Patch to show individual statement latencies in pgbench output
in [PgSql]
|
Prev: [HACKERS] Review of Synchronous Replication patches
Next: Proposal / proof of concept: Triggers on VIEWs
From: Florian Pflug on 4 Aug 2010 07:58 On Aug3, 2010, at 21:16 , Greg Smith wrote: >> That was a leftover of the trimming and comment skipping logic, which my patch moves to process_command. > > I think there's still a trimming error here--line 195 of the new patch is now removing the declaration of "i" just before it sets it to zero? Hm, I think it's just the diff thats miss-leading there. It correctly marks the "int i" line as "removed" with a "-", but for some reason marks the "i = 0" line (and its successors) with a "!", although they're removed too, and not modified. > On the coding standard side, I noticed all your for loops are missing a space between the for and the (; that should get fixed. Fixed > > Finally, now that the rest of the patch is looking in good shape and is something I think is worth considering to commit, it's time to work on the documentation SGML. I've added the "-r" option to the list of pgbench options in pgbench.sgml and also added a short section that shows how the output looks like, similar to how things are done for the "-l" option. > Also: when generating multiple versions of a patch like this, standard practice is to add something like "-vX" to the naming, so those of us trying to review can keep them straight. Will do from now on. Updated patch is attached. I've also pushed this as branch "pgbench_statementlatency" to git://github.com/fgp/postgres.git best regards, Florian Pflug |