Group by

A group-by step (builder label Group by + aggregate) collapses input rows to one output feature per distinct combination of group keys. Geometry for each group is always PostGIS ST_Union of the group's input geometries. Attributes are the group keys plus your declared aggregates.

Use after a spatial-join, after a filter, or directly on a source layer when you want a keyed summary. If you only need one union of the whole layer with no keys and no aggregates, use the plain Dissolve step instead.

Inputs

• Source layer.
• Group keys (groupBy): zero to eight field names. Distinct combinations define the groups. An empty key list collapses the whole input to a single feature (union + aggregates only).
• Aggregates (aggs): up to 16 entries, each (op, field?, outputName): - count — row count per group (field unused). - sum / avg / min / max — numeric field. - first — first value of a field in the group (no order-by control).

Not options: last, collect_into_array, or a geometry strategy picker (Drop / Bounding box / Convex hull / First / Dissolve). Output geometry is always the union of each group.

Output

One row per distinct group-key combination (or one row when groupBy is empty):

• Geometry: ST_Union of the group's non-null input geometries (often a multipolygon when parts do not touch).
• Properties: each group key (original name) plus each aggregate under its outputName.

How to use it on its own

1. New-item wizard → Derived layer.
2. Pick a Source.
3. Add a Group by step.
4. Pick group keys (or leave empty for a single overall rollup).
5. Add each aggregate (op, field when required, output name).
6. Save.

There is no separate geometry-strategy control.

From a data layer Analyze tab

On a data layer's Analyze tab, Group by for this layer opens one optional group key (or whole layer), one aggregate (count, sum, or avg on a numeric field), and an output name, then creates a private derived layer with a single aggregate step (same PostGIS ST_Union path as above). Multi-key and multi-agg authoring stay in Create → Derived layer. Expect a server job, not an instant browser result.

Examples

• Hydrant count per parcel id (after a spatial join that brought parcel_id onto hydrant rows, or on parcels that already carry a count). Group key: parcel_id. Aggregate: count → hydrant_count. Geometry: union of features in each parcel group.
• Average inspection score per neighborhood. Group key: neighborhood. Aggregate: avg(score) → avg_score.
• Single overall count. Empty group keys. Aggregate: count → n. Same shape idea as dissolve, but you keep an aggregate attribute instead of dropping properties.

Notes

• NULL is a group. Rows with NULL in a group key are grouped together. Filter them out first if you do not want that.
• Large unions are slow. Dissolving a huge group via ST_Union can take minutes; filter first when you can.
• Cardinality drops. Output row count is the number of distinct key combinations (or 1 when keys are empty).
• Keyed dissolve lives here. The no-param Dissolve step unions everything and drops attributes. For one polygon per attribute value plus counts or sums, use this step.