RFC for Sqlite3 Extended Error Codes

Hello,

My apologies if I violate any netiquette during this e-mail. This is my
first time doing something like this and I am following the guide from the
CONTRIBUTING.md to the best of my ability.

I would to extend the functionality of the current sqlite3 implementation
(in ext/sqlite3 and in ext/pdo_sqlite) by exposing the functions
related to extended result codes. I have already forked and made my changes
at
https://github.com/rkopack/php-src/commit/f6c4d1eeccd27a3c5bd836fd8205b1283bd5eabc.
From
my understanding Scott MacVicar and Christoph M. Becker are the primary
maintainers for the sqlite3 extension and Ilia Alshanetsky is the primary
maintainer for the pdo_sqlite extension; if any of you would like to
comment on this (it's around 50 lines of new code) so I can get feedback on
this I would greatly appreciate it. I made sure to follow the convention I
saw in the file(s) to the best of my ability but I am sure there are little
things I might have missed.

Thank you for your time,
Robert Kopack

I would to extend the functionality of the current sqlite3 implementation
(in ext/sqlite3 and in ext/pdo_sqlite) by exposing the functions
related to extended result codes. I have already forked and made my changes
at
https://github.com/rkopack/php-src/commit/f6c4d1eeccd27a3c5bd836fd8205b1283bd5eabc.
From
my understanding Scott MacVicar and Christoph M. Becker are the primary
maintainers for the sqlite3 extension and Ilia Alshanetsky is the primary
maintainer for the pdo_sqlite extension;

It's best to look at EXTENSIONS in master[1], which also reveals the
most up-to-date dates of maintainership – i.e. pdo_sqlite is effectively
maintained by the core maintainers.

if any of you would like to
comment on this (it's around 50 lines of new code) so I can get feedback on
this I would greatly appreciate it. I made sure to follow the convention I
saw in the file(s) to the best of my ability but I am sure there are little
things I might have missed.

I haven't reviewed closely for now, but basically I like these
additions. With regard to sqlite3_extended_result_codes() we have to
review existing error checks (they likely need small adjustments, if
extended error codes are enabled). And we likely want to have a few
tests for the new functionality.

Feel free to submit a pull request[2]; it seems to me any discussion
could happen on Github as well. :)

[1] https://github.com/php/php-src/blob/master/EXTENSIONS
[2] https://github.com/php/php-src/pulls

Thanks,
Christoph

On Wed, May 15, 2019 at 7:00 PM Christoph M. Becker cmbecker69@gmx.de
wrote:

I would to extend the functionality of the current sqlite3 implementation
(in ext/sqlite3 and in ext/pdo_sqlite) by exposing the functions
related to extended result codes. I have already forked and made my
changes
at

https://github.com/rkopack/php-src/commit/f6c4d1eeccd27a3c5bd836fd8205b1283bd5eabc
.

From
my understanding Scott MacVicar and Christoph M. Becker are the primary
maintainers for the sqlite3 extension and Ilia Alshanetsky is the primary
maintainer for the pdo_sqlite extension;

It's best to look at EXTENSIONS in master[1], which also reveals the
most up-to-date dates of maintainership – i.e. pdo_sqlite is effectively
maintained by the core maintainers.

That's strange as I did look in that file first which is how I found the
three names. Perhaps I read it wrong. Mea Cupla.

if any of you would like to
comment on this (it's around 50 lines of new code) so I can get feedback
on
this I would greatly appreciate it. I made sure to follow the convention
I
saw in the file(s) to the best of my ability but I am sure there are
little
things I might have missed.

I haven't reviewed closely for now, but basically I like these
additions. With regard to sqlite3_extended_result_codes() we have to
review existing error checks (they likely need small adjustments, if
extended error codes are enabled). And we likely want to have a few
tests for the new functionality.

I didn't consider that! And just to clarify are you referring to the
results from sqlite3_extended_errcode() or the return value from
sqlite3_extended_result_codes()?
If the former, I'll have to look around to see where those kind of checks
are done - do you know of any locations off the top of your head? From my
understanding we can just mask the incoming error code with 0xFF (That's
what sqlite does anyway) and always get "what we expected" it to be in the
past instead of trying to account for the additional extended error codes
in some other manner.
If it's the latter, and after glancing at the sqlite source, it looks like
that function returns either SQLITE_MISUSE (21) or SQLITE_OK (0). I put in
-1 instead of 21 as the return for that function in error because I felt
that was following what PHP normally does on error (perhaps FALSE would be
better?) In fact, on further thought, returning the 0 from success on that
might be confusing as well. Perhaps it would better suited to check the
return and return TRUE on SQLITE_OK and FALSE on SQLITE_MISUSE?

