[OPEN-ILS-GENERAL] Language problems in Evergreen

[Dan Scott]

On Mon, Feb 21, 2011 at 11:11:02AM +0400, Armen wrote:
> Dear Evergreen staff
> 
> I am from the Fundamental Scientific Library of Armenia. We have recently
> upgraded our Evergreen 1.6.0.6 version to 2.0.1 version and currently we are
> having some problems concerning our databases in Armenian language.
> Particularly, per se 
> 
> I want to remark that the same problems occurred when we  created a new
> database with command "createdb -T template0 -E UNICODE evergreen".

I don't think this command is at the root of your problems, for what
it's worth.

> 
> *         we can't register our users in Armenian (on the contrary to
> English)

I have reproduced this problem with Evergreen 2.0.1 and Evergreen trunk
running on both OpenSRF 1.6.2 and OpenSRF trunk.

On OpenSRF 1.6.2, the problem (as suggested by Galen Charlton) happens
when the FreezeThaw Perl module tries to "thaw" an OpenSRF message
containing Unicode characters outside of the basic multilingual plane.
Characters like 'èçàö' work fine, but your Armenian examples fail.

I was certain that this used to work (it works for you in Evergreen
1.6.0.6, correct?) and tried rolling back the version of FreezeThaw to
0.43 from 2002, which is what we had been using on Debian Lenny.
However, that did not resolve the problem - so how that used to work
remains a mystery.

On OpenSRF trunk (to become 2.0), FreezeThaw is no longer used and I
found and fixed a bug where the Unicode data was not decoded and encoded
correctly. So... if we release OpenSRF 2.0, that offers one way of
resolving your problem.
 
> *         and also we can't search our databases in Armenian.  When we are
> doing search in Armenian and while typing one letter with a (*) asterix (the
> results of the search are being retrieved in Armenian), however when we are
> doing search only with Armenian letters we are not able to retrieve the
> search results. 
> 
> *         Also when we want to search the user, we can do it only with bar
> code, however we can't do the search only by name and last name of the user
> (especially in Armenian).

These probably have the same underlying issue as when registering the
users.
 
> *         We also have some language issues concerning the interface of the
> circulation database, some part of it is in English and some party of it is
> in Armenian (that is the menu tab of the users registration is in Armenian,
> however when it comes to the registration details of the users it is in
> English, to the contrary of the previous version, here every menu was in
> Armenian). 

In 2.0, there are unfortunately some strings that have been hardcoded
in English. Bill Erickson has a working proof-of-concept for providing
translation support for Template::Tookit templates (which is what the
patron registration interface and many other interfaces have been built
on in 2.0) but we have not yet integrated that support into Evergreen.
Hopefully we can get some help in converting the existing interfaces
over to provide translation support once the option is available; our
French users would also appreciate this option!