]>
git.ipfire.org Git - thirdparty/psycopg.git/log
Denis Laxalde [Sun, 10 Oct 2021 09:27:13 +0000 (11:27 +0200)]
Rework PGconn.pipeline_status
We make this method a property, that returns an integer value, as do
other similar properties/methods on PGconn.
Then, for libpq version older than 14.0, we make this property always
return 0 (PQ_PIPELINE_OFF).
Denis Laxalde [Sun, 10 Oct 2021 09:22:14 +0000 (11:22 +0200)]
Define PGRES_PIPELINE_* ExecStatusType values as macros for libpq < 14.0
Daniele Varrazzo [Fri, 8 Oct 2021 01:25:46 +0000 (03:25 +0200)]
Hide long tracebacks on connect
Only hide tracebacks of errors coming from the database. They just come
from the database: seeing exactly where they explode inside psycopg is
not useful for the end user. Especially given the generator/waiting
interaction.
See https://github.com/psycopg/psycopg/issues/99 for an example of not
particularly helpful traceback (the message is the only useful bit).
Daniele Varrazzo [Fri, 8 Oct 2021 01:24:16 +0000 (03:24 +0200)]
Mark typing test slow
Daniele Varrazzo [Thu, 7 Oct 2021 21:20:49 +0000 (23:20 +0200)]
Don't generate random timestamps BC
It might happen to go below datetime.datetime.min if the timezone is
involved.
Today the random data generator is having a field day...
Daniele Varrazzo [Thu, 7 Oct 2021 19:47:02 +0000 (21:47 +0200)]
Add tests to verify that execute and enter respect cursor subclasses
Daniele Varrazzo [Thu, 7 Oct 2021 18:29:45 +0000 (20:29 +0200)]
Improve typing definition for cursor execute/enter
This is more redundant than using AnyCursor but it respects better the
interface returned.
Daniele Varrazzo [Thu, 7 Oct 2021 20:23:36 +0000 (22:23 +0200)]
Avoid to create emtpy ranges in tests
Daniele Varrazzo [Thu, 7 Oct 2021 20:15:47 +0000 (22:15 +0200)]
Bump up a timeout for a test which sometimes fails in the CI on windows
Federico Caselli [Thu, 7 Oct 2021 19:25:41 +0000 (21:25 +0200)]
Properly close the at hock event loop in test_windows_error
Daniele Varrazzo [Thu, 7 Oct 2021 12:51:10 +0000 (14:51 +0200)]
Test binary packages using Postgres 13 to avoid scram auth problems on aarch64
Daniele Varrazzo [Wed, 6 Oct 2021 22:17:47 +0000 (00:17 +0200)]
Build Windows binary package with libpq 14
Denis Laxalde [Thu, 7 Oct 2021 10:03:03 +0000 (12:03 +0200)]
Update Windows CI job for Postgres 14
Since image
20211003 .2, PostgreSQL 14 is installed:
https://github.com/actions/virtual-environments/blob/win19/
20211003 .2/images/win/Windows2019-Readme.md
Daniele Varrazzo [Thu, 7 Oct 2021 00:29:00 +0000 (02:29 +0200)]
Fix connection delay bias in pool test queue
The connection time was mistakenly counted in the execution time, which
made the test more brittle than already is.
Daniele Varrazzo [Wed, 6 Oct 2021 22:15:06 +0000 (00:15 +0200)]
Use PostgreSQL 14 to test in the CI
Daniele Varrazzo [Tue, 5 Oct 2021 13:41:55 +0000 (15:41 +0200)]
Install typing extensions for setuptools, pytest
Daniele Varrazzo [Tue, 5 Oct 2021 13:41:15 +0000 (15:41 +0200)]
Move mypy configuration to pyproject.toml
Daniele Varrazzo [Tue, 5 Oct 2021 11:31:26 +0000 (13:31 +0200)]
Dropped unneeded continue at the end of a loop
Daniele Varrazzo [Tue, 5 Oct 2021 13:09:48 +0000 (14:09 +0100)]
Merge pull request #91 from dlax/generators-c
Define and use send(), fetch_many() and fetch() generators in C implementation
Denis Laxalde [Sat, 2 Oct 2021 13:47:43 +0000 (15:47 +0200)]
Possibly use C impl of send() and fetch() in Cursor.stream()
Taking advantage of these functions being available per previous commit.
