Ed,
On Fri, Jun 17, 2011 at 10:53:00AM +0100, Edmund Chamberlain wrote:
> Firstly, hello! Its my first time posting and possibly somewhat
> predictably with a call for help with Unicode stuff.
Ah, yes...
> I've just checked the archive and seen this thread and am having a
> similar problem, a badly encoded character is causing a while loop
> through MARC::Batch->next to crash out with:
>
> utf8 "\x87" does not map to Unicode at
> /usr/lib64/perl5/5.8.8/x86_64-linux-thread-multi/Encode.pm line 173.
>
> I've tried pasting Al's modified decode subroutine and package to the
> script, but it is still failing. One of the offending records is
> isolated and attached.
>
> Any suggestions welcome with regards to further modifying the sub or
> alternatives to MARC::Batch->next would be welcome. For the scope of the
> project, I'm limited to large batch files of Marc21.
You could try MARC::Loop, which doesn't care about (or even detect)
character (mis)codings in any way -- it just sees everything as raw
bytes. Then you can go in and wipe out (or fix) bad byte sequences,
something like this:
use MARC::Loop qw(marcloop marcbuild);
marcloop {
# This code is called once for each record read from standard input
my ($leader, $fields, $rawref) = @_;
if ($$rawref =~ /[^\x00-\x7f]/) {
# The record contains one or more non-ASCII bytes
foreach my $field (@$fields) {
my ($tag, $valref) = @$field;
$strref =~ s{
( # Valid byte sequences for a single character:
[\x{00}-\x{7f}]
| [\x{c2}-\x{df}][\x{80}-\x{bf}]
| \x{e0} [\x{a0}-\x{bf}][\x{80}-\x{bf}]
| [\x{e1}-\x{ec}][\x{80}-\x{bf}][\x{80}-\x{bf}]
| \x{ed} [\x{80}-\x{9f}][\x{80}-\x{bf}]
| [\x{ee}-\x{ef}][\x{80}-\x{bf}][\x{80}-\x{bf}]
| \x{f0} [\x{90}-\x{bf}][\x{80}-\x{bf}]
|
[\x{f1}-\x{f3}][\x{80}-\x{bf}][\x{80}-\x{bf}][\x{80}-\x{bf}]
| \x{f4}
[\x{80}-\x{8f}][\x{80}-\x{bf}][\x{80}-\x{bf}]
)
|
(
# Oops! -- invalid byte sequence, we assume all
# non-ASCII bytes starting here are bad
[^\x00-\x7f]+
)
}{
if (defined $2) {
# Substitute a "fixed" version of the bad byte sequence
print STDERR "Fixing bad byte sequence in field $tag of
record $.\n";
fixed($2);
}
else {
# Leave it unchanged
$1;
}
}xeg;
}
print marcbuild($leader, $fields);
}
else {
print $$rawref;
}
} \*STDIN;
sub fixed {
my ($str) = @_;
# There are any number of actions you might take to "fix" invalid byte
# sequences; this is just one
return '?';
}
The downside is that MARC::Loop is a separate Perl module you'll have to
download from CPAN and install manually. I wrote it, so I'm biased, but
I think it's good for people who prefer to (or have to) work closer to
the raw MARC record. (Fixing miscoded records that MARC::Record et al.
couldn't handle was what motivated me to write it in the first place.)
Paul.
--
Paul Hoffman <nkui...@nkuitse.com>
