Erlend Egeberg Aasland added the comment:
> If possible, I would prefer to only replace macros with static inline
> functions if the changes avoids clear macro pitfalls.
Yes, of course. Should I create a meta issue for this, or do you prefer raising
one issue p
Erlend Egeberg Aasland added the comment:
Fixed by GH-2.
--
nosy: +erlendaasland, vstinner
___
Python tracker
<https://bugs.python.org/issue15698>
___
___
Erlend Egeberg Aasland added the comment:
This seems to have been fixed in 2018 in bpo-33016 by GH-6010.
--
nosy: +erlendaasland
___
Python tracker
<https://bugs.python.org/issue12
Erlend Egeberg Aasland added the comment:
FYI, the SEGMENT_SIZE define was removed by
af01f668173d4061893148b54a0f01b91c7716c2 (bpo-16136).
--
nosy: +erlendaasland
___
Python tracker
<https://bugs.python.org/issue1103
Erlend Egeberg Aasland added the comment:
> As I wrote previously, I dislike macros. If someone is changed, I would
> prefer to convert the function into a static inline function which doesn't
> have macros pitfalls.
Should we create a meta issue for this? Most macros are trivial
Erlend Egeberg Aasland added the comment:
See bpo-43492 for upgrading the macOS & Windows installers to SQLite 3.35.0.
--
___
Python tracker
<https://bugs.python.org/iss
New submission from Erlend Egeberg Aasland :
SQLite 3.35.0 was released a couple of days ago:
https://www.sqlite.org/releaselog/3_35_0.html
Suggesting to hold off for a week or two, to see if a bug-fix release happens.
--
components: Windows, macOS
messages: 388685
nosy
Erlend Egeberg Aasland added the comment:
$ python -m pyperf compare_to -G main.json patched.json
Faster (1):
- bind: 63.6 us +- 1.2 us -> 61.8 us +- 0.9 us: 1.03x faster
$ git diff --shortstat master
1 file changed, 41 insertions(+), 74 deletions(-)
--
keywords: +patch
Added f
Erlend Egeberg Aasland added the comment:
Unfortunately, there's no way to detect R*Tree support in sqlite3.h. We could
run a script that dumps the compile options, and generate a config.h file from
that:
$ for L in $(sqlite3 ":memory:" "pragma compile_options"); do ech
Erlend Egeberg Aasland added the comment:
Test run output (see attached test file):
$ ./python.exe test_rtree.py
ARGS: ((-80.77490234375, -80.77469635009766, 35.377593994140625,
35.377803802490234), (45.3, 22.9, 5.0))
KWARGS: {'num_queued': [0, 1], 'context': None, 'level': 0, 'max_level': 1
Erlend Egeberg Aasland added the comment:
FYI, PoC patch attached. Lacks tests and some #ifdefs. Let me know if I should
create a PR out of it.
--
keywords: +patch
Added file: https://bugs.python.org/file49866/patch.diff
___
Python tracker
<ht
Erlend Egeberg Aasland added the comment:
I've opened bpo-43454 for R*Tree callbacks.
--
___
Python tracker
<https://bugs.python.org/issue43440>
___
___
Pytho
New submission from Erlend Egeberg Aasland :
Ref. bpo-43440
Now that both Windows and macOS builds compile SQLite with R*Tree support, we
should consider adding support for R*Tree callbacks.
SQLite has two API's:
- sqlite3_rtree_query_callback() for SQLite 3.8.5 and newer
Erlend Egeberg Aasland added the comment:
Anytime :)
I'll create an issue for rtree callbacks.
--
___
Python tracker
<https://bugs.python.org/issue43
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23569
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24801
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
> does compile options set in setup.py propagate to the Windows build, or do we
> need to set it both places?
Er, forget it. setup.py is of course only used to build the sqlite3 module, not
s
New submission from Erlend Egeberg Aasland :
Berker, can we please move the MODULE_NAME define from setup.py to
Modules/_sqlite/module.h? I'm tired of all the undeclared identifier warnings.
No other module defines their MODULE_NAME in setup.py.
--
components: Library (Lib)
messages
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23565
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24797
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
We should also consider adding support for R*Tree query callbacks using
sqlite3_rtree_query_callback() for SQLite >= 3.8.5, and
sqlite3_rtree_geometry_callback() for older versions.