Denis Laxalde [Wed, 29 Sep 2021 12:09:29 +0000 (14:09 +0200)]
Extract send(), fetch_many() and fetch() from generators.execute() C impl
We extract send(), fetch_many() and fetch() functions from execute() C
implementation. This way, the code structure is similar to the Python
implementation.
Daniele Varrazzo [Mon, 4 Oct 2021 16:33:28 +0000 (18:33 +0200)]
Run mypy on pool/c Python files too
Denis Laxalde [Mon, 4 Oct 2021 09:20:55 +0000 (11:20 +0200)]
Explicitly disallow some mypy checks for tests/
When running mypy on the tests/ directory, the top-level mypy
configuration is used and strict mode is thus enforced whereas tests
code is not typed. As mypy might be run in people editors by default,
the current situation makes this uncomfortable.
By disabling obvious checks, we get a nicer developer experience. A more
straightforward option would be to set 'check_untyped_defs = False', but
this would prevent valid code to be checked, so the current proposal is
a trade-off.
Daniele Varrazzo [Mon, 4 Oct 2021 12:46:27 +0000 (14:46 +0200)]
Improve TypeInfo docs
Daniele Varrazzo [Mon, 4 Oct 2021 12:46:15 +0000 (14:46 +0200)]
Merge branch 'multirange'
Daniele Varrazzo [Mon, 4 Oct 2021 12:43:15 +0000 (14:43 +0200)]
Add Multirange documentation
Daniele Varrazzo [Mon, 4 Oct 2021 01:50:15 +0000 (03:50 +0200)]
Complete multirange implementation and test
- add register_multirange()
- check type of multirange items
- add extensive dump/load tests
Daniele Varrazzo [Sun, 3 Oct 2021 23:53:18 +0000 (01:53 +0200)]
Add total ordering on Multirange object
Also improved idiom for the Range ordering implementation.
Daniele Varrazzo [Sat, 2 Oct 2021 17:28:19 +0000 (19:28 +0200)]
Add random multirange tests
Daniele Varrazzo [Sat, 2 Oct 2021 03:31:23 +0000 (05:31 +0200)]
Add multirange adaptation
To be tested yet
Close #75
Daniele Varrazzo [Sat, 2 Oct 2021 01:48:34 +0000 (03:48 +0200)]
Refactor range adapters to make dumper and loader functions reusable
Daniele Varrazzo [Sat, 2 Oct 2021 19:23:51 +0000 (21:23 +0200)]
Print the unexpected warnings in tests
Daniele Varrazzo [Sat, 2 Oct 2021 03:30:01 +0000 (05:30 +0200)]
More compact parametrisation of range tests
Daniele Varrazzo [Sat, 2 Oct 2021 17:47:43 +0000 (19:47 +0200)]
Disable creation of empty ranges in random tests
If one end up as the first item in a list the array dumping the array
will fail. There are xfail tests in test_range for it.
Daniele Varrazzo [Sat, 2 Oct 2021 17:27:11 +0000 (19:27 +0200)]
Fix range comparisons in random tests
Daniele Varrazzo [Sat, 2 Oct 2021 17:22:26 +0000 (19:22 +0200)]
Modify range bounds to make them consistent with inf bounds
Without doing so, there could be a range '[,]' which would still be
reported as having not lower_inc or upper_inc.
Daniele Varrazzo [Sat, 2 Oct 2021 03:29:25 +0000 (05:29 +0200)]
Allow to use the pg(ver) marker with the module pytestmark
Denis Laxalde [Sat, 2 Oct 2021 16:57:21 +0000 (18:57 +0200)]
Add missing ExecStatusType enum values for pipeline mode
Follow-up
515e246b5f933ba93fdde93a03287c6940dbf85b ; these values were
only added for the Python definition of the enum.
