Checking a little more...

This gives the error: 
http://head.bibkat.no/cgi-bin/koha/opac-search.pl?tag=code&q=code

This works: 
http://head.bibkat.no/cgi-bin/koha/opac-search.pl?q=code

The problem seems to stem from Lingua::Stem::Snowball and specifically the call to it on line 741 of C4/Search.pm - in the _build_stemmed_operand function: 

    my $stemmer = Lingua::Stem::Snowball->new( lang => $lang,
                                               encoding => "UTF-8" );

For Norwegian, $lang holds "nb-NO". Replacing that temporarily with a literal value of "no" solves the problem. 

The documentation (e.g. http://search.cpan.org/dist/Lingua-Stem-Snowball/lib/Lingua/Stem/Snowball.pm) for Lingua::Stem::Snowball->new says this about the lang parameter: 

"lang: An ISO code taken from the table of supported languages, above." 

The table referenced only lists two-letter language codes, specifically these: da, nl, en, fi, fr, de, hu, it, no, pt, ro, ru, es, sv, tr. 

The syspref QueryStemming influences the observed behviour. When the chosen language in the OPAC is Norwegian (nb-NO), this works regardless of what QueryStemming is set to: 

http://head.bibkat.no/cgi-bin/koha/opac-search.pl?q=code

Clicking on a tag in the result list gives this URL: 

http://head.bibkat.no/cgi-bin/koha/opac-search.pl?tag=code&q=code

When QueryStemming = "Try" this fails with the "Language ... does not exist" message. 

When QueryStemming = "Don't try" a result list is shown, as expected. 

Still confused...

In 3.4.6 it works for Swedish, regardless of what QueryStemming is set to, but not for Norwegian. There's a difference between 3.4.6 and current master:

In 3.4.6 $lang is taken from C4::Templates::getlanguagecookie: 

389 sub getlanguagecookie {
390     my ($query) = @_;
391     my $lang;
392     if ($query->cookie('KohaOpacLanguage')){
393         $lang = $query->cookie('KohaOpacLanguage') ;
394     }else{
395         $lang = $ENV{HTTP_ACCEPT_LANGUAGE};
396         
397     }
398     $lang = substr($lang, 0, 2);
399 
400     return $lang;
401 }

In current master it comes from C4::Templates::getlanguage: 

311 sub getlanguage {
312     my ($query, $interface) = @_;
313 
314     # Select a language based on cookie, syspref available languages & browser
315     my $is_intranet = $interface eq 'intranet';
316     my @languages = split(",", C4::Context->preference(
317         $is_intranet ? 'language' : 'opaclanguages'));
318 
319     my $lang;
320 
321     # cookie
322     if ( $query->cookie('KohaOpacLanguage') ) {
323         $lang = $query->cookie('KohaOpacLanguage');
324         $lang =~ s/[^a-zA-Z_-]*//; # sanitize cookie
325     }
326 
327     # HTTP_ACCEPT_LANGUAGE
328     unless ($lang) {
329         my $http_accept_language = $ENV{ HTTP_ACCEPT_LANGUAGE };
330         $lang = accept_language( $http_accept_language,
331             getTranslatedLanguages($interface,'prog') );
332     }
333 
334     # Ignore a lang not selected in sysprefs
335     $lang = undef  unless first { $_ eq $lang } @languages;
336 
337     # Fall back to English if necessary
338     $lang = 'en' unless $lang;
339 
340     return $lang;
341 }

In 3.4.6 "sv-SE" is reduced to "sv" on line 398 and this is a valid language for Lingua::Stem::Snowball. "nb-NO" is reduced to "nb", which is not valid in the eyes of Lingua::Stem::Snowball. C4::Templates::getlanguage returns the unreduced strings, neither of which is valid for Lingua::Stem::Snowball.

Hi Magnus,

we are working on similar problems.  One site has es-ES language pack installed, 6.02.  All of the English tags work, but when switching to Spanish, the tags won't work if there is only a single item returned.