--
___
Python trac
Erlend Egeberg Aasland added the comment:
> does compile options set in setup.py propagate to the Windows build, or do we
> need to set it both places?
Seems like it doesn't; correct me if I'm wrong. PR coming up.
--
___
Python tracker
Erlend Egeberg Aasland added the comment:
Actually, the macOS build already builds with R*Tree support enabled, but it is
missing from PCbuild/sqlite3.vcxproj. I'm not very familiar with the Windows
build; does compile options set in setup.py propagate to the Windows build, or
do we need
Erlend Egeberg Aasland added the comment:
I can put up a PR for it. I don’t see any reason not to enable it.
--
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
Unless I’m mistaken, that’s enabled simply by compiling with
SQLITE_ENABLE_RTREE defined.
--
___
Python tracker
<https://bugs.python.org/issue43
Change by Erlend Egeberg Aasland :
--
keywords: +patch
Added file: https://bugs.python.org/file49858/patch.diff
___
Python tracker
<https://bugs.python.org/issue43
New submission from Erlend Egeberg Aasland :
The module level connect method is guarded by PySys_Audit(), but
sqlite3.Connection.__init__() is not. It is possible to bypass the module level
connect() method simply by creating a new sqlite3.Connection object directly.
Easily fixed by either
Change by Erlend Egeberg Aasland :
--
pull_requests: +23516
stage: needs patch -> patch review
pull_request: https://github.com/python/cpython/pull/24746
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
Attached patch consists of two commits: one to rename "conn" to "con", and one
to rename "c" to "cur". Are you ok with that, Berker?
--
Added file: https://bugs.python.org/file49851/patch.diff
__
Erlend Egeberg Aasland added the comment:
'con' and 'cur' seems to be used in a majority of the examples. I suggest
normalising to always using these two in code examples.
--
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
Right, got me confused as well! I guess a clarification would be an improvement
:)
--
___
Python tracker
<https://bugs.python.org/issue43
New submission from Erlend Egeberg Aasland :
If the connection factory __init__ method fails, we hit a seg. fault when
pysqlite_do_all_statements() is called to clean up the defect connection:
PyList_Size received a NULL pointer.
Suggested fix:
Split pysqlite_do_all_statements() in two: one
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23515
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24745
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
Correct, fetchone() is a cursor method. Thanks for the report!
--
nosy: +berker.peksag, erlendaasland
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
Quoting from the SQLite docs:
"As long as the input parameters are correct, these routines will only fail if
an out-of-memory error occurs during a format conversion"
--
___
Python track
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23496
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24723
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
> But converting the decimal c api may breaks the compatibility, because some
> macros like `PyDec_TypeCheck_INDEX` have been exposed in headers.
True. Is there many external users of this API? I could not find any relevant
examples
Erlend Egeberg Aasland added the comment:
Hm, I guess we could use sqlite3_stmt_busy() instead of the in_use flag.
--
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
> There are three calls of pysqlite_statement_reset() in
> _pysqlite_query_execute()
Yes, but only two before the cache is queried.
> So additional call of pysqlite_statement_reset() does not harm.
That's true, but removing redundant c
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23466
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24681
___
Python tracker
<https://bugs.python.org/issu
New submission from Erlend Egeberg Aasland :
In _pysqlite_query_execute(), if the cursor already has a statement object, it
is reset twice before the cache is queried.
--
components: Library (Lib)
messages: 387850
nosy: berker.peksag, erlendaasland, serhiy.storchaka
priority: normal
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23465
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24680
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
> FreeBSD only (?) system call [...]
Correction, it's not a sys call. It belongs to the misc manpage section.
--
___
Python tracker
<https://bugs.python.org/issu
New submission from Erlend Egeberg Aasland :
In https://docs.python.org/3.10/library/resource.html#resource.RLIMIT_SWAP,
tuning(7) points to https://manpages.debian.org/tuning(7), however this is a
FreeBSD only (?) system call, so the link is incorrect.
I suggest linking to either:
- https
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23459
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24674
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
> If sqlite3_value_blob() returns NULL, we should check if sqlite3_errcode()
> equals SQLITE_NOMEM and raise MemoryError if it does.
