martin-evans.me.uk Always finish the day on a success

I have released DBD::ODBC 1.36-1 to cpan. With hindsight I was probably too ambitious releasing 1-35 with the default for execute_for_fetch enabled. This decision was based on a) my experience with writing bound arrays of parameters in the ODBC-ODBC Bridge b) I knew some drivers would have issues but I thought this might encourage fixes c) it is so much faster. As it turns out I am seeing issues currently with SQLite, Freetds, MS Access and even MS SQL Server.

I've uploaded Perl DBD::ODBC 1.35 to CPAN.

This is the culmination of 7 development releases in the 1.34 chain and is a significant release containing a lot of changes and enhancements. As always I would like to thank everyone who has helped and especially CPAN testers. The full list of changes since the last official release is below.

My intention is to release this as 1.35 in the next week so if you find any issues please let me know soon.

=head2 Changes in DBD::ODBC 1.34_5 February 17 2012

[BUG FIXES]

* The 40UnicodeRoundTrip tests counts could be 1 off in some cases.

* Fix for t/03batt.t which could fail a test if the data source had
no table - Kenichi Ishigaki

* If a driver misbehaves during global destruction e.g. SQLFreeStmt
fails but no error is available DBD::ODBC issues an error saying
an error occurred but no error diagnostics could be found. This is

I've uploaded DBD::ODBC 1.34_4 to CPAN.

This release contains all changes/fixes/enhancements since 1.33 and also the new odbc_getdigarec and odbc_getdiagfield methods.

=head2 Changes in DBD::ODBC 1.34_4 February 5 2012

[BUG FIXES]

* When odbc_getdiag* methods were added they installed themselves
into DBI but did not set IMP_KEEP_ERR so calling them cleared
DBI's errors.

=head2 Changes in DBD::ODBC 1.34_3 February 3 2012

[BUG FIXES]

* Linking against unixODBC was working by accident on most UNIX

I have uploaded DBD::ODBC 1.34_1 to CPAN. This release adds very experimental support for a native execute_for_fetch method to DBD::ODBC which means you can do multiple row inserts/updates/deletes much quicker than using DBI's default execute_for_fetch (so long as you are using execute_for_fetch or execute_array and using column-wise binding).

Native execute_for_fetch support can often be as much as 30 or 40 times faster in my experiments (depending on data sent and odbc_batch_size) when inserting rows and even more if you do it in a transaction and even faster if you disable index updates until after the commit. Read on for why.

By default DBD::ODBC will use the new execute_for_fetch and there are some differences between what happens when you use DBD::ODBC with DBI's default execute_for_fetch and DBD::ODBC's. To disable the ODBC one you can set the new odbc_disable_array_operations attribute. I've listed the potential difference below. I'd love people to test this. It may be missing some special workarounds for a few drivers but I'm working my way through those.

I've just uploaded DBD::ODBC 1.33 to CPAN.

This release contains no new changes since the 1.32_5 development release but is the official release for all the 1.32 dev series. The complete changes can be found below. The main thrust has been Unicode improvements.

The most significant enhancements are:

• unicode support for catalog/schema/table/column names in table_info and column_info
• now using SQLGetTypeInfoW on unicode builds
• new odbc_driver_complete attribute to allow interaction with ODBC dialogs on Windows.
  I'd really like to hear from anyone who uses this as I've had no feedback so far - just drop me an email if you have

and bug fixes:

• remove a debugging printf - sorry, most embarrassing.

In a rare moment (I hope) of stupidity last weekend when I was a little bored (bad cold stopped me doing what I really wanted to do) I looked through the DBD::ODBC TO_DO list and saw the "implement execute_array/execute_for_fetch" item which has been in that file for years. I know for sure implementing it will be a lot faster than using DBI's default methods (which effectively do a row at a time) mostly because I wrote an ODBC tutorial years ago showing how much faster binding arrays of parameters is (Easysoft ODBC-ODBC Bridge Performance White Paper) but I also knew it would be a PITA to write. The main problem is that DBD::ODBC has loads of workarounds for broken drivers and special cases.

I've uploaded the 4th development release of DBD::ODBC 1.32. This version improves Unicode support in metadata functions even further. My intention is to release this as a full version in the next week unless I get a load of bug reports. The changes since 1.31 are:

I have just uploaded 1.32_2 development release of DBD::ODBC. This contains one rather nice enhancement for Windows users which adds the new odbc_driver_complete attribute described as below. This allows DBD::ODBC on Windows to obtain a window handle which means if the ODBC Driver Manager or ODBC Driver needs to throw a dialogue to complete the data source or perhaps re-enter an expired password it can. Any testing will be greatly appreciated. The full changes since 1.31 are:

I maintian DBD::ODBC but I am wrestling with problems coping with 64 bit platforms. The problem is rather complex but I'm hoping that by describing it here someone will help come up with a logical answer to at least stop DBD::ODBC building in circumstances which are suspect.