Daniele Varrazzo [Fri, 1 Oct 2021 16:27:59 +0000 (17:27 +0100)]
Merge pull request #89 from dlax/pipeline
add libpq interface for pipeline mode
Denis Laxalde [Fri, 1 Oct 2021 14:43:17 +0000 (16:43 +0200)]
Use now available "postgres:14" Docker image in CI
Denis Laxalde [Wed, 29 Sep 2021 07:42:51 +0000 (09:42 +0200)]
Install libpq 14 in for some tests CI jobs
We add "libpq" matrix axis and select version 14 against various server
versions.
Denis Laxalde [Tue, 22 Jun 2021 14:07:47 +0000 (16:07 +0200)]
Add libpq interface for pipeline mode
Bindings are conditional so that compilation works with libpq version
older than 14.
Add a couple of tests inspired by PostgreSQL's tests.
Daniele Varrazzo [Thu, 30 Sep 2021 19:14:14 +0000 (21:14 +0200)]
Merge branch 'python-3.10'
Daniele Varrazzo [Thu, 30 Sep 2021 19:02:26 +0000 (21:02 +0200)]
Document Python 3.10 supported
Daniele Varrazzo [Thu, 30 Sep 2021 19:02:11 +0000 (21:02 +0200)]
Build Python 3.10 packages
Daniele Varrazzo [Thu, 30 Sep 2021 19:01:41 +0000 (21:01 +0200)]
Test with Python 3.10
Daniele Varrazzo [Thu, 30 Sep 2021 18:59:35 +0000 (20:59 +0200)]
Upgrade testing dependencies
Some are required for Python 3.10
Daniele Varrazzo [Thu, 30 Sep 2021 18:58:43 +0000 (20:58 +0200)]
Upgrade flake to 3.9
This version reports an error if a name is declared but not defined, so
set the lazy imported objects to None and deal with mypy complaining.
Daniele Varrazzo [Thu, 30 Sep 2021 12:51:24 +0000 (14:51 +0200)]
Improve test workflow
- Quote Python version number because 3.10 will cause issue as number
- Use 'postgres' instead of 'image'
- Add vim folds
Daniele Varrazzo [Thu, 30 Sep 2021 10:56:23 +0000 (12:56 +0200)]
Work around new delocation failure on Windows
https://github.com/psycopg/psycopg/runs/
3754213309 ?check_suite_focus=true
Not sure this is the right thing to do, if we are pulling in the right
dll.
Daniele Varrazzo [Thu, 30 Sep 2021 09:21:56 +0000 (11:21 +0200)]
Merge branch 'postgres-14'
Daniele Varrazzo [Thu, 30 Sep 2021 09:21:33 +0000 (11:21 +0200)]
Add supported systems in the docs
Daniele Varrazzo [Thu, 30 Sep 2021 00:56:41 +0000 (00:56 +0000)]
Add info about multirange types
Daniele Varrazzo [Thu, 30 Sep 2021 00:55:23 +0000 (00:55 +0000)]
Replace 'TypesRegistry.get_range()' with more generic 'get_by_subtipe()'
Daniele Varrazzo [Wed, 29 Sep 2021 20:01:31 +0000 (20:01 +0000)]
Simplify the types registry using a single map
Daniele Varrazzo [Wed, 29 Sep 2021 19:20:31 +0000 (19:20 +0000)]
Allow PSYCOPG_TEST_DSN to be an empty string
This allows to configure the test database by only using the PG* env
vars, while still leaving the option to wreck a database opt-in.
Daniele Varrazzo [Wed, 29 Sep 2021 19:05:13 +0000 (21:05 +0200)]
Add PostgreSQL 14 to the test grid
Denis Laxalde [Wed, 29 Sep 2021 08:36:23 +0000 (10:36 +0200)]
Adjust test on PGconn.tty for libpq 14
Quoting https://www.postgresql.org/docs/14/libpq-status.html:
[PQtty] no longer does anything, but it remains for backwards
compatibility. The function always return an empty string.
Or perhaps we should drop 'tty' property (and binding), since this is at
least obsolete since PostgreSQL 7.4?
