue: 'sum' }}
  showExportButton
  exportMode="aggregated"       // 'source' (raw underlying rows) | 'aggregated' (the roll-up as shown)
  exportFormatted               // emit '$420.00' etc. instead of raw numerics
  onExport={(csv) => downloadCsv(csv)}
/>;

// …or build the CSV yourself from the core helpers:
// import { toCsv, treeToCsv, pivotToCsv } from '@tensorgrid/core';

<TensorGrid
  :data="data"
  :rows="['region', 'country']"
  :measures="{ revenue: 'sum' }"
  show-export-button
  export-mode="aggregated"
  export-formatted
  :on-export="downloadCsv"
/>

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

const grid = mountTensorGrid(el, {
  data,
  rows: ['region', 'country'],
  measures: { revenue: 'sum' },
  showExportButton: true,
  exportMode: 'aggregated',            // 'source' | 'aggregated'
  exportFormatted: true,
  onExport: (csv) => downloadCsv(csv),
});

grid.exportCsv();                       // SOURCE rows (selection-or-all), on demand
grid.aggregatedToCsv({ format: true }); // the AGGREGATED roll-up, formatted

// PIVOT cross-tab: the pivot handle flattens the CURRENT cross-tab to CSV.
const pivot = mountTensorPivot(el, {
  data, rows: ['region'], cols: ['category'], measures: { revenue: 'sum' },
});
pivot.toCsv();

Options

Source vs aggregated vs pivot. exportMode: 'source' (default) emits the raw underlying rows under the selection, or all rows if none is selected (core rowsForSelection + toCsv, engine to_csv). 'aggregated' emits the roll-up tree as shown — filter, sort, and top-N applied (core treeToCsv, engine tree_to_csv). A pivot flattens its cross-tab to one column per row dimension plus one per measure leaf (core pivotToCsv, engine pivot_to_csv).
exportFormatted — with exportMode: 'aggregated', render each measure through its formatStr ($420.00, 45%) instead of raw numerics. Ignored for 'source' export.
The engine owns serialization. The JS toCsv / treeToCsv / pivotToCsv are parity-verified ports of Polars write_csv (RFC-4180 quoting, null → empty field), so a Dash export and a React/Vue/Vanilla export of the same view are byte-identical.
Combine with aggregation or a pivot to decide what the exported view contains, and turn on exportFormatted when the CSV should read like the on-screen table.