Обсуждение: Beta of DBD::Pg 1.44 available
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The first beta of DBD::Pg version 1.44 is now available at: http://search.cpan.org/~dbdpg/ SHA1 and MD5 Checksums for the tarball: 1c8593d2bf66562ac9ff1f169bb02aabb178ad00 DBD-Pg-1.43_1.tar.gz a5752ac6d69928f7a8b2254ae8377ee0 DBD-Pg-1.43_1.tar.gz This version has some deep changes, so testing is much appreciated. In particular, both the do() and quote() methods have been rewritten. In addition, the get_info() information has been greatly expanded. Information I was not sure about is marked with a "## ??" in the Pg.pm code - anyone familiar with the ODBC / ANSI standards, please take a look. Here is the complete list of changes: - Added $dbh->{pg_default_port} method [GSM]- Overhaul get_info data, add many more values [GSM]- Overhaul type_info data(fixes CPAN bug #13806) [GSM]- Explicitly prevent any parsing of statements for placeholders unless they begin with"SELECT", "INSERT", "UPDATE", or "DELETE" Suggested by (and fixes) CPAN bug #13608 [GSM]- Rewrite some of the quotingfunctions, reduce dependence on libpq versions. [GSM]- Rewrite and optimize the do() method. Should be much faster when called without placeholders. Thanks to Tom Lane for suggesting this. [GSM]- Double check PQserverVersion returnand use alternate method if it returns 0 (CPAN bug #14302)- Add support for specifying type in $dbh->quote(), suchas $dbh->quote($var, {pg_type => DBD::Pg::PG_BYTEA}). Also support type => SQL_xx [GSM] (fixes CPAN bug #13942)- Fixpg_notifies() bug (CPAN bug #14232) [door@lcpi.ru]- Add pg_ping() method [GSM]- Make sure ping returns true, even if infailed transaction state. [thanks to Bill Moseley] [GSM]- Fix COPY-related core dump [GSM]- Fix strncpy bug in quote.c[Jun Kuriyama]- Fix error in is_high_bit_set(). [Alexey Tourbin] (CPAN bug #13406) - -- Greg Sabino Mullane greg@turnstep.com PGP Key: 0x14964AC8 200509132018 http://biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8 -----BEGIN PGP SIGNATURE----- iD8DBQFDJ2yFvJuQZxSWSsgRAg+xAKD6sNJTITEnLDYxpENjPWZinotGpwCfdI/+ tvg9CbWzz9B8jBf132dv/6o= =ev63 -----END PGP SIGNATURE-----
On 9/13/05 8:24 PM, "Greg Sabino Mullane" <greg@turnstep.com> wrote: > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > > The first beta of DBD::Pg version 1.44 is now available at: > > http://search.cpan.org/~dbdpg/ > > SHA1 and MD5 Checksums for the tarball: > > 1c8593d2bf66562ac9ff1f169bb02aabb178ad00 DBD-Pg-1.43_1.tar.gz > a5752ac6d69928f7a8b2254ae8377ee0 DBD-Pg-1.43_1.tar.gz > > This version has some deep changes, so testing is much appreciated. In > particular, both the do() and quote() methods have been rewritten. > In addition, the get_info() information has been greatly expanded. > Information I was not sure about is marked with a "## ??" in the Pg.pm > code - anyone familiar with the ODBC / ANSI standards, please take a look. > > Here is the complete list of changes: > > - Added $dbh->{pg_default_port} method [GSM] > - Overhaul get_info data, add many more values [GSM] > - Overhaul type_info data (fixes CPAN bug #13806) [GSM] > - Explicitly prevent any parsing of statements for placeholders > unless they begin with "SELECT", "INSERT", "UPDATE", or "DELETE" > Suggested by (and fixes) CPAN bug #13608 [GSM] Greg, Thanks for doing all of this. I will give it a whirl. Just a quick question on the point directly above: does this imply that explain and explain analyze are excluded from placeholder parsing? I don't personally use them programmatically, but I think that particularly for count(*) types of queries or receiving estimates of the number of returned tuples without executing the query, people could potentially use explain select.... Pardon my laziness for not looking at the code for my own answers. Thanks, Sean
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 >> - Explicitly prevent any parsing of statements for placeholders >> unless they begin with "SELECT", "INSERT", "UPDATE", or "DELETE" >> Suggested by (and fixes) CPAN bug #13608 [GSM] > Thanks for doing all of this. I will give it a whirl. Just a quick > question on the point directly above: does this imply that explain and > explain analyze are excluded from placeholder parsing? I don't personally > use them programmatically, but I think that particularly for count(*) types > of queries or receiving estimates of the number of returned tuples without > executing the query, people could potentially use explain select.... Pardon > my laziness for not looking at the code for my own answers. Excellent question. There obviously needs to be a way around this, as your explain scenario points out. I have another parameter "pg_client_prepare" waiting in the wings for just such an occasion. The question becomes, do we make parsing the default, and break lots of future apps that start using the "dollar-quoting" feature, or do we make the default to only parse DML statements, possibly breaking lots of older apps? I'm leaning heavily towards the latter, not least of which is because the backend's PREPARE only accepts DML as well, but welcome any discussion on the subject. A third (and probably best) option is to make the parser inside of DBD::Pg recognize dollar-quoting. I'll play around with that one some more. - -- Greg Sabino Mullane greg@turnstep.com PGP Key: 0x14964AC8 200509141424 https://www.biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8 -----BEGIN PGP SIGNATURE----- iD8DBQFDKLeWvJuQZxSWSsgRAriGAJsG5AML1mv8qYGvaFYqQ1VYohaicgCgsHkN gqzxS8R7GNH6NsF1kyBGfr0= =GWQc -----END PGP SIGNATURE-----
"Greg Sabino Mullane" <greg@turnstep.com> writes:
>> Thanks for doing all of this. I will give it a whirl. Just a quick
>> question on the point directly above: does this imply that explain and
>> explain analyze are excluded from placeholder parsing?
The existing approach for that isPREPARE foo(parameter types) AS SELECT ...EXPLAIN [ANALYZE] EXECUTE foo(parameter
values)
> ... A third (and probably best) option is to make the parser
> inside of DBD::Pg recognize dollar-quoting.
I think you need to do that, independently of anything else ...
regards, tom lane
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 >> ... A third (and probably best) option is to make the parser >> inside of DBD::Pg recognize dollar-quoting. > I think you need to do that, independently of anything else ... Yes, I agree, this is the route to go. Which will also solve the original problem. Stay tuned for a beta2. - -- Greg Sabino Mullane greg@turnstep.com PGP Key: 0x14964AC8 200509152041 http://biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8 -----BEGIN PGP SIGNATURE----- iD8DBQFDKhTQvJuQZxSWSsgRAh6VAJ4jNeRfrMfaTpGdcWzXuJUhd40MhQCguJFa YeQqUW5tk2A7oBk/myEyPOA= =2sWA -----END PGP SIGNATURE-----