| CC: | cristiano.passerini [...] gmail.com |
| Subject: | DBD::CSV cannot read CSV files with dots on the first line |
DBD::CSV interprets the first line of a CSV file it reads as the list of
column names.
This interpretation fails when the column names contain dots.
Code example:
my $csvh = DBI->connect('dbi:CSV:')
or die "the CSV parser doesn't work (not installed?)\n";
my @colnames;
$csvh->{'csv_tables'}->{'input'} =
{
file => $csvfile,
sep_char => ','
};
my $sth = $csvh->prepare('SELECT * FROM input');
$sth->execute;
The execute statement will execute SQL::Statement->SELECT,
which invokes $self->verify_columns( $data,$eval, $all_cols );
to convert the column names into SQL::Statement::Column objects.
It is when creating these objects that names with dots are interpreted
as standing for $tablename.$columnname combinations, wrongly in this use
case, and the resulting shortened $columnname can no longer be found in
a hash, with the result that fetch_hashref and related return column
names with a values of undef.
It is not 100% clear to me where exactly this problem should be
rectified, but probably in DBD::CSV.
The problem is worsened by the fact that the documentation is unclear on
how to work around the bug: it does mention a facility to avoid
interpreting the first row as the column names, but I can't figure out
how to use it. This limitation should at least be mentioned.
This bug has been reported before, but to a different RT:
http://rt.perl.org/rt3//Public/Bug/Display.html?id=39466
and was rejected there for that reason.