Re: AnyData patches
On Wed, 05 Sep 2012 11:04:23 +1000, Sven Dowideit svendowid...@home.org.au wrote: yup :) http://github.com/SvenDowideit/AnyData tl;dr; -- shall I take this off-list? A random number of remarks … I have a *lot* of modules - for testing - installed, but still run into new ones, and neither is mentioned when running Makefile.PL t/01distributtion.t .. skipped: Test::Distribution not installed t/01pod_snippets.t ... Can't locate Test/Pod/Snippets.pm And Test::Pod::Snippets depends on a lot of unneeded stuff like Object::InsideOut and Module::Locate, two I also had not installed. If you really want to depend on those just locally, disable these tests in the distribution if the module cannot be loaded. Even after having the modules installed, I get t/01distributtion.t .. skipped: Test::Distribution not installed t/01pod_snippets.t ... skipped: Need to do more work fixing pod code snippets so I guess it is better to just drop that test and use Test::Pod and Test::Pod::Coverage instead: t/00_pod.t -- #!/usr/bin/perl use strict; use warnings; use Test::More; eval use Test::Pod 1.00; plan skip_all = Test::Pod 1.00 required for testing POD if $@; all_pod_files_ok (); -- t/01_pod.t -- #!/usr/bin/perl use strict; use warnings; use Test::More; eval use Test::Pod::Coverage tests = 1; plan skip_all = Test::Pod::Coverage required for testing POD Coverage if $@; pod_coverage_ok (AnyData, AnyData is covered); -- I miss an examples folder. Docs are fine, but examples are great Makefile.PL requires 5.004. I'd say that by now it is safe to require at least 5.6.2, but when also using DBI, require 5.8.1. Additional, you can then add: * use_warnings Add 'use warnings' to all modules. (This will require perl 5.6) MYMETA.json/yml should include the git repo under resources Spell-checking - after filtering known words. I can give you the tools on IRC DBD-AnyData-git 529 pod-spell-check --aspell --ispell ok 1 - AnyData.pm ok 2 - AnyData/Format/Base.pm ok 3 - AnyData/Format/CSV.pm ok 4 - AnyData/Format/FileSys.pm ok 5 - AnyData/Format/Fixed.pm ok 6 - AnyData/Format/HTMLtable.pm ok 7 - AnyData/Format/Ini.pm ok 8 - AnyData/Format/Mp3.pm ok 9 - AnyData/Format/Paragraph.pm ok 10 - AnyData/Format/Passwd.pm ok 11 - AnyData/Format/Pipe.pm ok 12 - AnyData/Format/Tab.pm ok 13 - AnyData/Format/Text.pm ok 14 - AnyData/Format/Weblog.pm ok 15 - AnyData/Format/XML.pm ok 16 - AnyData/Storage/File.pm ok 17 - AnyData/Storage/FileSys.pm ok 18 - AnyData/Storage/PassThru.pm ok 19 - AnyData/Storage/RAM.pm ok 20 - AnyData/Storage/TiedHash.pm ok 21 - sandbox/genMETA.pm 1..21 ok 1 - CommonMistakes not ok 1 - AnyData.pm # Failed test 'AnyData.pm' # at /pro/bin/pod-spell-check line 96. # got: ''additonal' = (additional additionally atonal ordinal) # 'adTie' = (ad Tie ad-Tie Addie Artie audit date adieu edit added ate artier EDT attire auntie eddied adze ante anti tattie admit Eddie Addie's Artie's) # 'containng' = (containing contain contains container contained counting contagion continuing canting condoning containers continua continue container's) # 'DTD' = (DDT TDD STD TD DD DOD DAD DID DUD DMD DTP DVD ETD LTD IT'D) # 'insdie' = (inside ins die ins-die insider onside indie insides inset Indies indies indite incite Inst inst India inside's) # 'itslef' = (itself outsell oneself Edsel Adolf outsells dissolve) # 'Malcom' = (Malcolm Talcum LCM Mailbomb Maalox Malacca Holcomb Welcome Mulct Melanoma Slocum Amalgam Locum Glaucoma Magma) # 'mltiple' = (multiple multiply multiplier multiples multiplied multiplies multiple's) # 'possiblity' = (possibility possibly possibility's possible) # 'sqaure' = (square squarer squire scare secure squared squares Sucre sure square's) # 'unpdated' = (undated updated) # 've' = (vie V v veg vet Be Ce be E e vex VA VI Va vi we VD VF VG VJ VP VT Vt vb vs DE Fe GE Ge He IE Le ME Me NE Ne OE PE Re SE Se Te Xe he me re ye Ave Eve ave eve V's I've)' # expected: '' ok 2 - AnyData/Format/Base.pm ok 3 - AnyData/Format/CSV.pm ok 4 - AnyData/Format/FileSys.pm not ok 5 - AnyData/Format/Fixed.pm # Failed test 'AnyData/Format/Fixed.pm' # at /pro/bin/pod-spell-check line 96. # got: ''formating' = (for mating for-mating formatting formation forming formulating formalin formative foaming firming fomenting formatting's floating fating format mating reformatting farming farting footing fording formfitting)' # expected: '' not ok 6 - AnyData/Format/HTMLtable.pm # Failed test 'AnyData/Format/HTMLtable.pm' # at /pro/bin/pod-spell-check line 96. # got: ''Sisk' = (Disk Sis Sask Sick Fisk Risk Silk Sink Sis's)' # expected: '' not ok 7 - AnyData/Format/Ini.pm # Failed test 'AnyData/Format/Ini.pm' # at
