Skip to content

Counting rows

POST /api/v1/datasets/{name}/count
Content-Type: application/json

Same predicate shape as /query, but only the predicates field is read — columns, page, page_size, limit are all ignored. An empty body (or {}) counts every row.

Response

{ "count": 7728394 }

Examples

# Total row count.
curl -s -X POST http://localhost:8080/api/v1/datasets/accidents/count \
     -H 'content-type: application/json' -d '{}'
# → {"count":7728394}

# Filtered count — same operators as /query.
curl -s -X POST http://localhost:8080/api/v1/datasets/accidents/count \
     -H 'content-type: application/json' \
     -d '{
       "predicates": [
         { "col": "state",    "op": "in",  "val": ["CA","TX"] },
         { "col": "severity", "op": "gte", "val": 3 }
       ]
     }'
# → {"count":418217}

Execution paths

DataFusion backend:

  1. No predicates on a materialised dataset → resident num_rows(). No scan.
  2. All eq / in on indexed columns → length of the merged row-id list from the equality index. No scan.
  3. Anything else, or lazy datasets → SELECT COUNT(*) FROM … WHERE … through DataFusion / DuckDB.

This makes /count ideal for "results so far" UI badges that update as the user adjusts filters.