Better conformance for UNIMARC Authorities Encoding
encoding is on 8th character and not 9th
This commit is contained in:
parent
03c51600a0
commit
9d52c46dda
|
|
@ -255,7 +255,7 @@ sub SetMarcUnicodeFlag {
|
|||
$marc_record->leader($leader);
|
||||
} elsif ($marc_flavour =~/UNIMARC/) {
|
||||
my $string;
|
||||
my ($subflength,$encodingposition)=($marc_flavour=~/AUTH/?(21,8):(36,22));
|
||||
my ($subflength,$encodingposition)=($marc_flavour=~/AUTH/?(21,9):(36,22));
|
||||
$string=$marc_record->subfield( 100, "a" );
|
||||
if (length($string)==$subflength) {
|
||||
$string = substr $string, 0,$subflength if (length($string)>$subflength);
|
||||
|
|
|
|||
Loading…
Reference in New Issue