On Tue, 25 Apr 2017 07:27:29 -0400, "Felix Antonius Wilhelm Ostmann via
RT" <bug-Text-CSV_XS@rt.cpan.org> wrote:
Show quoted text> Queue: Text-CSV_XS
> Ticket <URL:
https://rt.cpan.org/Ticket/Display.html?id=121350 >
>
> Am Di 25. Apr 2017, 06:04:44, h.m.brand@xs4all.nl schrieb:
> > > Ok, i understand that approach, but i would suggest the saver way:
> > > make the modul work with default options in 100% of input data and
> > > for someone who needs the speed add something like a option to
> > > disable the save way.
> >
> > Absolutely not. The way it currently works *is* safe. I *know* of
> > processes that *depend* on this behavior.
>
> OK, this is a point where religion wars can start ...
>
> I get input from customers and end up with wrong data. In my opinion
> a good modul should save me with the default options.
A good module offers not only basic functionality, but also a test
suite and documentation that covers all options. Most options are
implemented as attributes and stem from direct wishes from users.
They were not all a result of my devious mind. (Some are though)
Keep in mind that this module is written with SPEED (yes, in capitals)
is a primary goal (after correctness of course).
If *you* choose to use bind_column because of speed (which you should,
as it can cause a lot of speed-up), you inherently choose to take the
side-effects that come with it (and are documented).
If you would not use bind_columns, you could check the size of the row
inside your parsing loop. If you choose to use bind_columns, with the
current implementation and documented side-effects, you can not state
that it has erroneous behavior. There are (many) alternatives to work
around that issue.
Show quoted text> I see that at this point it is not possible to switch that anymore
> (because somone might depend on the current behavior).
not might. I'm very sure about it.
Even if it were possible, I would not. It is not a rare case that data
is set up so that trailing columns are expected to be copies of the row
above.
Show quoted text> > > In my case it was horrible wrong to have values from old rows and
> > > i guess that can hit others easily too.
> >
> > In that case, you might be (very) interested in the new strict
> > attribute that just got pushed to github. You can checkout and test
> > if that fits your needs. This option was on my TODO list for a long
> > time, but your ticket triggered me to actually implement and test
> > it
>
> yeah, works as expected and is my favorite beloved possibility to
> deal with ambignious data :) don't accept it! If only this could be
> the default behavior ... ;)
The only default behavior that I seriously want to change - but cannot
due to backward incompatibilities - is the default for the binary
attribute. That should have been default from the start. But alas, I
did not write the first version. As you will probably need a default
list of attributes like binary and auto_diag, having strict in there as
well shouldn't be too much of a strain
my $csv = Text::CSV_XS->new ({ binary => 1, strict => 1, auto_diag => 1 });
FWIW I also now implemented the same option in Text::CSV for perl6,
where it *also* defaults to False. The invocation is simpler though, as
in the perl6 module I was able to make binary the default
my $csv = Text::CSV.new (:strict, :auto-diag);
I'll start the test phase and will try to get out this new version asap
for you.
Show quoted text> > $ cat rt121350.pl
> > use 5.18.1;
> > use warnings;
> > use Text::CSV_XS;
> >
> > my $data = <<"__CSV__";
> > owner,admin
> > 1,1
> > 2
> > 3,
> > 4,4
> > __CSV__
> >
> > open my $fh, "<", \$data;
> >
> > my $csv = Text::CSV_XS->new ({ auto_diag => 1, strict => 1 });
> > my %row;
> > $csv->bind_columns (\@row{@{$csv->getline ($fh)}});
> > while ($csv->getline ($fh)) {
> > printf "owner: %s ; admin: %s\n", $row{owner}, $row{admin};
> > }
> > $ perl rt121350.pl
> > owner: 1 ; admin: 1
> > # CSV_XS ERROR: 2014 - ENF - Inconsistent number of fields @ rec 3
> > pos 3 field 1
--
H.Merijn Brand
http://tux.nl Perl Monger
http://amsterdam.pm.org/
using perl5.00307 .. 5.25 porting perl5 on HP-UX, AIX, and openSUSE
http://mirrors.develooper.com/hpux/ http://www.test-smoke.org/
http://qa.perl.org http://www.goldmark.org/jeff/stupid-disclaimers/