Error:
Language 'es-es' does not exist 

At another site, with only English, none of the tags works:
Error:
Language 'en-us,en;q=0.5' does not exist

Turning off Stemming, does allow the tags to work.

Best regards,
Albert

Looks to me like there are two problems here:

> Error:
> Language 'es-es' does not exist 
> 
> At another site, with only English, none of the tags works:
> Error:
> Language 'en-us,en;q=0.5' does not exist

1. These "long" language codes need to be reduced to two-letter codes that Lingua::Stem::Snowball will accept. 

2. Why does the error from the stemmer only show up for opac-search.pl?tag=code&q=code and not for opac-search.pl?q=code ?

(In reply to comment #5)
> 
> 1. These "long" language codes need to be reduced to two-letter codes that
> Lingua::Stem::Snowball will accept. 


I confirm Lingua::Stem::Snowball needs a 2-characters lang.

It can be simply corrected in Search.pm : 
my $stemmer = Lingua::Stem::Snowball->new( lang => lc(substr($lang, 0, 2)), encoding => "UTF-8" );

See : http://annocpan.org/~CREAMYG/Lingua-Stem-Snowball-0.952/lib/Lingua/Stem/Snowball.pm

(Sorry, I don't have time to create a patch).

(In reply to comment #6)
> my $stemmer = Lingua::Stem::Snowball->new( lang => lc(substr($lang, 0, 2)),

If only it were that simple... ;-) 

For the two variants of Norwegian, this would result in
nb-NO -> nb
nn-NO -> nn
but neither of these are valid for Lingua::Stem::Snowball, which expects Norwegian to be coded as "no". 

I have done a quick and dirty hack for my installations which looks like this: 

 $lang = substr($lang, 0, 2);
 if ($lang eq 'nb' || $lang eq 'nn') {
     $lang = 'no';
 }

but that is probably not the way to go...

The solution of Fridolyn Somers works on my server for German and French, and I expect it to work for Italian as well.

I think the most pragmatic solution would be to implement the code of Magnus Enger, because Norwegian seems to be the only exception for the list of languages supported by Snowball, see:

http://annocpan.org/~CREAMYG/Lingua-Stem-Snowball-0.952/lib/Lingua/Stem/Snowball.pm
 
 Language   | ISO code 
-----------------------
 Danish     | da       
 Dutch      | nl       
 English    | en       
 Finnish    | fi       
 French     | fr       
 German     | de       
 Hungarian  | hu       
 Italian    | it       
 Norwegian  | no       
 Portuguese | pt       
 Romanian   | ro       
 Russian    | ru       
 Spanish    | es       
 Swedish    | sv       
 Turkish    | tr

Created attachment 8888 [details] [review]
[SIGNED-OFF] Bug 7445 - Clicking on a tag gives "Language ... does not exist"

To test:
- Syspref QueryStemming = Try
- Install Norwegian bokmål:
  cd misc/translator/
  perl translate install nb-NO
- Go to Home › Administration › System Preferences > I18N/L10N
  and enable "Norsk bokmål(nb-NO)" for opaclanguages as well as
  setting opaclanguagesdisplay = Allow
- Make sure you have selected "Norsk bokmål" as the active language
  in the OPAC
- Find a record that has a tag (which does not contain any digits)
- Click on the tag and see that you get the error in the title of
  this bug
- Apply the patch
- Click on the tag again and the error should be gone

Signed-off-by: Katrin Fischer <Katrin.Fischer.83@web.de>
Easy to test with a great test plan. Works nicely.

This patch just does a substring call to give Snowball what it needs, and adds handling for the only exception we currently know about.  If we start finding more exceptions, we may want to revisit our methodology, but for now, I think having 1 special case is okay.

Passed QA.

QA comment: this code should, in theory, not be accepted because it's a workaround using something hardcoded. But Marc Veron, in comment 8 seems to prove that's really the only exception, and it's more a problem with Lingua::Stem than a problem in Koha itself, so OK to push

... patch pushed

This bug will be included in the Koha 3.6.5 release.