Daniele Varrazzo [Tue, 28 Sep 2021 20:35:52 +0000 (22:35 +0200)]
Document that the ProactorEventLoop is incompatible with Psycopg
Daniele Varrazzo [Tue, 28 Sep 2021 20:14:28 +0000 (22:14 +0200)]
Improve error message if proactor event loop is used on Windows
Daniele Varrazzo [Tue, 28 Sep 2021 19:59:13 +0000 (20:59 +0100)]
Merge pull request #84 from CaselIT/asyncio_and_windows
Improve async story in the test on windows
Federico Caselli [Mon, 27 Sep 2021 20:39:31 +0000 (22:39 +0200)]
Support py36, test exception
Federico Caselli [Sun, 26 Sep 2021 19:53:22 +0000 (21:53 +0200)]
Require selector event loop on windows; enable aync tests
Fixes #76
Daniele Varrazzo [Mon, 27 Sep 2021 09:23:26 +0000 (11:23 +0200)]
Fix doc for Notify
Probably autoclass is confused by the class module being changed.
Daniele Varrazzo [Mon, 27 Sep 2021 00:01:30 +0000 (02:01 +0200)]
Print a log instead of a warning on swallowed exit exception
Daniele Varrazzo [Sun, 26 Sep 2021 23:33:16 +0000 (01:33 +0200)]
Drop TODO points covered by the _dns module
Daniele Varrazzo [Sun, 26 Sep 2021 23:37:02 +0000 (01:37 +0200)]
Merge branch 'no-connection-encoding'
Daniele Varrazzo [Sun, 26 Sep 2021 23:26:59 +0000 (01:26 +0200)]
Document that client_encoding is gone
Daniele Varrazzo [Fri, 14 May 2021 22:24:24 +0000 (00:24 +0200)]
Drop Connection.client_encoding attribute
Expose it as `Connection.info.encoding`, but use pgconn_encoding()
internally.
Daniele Varrazzo [Fri, 14 May 2021 21:55:31 +0000 (23:55 +0200)]
Add Cursor._pgconn internal property
Daniele Varrazzo [Fri, 14 May 2021 21:30:54 +0000 (23:30 +0200)]
Make encodings module private
Daniele Varrazzo [Fri, 14 May 2021 21:11:50 +0000 (23:11 +0200)]
Drop connection.set_client_encoding
The property might have been super useful in pre-unicode times, but now
it is really an implementation detail.
The connection encoding can be still changed setting the GUC or with a
PGOPTION or option in the connection params.
Daniele Varrazzo [Sun, 26 Sep 2021 22:08:10 +0000 (00:08 +0200)]
Fix beta release number to normalised form... again!
I seem unable to learn what that form is.
Daniele Varrazzo [Sun, 26 Sep 2021 21:08:11 +0000 (23:08 +0200)]
Drop "installing from source" from readme and install docs cleanup
Daniele Varrazzo [Sun, 26 Sep 2021 20:31:33 +0000 (22:31 +0200)]
Merge branch 'fix-dumping-by-oid'
Daniele Varrazzo [Sun, 26 Sep 2021 20:02:22 +0000 (22:02 +0200)]
Note fix of dumping by oid in news file
Daniele Varrazzo [Sun, 26 Sep 2021 20:18:54 +0000 (22:18 +0200)]
Improve AdaptersMap and Copy.set_types() docs
Daniele Varrazzo [Sun, 26 Sep 2021 19:47:06 +0000 (21:47 +0200)]
Fix BytesDumper return type annotation
Daniele Varrazzo [Thu, 23 Sep 2021 14:28:12 +0000 (16:28 +0200)]
Add C version of numeric dumpers by oid too
Daniele Varrazzo [Sun, 26 Sep 2021 19:02:22 +0000 (21:02 +0200)]
Better wording: writing to attribute -> setting the attribute
Daniele Varrazzo [Thu, 23 Sep 2021 14:22:26 +0000 (16:22 +0200)]
Fix selection of dumper by oid
Add a few dumpers which can deal with the different types and make sure
to register them so that the lookup by oid would use them. Because they
are slightly less efficient (with isinstance tests) we prefer to have
them only used by oid, instead of extending the domain of the dumpers
used in normal adaptation by type.
