This patch aims to support the execution of a specific query pattern without requiring a sort (and potentially some temporary files).
I can quote Ants Aasma <ants(dot)aasma(at)cybertec(dot)at> [1] > Think of the social media timeline, showing posts from > all followed accounts in timestamp order, returned in reasonably sized > batches. I tried this before as an access method, and now I am going the easy way by doing a plan transformation. Since skip-scan is implemented directly as a feature of nbtree access method [2], my best shot was to use the same approach and implement this feature as an access method [3]. I am glad the idea got some traction, but some questions related to how to cost it properly can be hard, and to be honest, that patch was uncomfortably large, this patch runs index scans through Merge Append, which as pointed out in previous discussions [4], produces the same results. Kind Regards, Alexandre --- [1] https://www.postgresql.org/message-id/CANwKhkNd85u%2B4joaKR3YHoDOQSMg5SmJmsYJGo-tMyW%3DXVXTew%40mail.gmail.com [2] https://github.com/postgres/postgres/commit/92fe23d93 [3] https://www.postgresql.org/message-id/flat/AS1PR02MB784695AFEC37179FFAF7EAE19A9DA%40AS1PR02MB7846.eurprd02.prod.outlook.com [4] https://www.postgresql.org/message-id/CANwKhkNd85u%2B4joaKR3YHoDOQSMg5SmJmsYJGo-tMyW%3DXVXTew%40mail.gmail.com
v1-0001-SKIP-MERGE-Implementation.patch
Description: Binary data
v1-0002-SKIP-MERGE-tests.patch
Description: Binary data
