]> granicus.if.org Git - postgresql/commit
tableam: bitmap table scan.
authorAndres Freund <andres@anarazel.de>
Mon, 1 Apr 2019 00:51:49 +0000 (17:51 -0700)
committerAndres Freund <andres@anarazel.de>
Mon, 1 Apr 2019 01:37:57 +0000 (18:37 -0700)
commitbfbcad478f05794e5e7ea1339e62a1d258c99b6c
treeeec25ff67307f13fa7476d916ed04dc53da2c9b0
parent73c954d24896aeb05de0f81d75e891a858e439e9
tableam: bitmap table scan.

This moves bitmap heap scan support to below an optional tableam
callback. It's optional as the whole concept of bitmap heapscans is
fairly block specific.

This basically moves the work previously done in bitgetpage() into the
new scan_bitmap_next_block callback, and the direct poking into the
buffer done in BitmapHeapNext() into the new scan_bitmap_next_tuple()
callback.

The abstraction is currently somewhat leaky because
nodeBitmapHeapscan.c's prefetching and visibilitymap based logic
remains - it's likely that we'll later have to move more into the
AM. But it's not trivial to do so without introducing a significant
amount of code duplication between the AMs, so that's a project for
later.

Note that now nodeBitmapHeapscan.c and the associated node types are a
bit misnamed. But it's not clear whether renaming wouldn't be a cure
worse than the disease. Either way, that'd be best done in a separate
commit.

Author: Andres Freund
Reviewed-By: Robert Haas (in an older version)
Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
src/backend/access/heap/heapam_handler.c
src/backend/access/table/tableamapi.c
src/backend/executor/nodeBitmapHeapscan.c
src/backend/optimizer/util/plancat.c
src/include/access/tableam.h
src/include/nodes/execnodes.h
src/include/nodes/tidbitmap.h