vcswatch reports that
this package seems to have new commits in its VCS but has
not yet updated debian/changelog. You should consider updating
the Debian changelog and uploading this new version into the archive.
Here are the relevant commit logs:
commit f33a3ac3b82eeb37882d7613a794eabf42985e8a
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Thu Jan 9 16:39:54 2020 +0100
Fix compatibility with version 9.4
- don't use named arguments for plpgsql function call
- use json rather jsonb
commit f4410afca02837bb9f1586970b83334b901f8f7c
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Thu Jan 9 16:23:05 2020 +0100
Fix mutliple problems with regression tests.
- the ^@ operator does not exist in all supported version, so use instead the
"text ~ text" non-commutable operator
- first tests could fail is used against an instance with preexisting data in
pg_qualstats' shmem
- a test would give different results whether pg_stat_statements is installed
or not
commit 1fc605faaab83e4e0d61bdafae7de674d8abb210
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Thu Jan 9 15:58:48 2020 +0100
Fix bug in index advisor when all qual can be optimized
commit 0732013223040898f326bdf754b25771c015a510
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Thu Jan 9 15:45:53 2020 +0100
Return empty array rather than null in index advisor.
Per report from github user GrandFelix. This allows to use
jsonb_array_elements() whether the advisor returned indexes/unoptimized quals
or not.
commit 1c7aa0e7f368e80b518950d9d1ad5e504f6ceeaa
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Thu Jan 9 14:48:40 2020 +0100
Fix index advisor for older versions
Reported by github user GrandFelix.
Older postgres versions don't have the behavior for the || operator with arrays
and cast from jsonb to bigint.
There are still some broken regression tests with older versions that will be
fixed in later commits.
commit 191640c3a818fbb2bb50f76fb209abc128d30cf8
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Tue Dec 3 15:11:27 2019 +0100
Update changelog
commit c08a337f56e75280dd1bc891b3415a00739a369b
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Fri Dec 13 16:04:26 2019 +0100
Fix (unique)qualid processing.
(unique)qualid should only consider AND-ed quals, but this has been broken for
a long time, as the values were saved in the entry before being reset in case
of OR-ed quals.
Add some basic regression tests to make sure this doesn't get broken again.
commit 83f16784a7c27bb966012adf7a5c9b648b8afcae
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Fri Dec 13 15:40:34 2019 +0100
Run pgindent
commit 940f8a7646498dc151af6497b5e194dfe26ef1e0
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Tue Dec 3 14:48:50 2019 +0100
Add an index advisor feature.
This uses the same approach as PoWA wizard feature, but can be used with only
pg_qualstats and SQL. As a result, the suggestion can be less good, as it's
only considering quals currently present in shared memory.
commit 6619c1090bcebf3d3642bec6e9cb504859e2142d
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Tue Dec 3 14:29:12 2019 +0100
Remove outdated index suggestion queries.
Those queries are too naive, and will be replaced by a better approach in a
following commit.
commit eb7d5826e5794afbe281e84676872b65aea402fc
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Tue Dec 3 14:25:55 2019 +0100
Add missing fields description in the documentation.
commit 5ef44ed7bae82d0730c634bb919076de6a5928f5
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Fri Dec 13 11:06:22 2019 +0100
Handle statistics on selectivity estimation errors.
Per idea from Oleg Bartunov.
We now compute the selectivity estimation error, but in a ratio and raw number
of rows, per qual, and keep track of the minimum, maximum, mean and standard
deviation for both values.
Also add two new GUCS, to store only the quals that have at least a certain
selectivity estimation error ratio or number of rows.
commit 8d41caf0a410f1bea66d29654984140cf2487087
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Fri Dec 13 11:07:56 2019 +0100
Refactor pgqs_process_opexpr
Previous code wasn't terribly readable, had multiple inconsistencies and failed
to consider some OpExpr. The changes are:
- we now compute the various qual identifier the same way we store the qual,
based on the possibly commuted OpExpr
- we now also track OpExpr of the form Constant operator Val where the operator
doesn't have a commutator
- we now correctly discard quals not referencing a relation
Some regression tests are added to cover those points.
commit 87b452c9d598d74baefafb01b8801eeb7581b507
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Wed Nov 27 20:15:45 2019 +0100
Minor refactoring with some more comments.
commit 25b0f7850356bce8907a504a135bc578a7d94824
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Wed Nov 27 17:31:56 2019 +0100
Start working on 2.0.0dev
commit 97cbc09698925eeac8b62ed9d133ec19c06db112
Author: Julien Rouhaud <julien.rouhaud@free.fr>
Date: Sun Dec 29 15:51:42 2019 +0100
Don't expose recorded constants to unauthorized users.
![[bug history graph]](https://qa.debian.org/data/bts/graphs/p/pg-qualstats.png){kind=link}