Data vs spec¶

GridViewSpec is structure only. Row data never lives inside the spec.

Rules¶

In spec	Outside spec
Block ids, column ids, labels	Row dicts / ORM rows
Filter schema and current state	Queryset filtering in Python
Chart axes and series keys	Numeric series values
`GridViewTable.datasource.endpoint` URL	API response bodies
Renderer ids (`badge`, `money`, …)	Renderer implementation (host JS registry)

# Correct: rows passed at render time
html = render_grid_view_spec(spec, rows, host=host)

# Wrong: embedding row data in blocks (except inline demos)
GridViewTable(rows=({"id": 1},))  # only for small static fixtures

Page loader pattern¶

request → page_data loader → rows + GridViewSpec → view → template

One loader should feed HTML render and export builders. See Page export pattern.

Dynamic columns¶

GridViewTable.column_source fetches column definitions at runtime (e.g. per-region price tiers), not row payloads. Rows still come from the host row API or server render path.

Validation boundary¶

validate_spec(spec) checks structure, id uniqueness, toolbar XOR search, and policy gates. It does not validate business data in rows.