Query

HeapQL is a read-only SQL-like language plus object navigation commands. Use highlighting, autocomplete, help, result tables, and history to investigate the same analyzed dump.

SELECT object_id, class_name, shallow_size, retained_size
FROM instances
WHERE retained_size > 1MB
ORDER BY retained_size DESC
LIMIT 20

Tables and counting

Supported tables: instances, class_histogram, dominator_tree, leak_suspects. FROM objects is not valid.

SELECT COUNT(*) AS class_rows, SUM(instance_count) AS instances
FROM class_histogram

COUNT counts rows; in the histogram it counts classes, not instances. SUM(instance_count) totals recorded instances. Retained sizes can overlap: SUM(retained_size) is not whole-heap live usage.

LIKE '%Cache%' matches names, not every HashMap used as a cache. Inspect semantics and actual owners.

Results and navigation

Use current host page controls when supplied. The paged API separates page size from SQL LIMIT; not every released renderer exposes identical Prev/Next buttons. Query history can repopulate/run saved text according to host controls—inspect it before execution. History persistence differs by host; avoid storing secrets in queries.

Object navigation commands :path, :refs, :children, and :info take an ID from this dump. Unknown objects/errors should remain visible without replacing valid results with fabricated rows.

Cost and limitations

LIMIT bounds returned rows, not necessarily scanning, sorting, join materialization, or graph memory. Use filters, bounded output, and a safe dump first. Query errors may differ in wording by host. Server Retry must restore query routing.

See the complete reference for JOINs, GROUP BY, aggregates, subqueries, binary size literals, special dominator-table semantics, and unsupported syntax.