This also applies to sqlite3_value_text().
It also applies to sqlite3_value_bytes() if a conversion takes pla
Change by Erlend Egeberg Aasland :
--
title: [sqlite3] sqlite3_value_bytes() should be called after
sqlite3_value_blob() -> [sqlite3] Fix sqlite3_value_blob() usage
___
Python tracker
<https://bugs.python.org/issu
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23443
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24658
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
Addendum to msg387641:
> The latter only leaves a valid Cursor->statement->st pointer (sqlite3_stmt
> pointer) if the Statement object was successfully created, and the
> sqlite3_stmt successfully prepared.
sqlite3_prepare_v2() a
Erlend Egeberg Aasland added the comment:
msg387668 was a little bit hasty. I'll try again:
Dong-hee Na:
> Hmm by the way the current implementation returns SQLITE_OK if the statement
> is NULL, but it looks like return SQLITE_MISUSE if we apply this patch.
> Does it not cause any
Erlend Egeberg Aasland added the comment:
> int rc = sqlite3_reset(NULL);
>printf("reset with NULL: %d %s\n", rc, sqlite3_errstr(rc));
Sorry, wrong test:
int rc = sqlite3_step(NULL);
printf("step with NULL: %d %s\n", rc, sqlite3_errstr(rc));
$ ./a.
Erlend Egeberg Aasland added the comment:
Ah, at last I found the source of the confusion: The SQLite changelog.
Quoting from msg387619 and https://sqlite.org/changes.html:
> From the SQLite 3.5.3 changelog:
> - sqlite3_step() returns SQLITE_MISUSE instead of crashing when
Erlend Egeberg Aasland added the comment:
Introduced by Christian Heimes 2007-12-14 in commit
380532117c2547bb0dedf6f85efa66d18a9abb88, which is a merge from SVN (?) checkin
r59471 by Gerhard Häring
(https://mail.python.org/pipermail/python-checkins/2007-December/063968.html
Erlend Egeberg Aasland added the comment:
> Look at issue in which the workaround was added. Does it contain some
> examples?
I'll check. Thanks.
--
___
Python tracker
<https://bugs.python.org/i
Erlend Egeberg Aasland added the comment:
> I wonder if it is possible at all to reach this branch. If it is not, then
> I'm pretty sure Cursor.in_use is redundant
Typo: should be Statement.in_use
Corner error cases may cause the _pysqlite_query_execute() loop to exit w
Erlend Egeberg Aasland added the comment:
> _pysqlite_query_execute() has one coverage gap: lines 478-488 (if
> (self->statement->in_use)) are never executed.
I wonder if it is possible at all to reach this branch. If it is not, then I'm
pretty sure Cursor.in_use is redundant,
Erlend Egeberg Aasland added the comment:
Regarding test coverage:
_pysqlite_query_execute() has one coverage gap: lines 478-488 (if
(self->statement->in_use)) are never executed. Except from that, the only
missing are the hard-to-trigger goto error's (for example PyLi
Erlend Egeberg Aasland added the comment:
pysqlite_cursor_iternext() has four users:
- sqlite3.Cursor.fetchone()
- sqlite3.Cursor.fetchall()
- sqlite3.Cursor.fetchmany()
- sqlite3.Cursor.__next__()
All of these methods pass self to pysqlite_cursor_iternext().
pysqlite_cursor_iternext() starts
Erlend Egeberg Aasland added the comment:
There are six users of pysqlite_step():
$ grep -nrE "\" Modules/_sqlite
Modules/_sqlite/util.c:27:int pysqlite_step(sqlite3_stmt* statement,
pysqlite_Connection* connection)
Modules/_sqlite/connection.c:393:rc = pysqlite_step(state
Erlend Egeberg Aasland added the comment:
I’ll check all uses and see if we’ve got everything covered by the test suite.
This is one of the core functions of the sqlite3 module (all queries call step
at least once, but often multiple times), so I’d expect coverage is pretty good
Erlend Egeberg Aasland added the comment:
I believe we can proceed with this as planned. Serhiy, do you have additional
comments or change requests?