Re: More problems with bound parameters in DBD::ODBC
On 04/09/12 23:15, Tim Bunce wrote:
On Tue, Sep 04, 2012 at 11:28:17AM +0100, Martin J. Evans wrote:
This issue cropped up because runrig posted a DBI bind_param_inout trick node
(http://perlmonks.org/?node_id=989136) on perl monks which when I ran through DBD::ODBC
did not work. The code is basically:
my @cols = qw(foo bar);
my $sql = SQL;
SELECT :foo, :bar
SQL
my $sth = $dbh-prepare($sql);
my %hsh;
for (@cols) {
$sth-bind_param_inout( $_ = \$hsh{$_}, 0 );
}
$hsh{foo} = 'abc';
$hsh{bar} = 123;
$sth-execute();
while (my @arr = $sth-fetchrow_array) {
print @arr\n;
}
$hsh{bar} = 456;
$sth-execute();
while (my @arr = $sth-fetchrow_array) {
print @arr\n;
}
which should output:
abc 123
abc 456
but actually outputs:
Use of uninitialized value $arr[0] in join or string at /tmp/x.pl line 33.
123
Use of uninitialized value $arr[0] in join or string at /tmp/x.pl line 39.
456
The problem is that DBD::ODBC binds the parameter initially when
bind_param_inout is called then when execute is called it needs to determine if
the parameters need to be rebound (if something significant has changed). It
uses the following test:
if (SvTYPE(phs-sv) != phs-sv_type /* has the type changed? */
|| (SvOK(phs-sv) !SvPOK(phs-sv)) /* is there still a string? */
|| (SvPVX(phs-sv) != phs-sv_buf) /* has the string buffer moved? */
) {
/* rebind the parameter */
}
I have some issues with this (in addition to it not working):
1. DBD::ODBC always calls SvUPGRADE(phs-sv, SVt_PVNV) on output parameters so
the type is unlikely to change.
Anything can happn to the sv between the bind_param_inout and the execute.
Yes, I realised that but in the majority of cases (all of them in the test
suite for a start) the test always ends up is if 6 != 6 (Svt_PVNV) - more
coverage required.
2. DBD::ODBC always calls SvGROW on output parameters to grow them to
28 chrs (some magic about 28 I don't know) to avoid mutation in most
cases. As a result, if you change the test code so the first param is
bigger than 28 chrs it works.
I vaguely recall some magic about the value, but not the specifics.
It just meant it looked a little confusing when strings 28 chrs worked and ones
28 did not.
3. I don't understand what the (SvOK(phs-sv) !SvPOK(phs-sv)) is
for. I know what those macros do but not why that test is present. Any
ideas? It is in other DBDs too.
That's saying rebind if the sv is defined but doesn't have a string.
I think the not defined case is (or should be) handled elsewhere, so
the test is mainly to check that the sv still contains a string.
(Before then testing SvPVX()).
Of course, been away on holiday for too long and not back in the swing of it
properly.
4. I'm unsure how to make this work although if I simply add a test to
say has SvOK(phs-sv) changed since binding it makes this example
work. Anyone any ideas if this is sufficient?
Something along those lines should be fine. Looking at DBD::Oracle,
in which I probably first wrote that code (perhaps around 1996 :-)
I see a few lines higher up:
/* is the value a null? */
phs-indp = (SvOK(sv)) ? 0 : -1;
so I presume that oracle 'indicator parameter' handles the 'is currently
null' case, so the if() statement only has to deal with the not-null case.
Oracle's need for rebinding probably differs in subtle ways from ODBC's.
In this case ODBC has already bound it with an indicator saying is NULL and
needs to rebind it. I changed it to test if SvOK had changed since the last
bind.
This particular situation is transitioning from an undef to defined.
And this particular kind of undef has SvTYPE == 0, in case that's relevant.
Hope that helps.
Tim.
Thanks for the help.
This is released on the CPAN as 1.40_1 + some other fixes.
Martin
--
Martin J. Evans
Easysoft Limited
http://www.easysoft.com
