September 8, 1996
Fixed another completion buglet in pmsql (complete-table-or-field chose to query an inexistant statement handle in some cases).
Msql::Statement now 'use strict' clean, except for the autoloader.
Added my recent posting about 'insert' closures to the FAQ.
Reformatted both Changes and FAQ (Hi, Doug!).
As I'm one of the privileged people getting prereleases of Msql2, I have already added support for Msql2 into this release. Things are still subject to change, so I cannot make any promises. I left all the new features undocumented in case Bambi changes his mind about his intentions.
Included the patch for the lost table bug.
August 28, 1996
Made Msql and Msql::Statement easier subclassable. Thanks to jake@vivid.com (Jake Donham) for the idea.
July 29, 1996
Made t/msql.t more strict in the sense that it may now be run in parallel.
Added a test that checks for the "lost table" phenomenon.
Assorted minor improvements to the test suite.
July 9, 1996
A few cleanups in the XS logic to reduce the code below 700 lines again.
Added Msql->errmsg and Msql->quote (see man page)
Fixed display of control characters in pmsql.
Moved the ->as_string method into Msql/Statement.pm to make it accessible from anywhere.
Improved test script to output query in more places in case of an error.
July 1, 1996
pmsql: hostname were truncated at the first dot mostly.
pmsql: completion for drop table statement issued unnecessary warning.
Msql.xs eliminated warning from the cc compiler on solaris 5
Documented how to connect to a different than the default port in the manpage.
Added gethostinfo, getserverinfo, and getprotoinfo as methods on the database handle.
Added the admin API calls shutdown, creatdb, dropdb, reloadacls.
Internally all constants are implemented in the standard uppercase subroutines way again. Lowercase is supported but not recommended. It was not a good idea to allow lowercase for caonstants too. It helps nobody and confuses others. But it's no big deal. In the name of efficiency, write CHAR_TYPE, INT_TYPE, etc. again!
Prepared a few things for msql 2.0, but don't support it yet, because it is broken on my development platform. The tests though, they should perform work for msql 2.0.
June 7, 1996
Improved pmsql's as_string method and documented it. Control characters and backslashes are escaped now.
Improved the manpage wrt listfields. Added precanned reply to the FAQ.
June 1, 1996
Made all methods lowercase internally and ignore case to enable backwards compatibility.
Added basic completion to the pmsql monitor.
Added a fetchhash method. Benchmark in the FAQ.
May 28, 1996
Added pmsql(1), version 1.14 or so, which integrates relshow as a method. Deleted standalone relshow(1).
Added prototypes just to get rid of the warning (method prototypes are ignored by perl 5.002 anyway).
Added an example of ListFields usage to msql.t.
Deleted the (anyway outcommented) Tie code in Msql.pm as it won't come back for sure.
April 30, 1996
Added a relshow(1) clone to the distribution just to have a sample program that actually does something.
March 15, 1996
Metadata access simplified: in array context a statement handle method does not return an array reference but the array itself.
Non-selects now return 1 on success instead of a worthless statement handle.
The outdated FAQ is not in the distribution anymore.
Fixed the manpages, so perl5.002's pod2man doesn't grok.
Added 7 tests to test the above mentioned changes.
Simplified Makefile.PL with VERSION_FROM. Needs perl5.002.
October 1, 1995
Fixed a serious bug. Since we do not store $db->{SOCK} and family anymore, it's legal that the hash-entry doesn't exist. 1.02 returned the not very helpful message "Couldn't read svp".
July 30, 1995
Fixed a few documentation bugs. Reintroduced -lmisc in Makefile.PL. Introduced $Msql::QUIET, a variable to suppress error messages from the msql daemon when $^W is true (i.e. when the -w switch is on)
July 8, 1995
Implemented a new interface that uses methods to access the object attributes. The resulting version is a hermaphrodite that is fully backwards compatible, but also implements the new interface. Future versions will not support the old interface anymore. The reason for the change is a greater flexibility and also greater efficiency.
Fixed a memory leak (Thanks to Dan McGuirk <mcguirk@indirect.com> for the report)
June 4, 1995
Introduced the handling of NULL fields that comes with msql 1.0.6. That should be a fully backwards compatible solution. (Thanks to Dirk vanGulik <Dirk.vanGulik@jrc.it> for giving it a go)
IQuery is gone, it turned out to be of no use at all. FastQuery is still there.
The Msql::Tie package has gone in a different file and is still considered experimental. If you have an opinion about it, tell me, if not, don't ever use this file.
April 22, 1995
Introduced masking of single ticks (') within the Insert method.
Feb 28, 1995
Connect(host, db) now connects to UNIX socket when hostname is the null string. Did only so, if called as Connect() followed by SelectDB(db).
Added a bench/ directory with two sample benchmarks.
$sth-{HOST}, $sth->{DATABASE}, and $sth->{SOCK} depreciated (will not supported in future releases). Use their $dbh equivalent instead!
Feb 27, 1995
Fetching of metadata via tie() on demand with caching for TieQuery and TieListFields implemeted. Turns out to be a performance PENALTY, as the benefits of tie are eaten by the tie overhead. As a byproduct new methods are available: $sth->fetchinternal(KEY), $dbh->FastQuery(), $dbh->IQuery(), $dbh->FastListFields(). The performance gain is much smaller than expected. Will have to be reevaluated with perl5.001. The new functions are only important for testing purposes, esp. for the implementation of the DBI switch.
Insert() now checks the field type instead of the input type when it decides, if a value should be enclosed in ticks. Insert is faster now, but will be much faster, I suppose, once closures are available in perl.
Feb 13, 1995
$sth->{LENGTH} added.
About August 1994
First alpha release.