--
___
Python tracker
<https://bugs.python.org/issue43
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23423
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24638
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
>From the SQLite 3.5.3 changelog:
- sqlite3_step() returns SQLITE_MISUSE instead of crashing when called with a
NULL parameter.
--
___
Python tracker
<https://bugs.python.org/issu
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23422
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24637
___
Python tracker
<https://bugs.python.org/issu
New submission from Erlend Egeberg Aasland :
I missed the SQLITE_OPEN_URI #ifdef in GH-24106. The "open URI" interface was
added in 3.7.7. The dead code can safely be removed.
$ grep -r SQLITE_OPEN_URI sqlite-3.7.6/
$ grep -r SQLITE_OPEN_URI sqlite-3.7.7/
[...]
sqlite-3.7.7/sqlite3
Erlend Egeberg Aasland added the comment:
Larry Brasfield's comment https://sqlite.org/forum/forumpost/6430fc589d?t=h
aligns with https://bugs.python.org/issue43251#msg387428
I'll think twice before posting there again, though.
--
___
Python
Erlend Egeberg Aasland added the comment:
Sure! Here it is: https://sqlite.org/forum/forumpost/574c6bc66c
--
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
After discussing the matter briefly on the SQLite forum, I'm no longer 100%
certain about this. There seems to be uncertainty about which other conditions
can produce NULL, although memory error seems to be the most probable. I
consider closing both
Change by Erlend Egeberg Aasland :
--
nosy: +erlendaasland
___
Python tracker
<https://bugs.python.org/issue35056>
___
___
Python-bugs-list mailing list
Unsub
Erlend Egeberg Aasland added the comment:
Related:
If sqlite3_value_blob() returns NULL, we should check if sqlite3_errcode()
equals SQLITE_NOMEM and raise MemoryError if it does.
If not, we should initialise cur_py_value to None, because as the
PyBytes_FromStringAndSize docs says: "
Change by Erlend Egeberg Aasland :
--
nosy: +erlendaasland
___
Python tracker
<https://bugs.python.org/issue24905>
___
___
Python-bugs-list mailing list
Unsub
New submission from Erlend Egeberg Aasland :
The sqlite3_value_*() API is almost identical to the sqlite3_column_*() API.
sqlite3_value_bytes() should be called after we've converted the value using
sqlite3_value_blob().
See also bpo-43249.
--
components: Library (Lib)
messages
New submission from Erlend Egeberg Aasland :
In Modules/_sqlite/connection.c:
pysqlite_connection_commit_impl(), pysqlite_connection_rollback_impl(), and
_pysqlite_connection_begin() all call sqlite3_prepare_v2() with the fourth
parameter (pzTail) set. This (output) parameter is never used
New submission from Erlend Egeberg Aasland :
pysqlite_step() contains a NULL check needed for SQLite 3.5 and earlier. This
can be removed.
--
components: Library (Lib)
messages: 387476
nosy: berker.peksag, erlendaasland, serhiy.storchaka
priority: normal
severity: normal
status: open
Erlend Egeberg Aasland added the comment:
> I am not sure that the difference is significant enough to justify the
> optimization.
I would guess it's negligible. I did some quick timeit tests, but the results
were pretty much equal, so I guess I'd have to use a more precise ben
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23388
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24609
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
Will do. Thanks for pushing the investigation.
--
___
Python tracker
<https://bugs.python.org/issue43251>
___
___
Pytho
Change by Erlend Egeberg Aasland :
--
pull_requests: +23383
pull_request: https://github.com/python/cpython/pull/24605
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
Serhiy also mentioned that marking functions definitions with the extern
keyword is not very common in the CPython source base.
https://github.com/python/cpython/pull/24578#discussion_r579506678
I suggest cleaning up those as well, while we're
Erlend Egeberg Aasland added the comment:
"SELECT 1" yields "1" as column name.
"SELECT <...> AS N" yields "N" as column name.
"SELECT NULL" yields "NULL" as column name.
We can't set the column name via the API, so that's
Erlend Egeberg Aasland added the comment:
Regarding 1: After thinking about it, I see nothing wrong with raising
ValueError error. No need to change current behaviour.
--
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
3.9, I presume? Yeah, I guess so.