With regards to Unit Tests that is a fantastic idea (that I completely
forgot about and i'm sure my team lead would lecture me on much to my
chagrin) and I will be sure to add some to cover the new functions at a
minimum.

Feel free to submit a pull request[2]; it seems to me any discussion
could happen on Github as well. :)

I have already done so ( https://github.com/php/php-src/pull/4166 ) and
i'll be sure to keep any more discussion beyond these last questions there.
Thank you for your time!

On Wed, May 15, 2019 at 7:00 PM Christoph M. Becker cmbecker69@gmx.de
wrote:

I haven't reviewed closely for now, but basically I like these
additions. With regard to sqlite3_extended_result_codes() we have to
review existing error checks (they likely need small adjustments, if
extended error codes are enabled). And we likely want to have a few
tests for the new functionality.

I didn't consider that! And just to clarify are you referring to the
results from sqlite3_extended_errcode() or the return value from
sqlite3_extended_result_codes()?

I was referring to the slight behavioral change that is caused by
calling sqlite3_extended_result_codes(db, 1). As I understand it, this
will cause other functions to possibly return extended error codes
instead of primary result codes. It seems, though, that there would not
be any conflict right now, since only SQLITE_OK, SQLITE_ROW and
SQLITE_DONE are checked in ext/sqlite3, and these do not have extended
error codes. Still it might make sense to mask these checks to prevent
future issues. And of course, we would have to be careful, if we add
checks for e.g. SQLITE_BUSY in the future.

If it's the latter, and after glancing at the sqlite source, it looks like
that function returns either SQLITE_MISUSE (21) or SQLITE_OK (0). I put in
-1 instead of 21 as the return for that function in error because I felt
that was following what PHP normally does on error (perhaps FALSE would be
better?) In fact, on further thought, returning the 0 from success on that
might be confusing as well. Perhaps it would better suited to check the
return and return TRUE on SQLITE_OK and FALSE on SQLITE_MISUSE?

I think the method should return TRUE on success, and FALSE on failure.
An alternative might be to return the old value, like done by
SQLite3::enableExceptions().

Thank you for your time!

Thanks for your's. :)

--
Christoph M. Becker

Hello everybody,

I've had the PR open for this a while now
(https://github.com/php/php-src/pull/4166) and was hoping to get more
eyes on it; Kalle was looking at it but he appears to have not been as
active lately and as I believe it is in a good state right now I was
hoping to have some other feedback on it. I was also curious if this
would go into a 7.x version of PHP or if something like this would
only be for the next major version (8.0).

Thank you,
Robert

On Wed, May 15, 2019 at 7:00 PM Christoph M. Becker cmbecker69@gmx.de
wrote:

I haven't reviewed closely for now, but basically I like these
additions. With regard to sqlite3_extended_result_codes() we have to
review existing error checks (they likely need small adjustments, if
extended error codes are enabled). And we likely want to have a few
tests for the new functionality.

I didn't consider that! And just to clarify are you referring to the
results from sqlite3_extended_errcode() or the return value from
sqlite3_extended_result_codes()?

I was referring to the slight behavioral change that is caused by
calling sqlite3_extended_result_codes(db, 1). As I understand it, this
will cause other functions to possibly return extended error codes
instead of primary result codes. It seems, though, that there would not
be any conflict right now, since only SQLITE_OK, SQLITE_ROW and
SQLITE_DONE are checked in ext/sqlite3, and these do not have extended
error codes. Still it might make sense to mask these checks to prevent
future issues. And of course, we would have to be careful, if we add
checks for e.g. SQLITE_BUSY in the future.

If it's the latter, and after glancing at the sqlite source, it looks like
that function returns either SQLITE_MISUSE (21) or SQLITE_OK (0). I put in
-1 instead of 21 as the return for that function in error because I felt
that was following what PHP normally does on error (perhaps FALSE would be
better?) In fact, on further thought, returning the 0 from success on that
might be confusing as well. Perhaps it would better suited to check the
return and return TRUE on SQLITE_OK and FALSE on SQLITE_MISUSE?

I think the method should return TRUE on success, and FALSE on failure.
An alternative might be to return the old value, like done by
SQLite3::enableExceptions().

Thank you for your time!

Thanks for your's. :)

--
Christoph M. Becker