This commit only fixes the Python side. The C numeric dumpers require
more work.
Daniele Varrazzo [Thu, 23 Sep 2021 14:30:16 +0000 (16:30 +0200)]
Use natural iteration over pq.Format in parametrised tests
Daniele Varrazzo [Sun, 26 Sep 2021 19:41:08 +0000 (21:41 +0200)]
Better tests of dumping after set_types
These tests reveal that When there is a 1-n mapping between oid and
python classes, we might pick a dumper that cannot really deal with the
object to dump (for instance trying to dump an int to numeric with the
Decimal dumper).
Daniele Varrazzo [Sun, 26 Sep 2021 18:32:36 +0000 (20:32 +0200)]
Add news file to the docs
Daniele Varrazzo [Thu, 23 Sep 2021 15:23:39 +0000 (17:23 +0200)]
Allow to specify a custom adapt context on connect
Close #83
Daniele Varrazzo [Fri, 24 Sep 2021 15:36:22 +0000 (17:36 +0200)]
Fix beta release number to normalised form
Daniele Varrazzo [Thu, 23 Sep 2021 01:59:19 +0000 (03:59 +0200)]
Fix partially written comment
The issue was addressed before the comment could be written in all its ugliness.
Daniele Varrazzo [Wed, 22 Sep 2021 20:10:09 +0000 (22:10 +0200)]
Clear the prepared statements if the catalog might have changed
This could happen on DROP (and re-creation) of some object or on the
implicit drop caused by a rollback.
Issues happen if a prepared transaction is executed after an object has
been dropped: see
https://github.com/sqlalchemy/sqlalchemy/issues/6842#issuecomment-
925131836
Postgres devs suggestion is to avoid to do that: see
https://www.postgresql.org/message-id/
2220999 .
1632335261 %40sss.pgh.pa.us
Daniele Varrazzo [Thu, 23 Sep 2021 01:15:52 +0000 (03:15 +0200)]
Improve range random testing
Make sure to create a numrange[] field instead of text, which would have
happened if the random selection in Faker.types_names() had picked an
array whose first object is an empty range.
Also extend the tests to certain binary range types which previously
couldn't be managed.
Daniele Varrazzo [Fri, 14 May 2021 21:18:38 +0000 (23:18 +0200)]
Fix a couple of imports as relative
Daniele Varrazzo [Wed, 22 Sep 2021 13:55:19 +0000 (15:55 +0200)]
Specify better that registering adapters only affects new object
It is unclear that registering adapters globally doesn't affect existing
connections, including the one used to fetch info... See
https://github.com/sqlalchemy/sqlalchemy/issues/6842#issuecomment-
924425360
for instance.
Daniele Varrazzo [Tue, 21 Sep 2021 19:01:48 +0000 (20:01 +0100)]
Improve pq.version() doc
Daniele Varrazzo [Tue, 21 Sep 2021 18:54:24 +0000 (19:54 +0100)]
Expose the libpq build version as psycopg.pq.__build_version__
Daniele Varrazzo [Tue, 21 Sep 2021 18:27:41 +0000 (19:27 +0100)]
Merge branch 'add-shapely-adapters'
Daniele Varrazzo [Tue, 21 Sep 2021 16:09:23 +0000 (17:09 +0100)]
Use class instead of type to create shapely dumpers
Daniele Varrazzo [Tue, 21 Sep 2021 15:54:14 +0000 (16:54 +0100)]
Report a friendly message if the geometry info is not available
Daniele Varrazzo [Tue, 21 Sep 2021 15:32:01 +0000 (16:32 +0100)]
Create postgis extension if needed by tests, shapely tests cleanup
Daniele Varrazzo [Tue, 21 Sep 2021 15:24:32 +0000 (16:24 +0100)]
Move shapely test to the types package
Daniele Varrazzo [Tue, 21 Sep 2021 15:19:17 +0000 (16:19 +0100)]
Shapely docs cleanup
Jacopo Farina [Mon, 20 Sep 2021 12:57:30 +0000 (14:57 +0200)]
Fix test MultiPolygon geometry, add test for it