--
___
Python tracker
<https://bugs.python.org/issue43265>
___
___
Python-bugs-list m
Erlend Egeberg Aasland added the comment:
That's inside sqlite3_value_text() and friends, then? Let's investigate further
before concluding.
--
___
Python tracker
<https://bugs.python.org/issue43
Erlend Egeberg Aasland added the comment:
The unit test suite shows one case of improved "exception text". I'd say it's
an improvement.
$ ./python.exe # with GH-24586 applied
>>> import sqlite3
>>> c1 = sqlite3.connect(":memory:")
>>> c2 = sql
Change by Erlend Egeberg Aasland :
--
title: Improve sqlite3 backup error handling -> [sqlite3] Improve backup error
handling
___
Python tracker
<https://bugs.python.org/issu
Change by Erlend Egeberg Aasland :
--
title: sqlite3_column_name() failures should raise MemoryError -> [sqlite3]
sqlite3_column_name() failures should raise MemoryError
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
> Well, it returns NULL in case of out of memory, but is it the only cause? Can
> NULL be returned for other reasons?
According to the SQLite docs, no. Looking at the source code, we see that it
also returns NULL if the second parameter (
Erlend Egeberg Aasland added the comment:
> It is a kind of optimization.
PyLong_Check is very fast (only one comparison, AFAICS), so there is no gain in
first doing PyLong_CheckExact and then PyLong_Check. Ditto for unicode.
PyFloat_Check is the most expensive check, but it comes bef
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23366
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24586
___
Python tracker
<https://bugs.python.org/issu
Erlend Egeberg Aasland added the comment:
> I'm not sure about 1) because if target == source it means a user error.
> OperationalError is usually used for non-user errors.
Yes, my bad. ProgrammingError would be better.
target == source results in SQLITE_ERROR, BTW.
I'll throw up a PR
Erlend Egeberg Aasland added the comment:
Ref. https://github.com/python/cpython/pull/24569#issuecomment-782014177,
Berker.
--
___
Python tracker
<https://bugs.python.org/issue43
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23357
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24578
___
Python tracker
<https://bugs.python.org/issu
New submission from Erlend Egeberg Aasland :
There's a lot of file scoped functions without the static storage-class
specifier. All file local functions should have the static storage-class
specifier.
--
components: Library (Lib)
messages: 387310
nosy: berker.peksag, erlendaasland
Change by Erlend Egeberg Aasland :
--
pull_requests: +23353
pull_request: https://github.com/python/cpython/pull/24574
___
Python tracker
<https://bugs.python.org/issue43
New submission from Erlend Egeberg Aasland :
The type checks at the start of pysqlite_statement_bind_parameter() are
redundant:
We first check for exact types (*_CheckExact()), and then we check again for
exact or subtyped versions (*_Check()).
(Adding to the redundantness: the result
Change by Erlend Egeberg Aasland :
--
title: sqlite3_column_name() failures should call PyErr_NoMemory() ->
sqlite3_column_name() failures should raise MemoryError
___
Python tracker
<https://bugs.python.org/issu
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23352
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24573
___
Python tracker
<https://bugs.python.org/issu
New submission from Erlend Egeberg Aasland :
str[size-1] is rendered incorrectly in
https://docs.python.org/3.10/c-api/conversion.html
The original intent was probably to apply italics to the variables names only.
This can be fixed by prefixing the first bracket with a backslash. However
New submission from Erlend Egeberg Aasland :
There are some issues with the error handling in
pysqlite_connection_backup_impl():
1. ValueError is returned if the target connection equals source connection.
Should be OperationalError, IMHO.
2. The aforementioned check is already performed
Change by Erlend Egeberg Aasland :
--
keywords: +patch
pull_requests: +23348
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/24569
___
Python tracker
<https://bugs.python.org/issu
New submission from Erlend Egeberg Aasland :
If no rows match an aggregate query, _pysqlite_step_callback() is never called,
and _pysqlite_final_callback() is called once. In order to prevent needless
context allocation, we should pass 0 as the second argument to
sqlite3_aggregate_context
Change by Erlend Egeberg Aasland :
--
pull_requests: +23346
pull_request: https://github.com/python/cpython/pull/24565
___
Python tracker
<https://bugs.python.org/issue43
201 - 300 of 656 matches
Mail list logo
