Re: multi-language web application: is it possible?

Elim Qiu schrieb:
> I have many tables like the table Person:below, in mysql database.
>
> person_id, first_name,last_name, mi, gb_first_name, gb_last_name,
> b5_first_name, b5_last_name, gender, dob
>
> where different columns storing strings in different encodings. At
> anytime, a web user can switch the language and the application will get
> the values in the right columns to generate web pages. The purpose of
> Multi-language tables is to make multilanguage dynamic content
> management easier for web applications.  For example, to add a person
> record, the user enter the English name, then switch the session
> language to gb2312, enter the Chinese name in gb2312, and then switch
> the session language to big5, enter the name in big5. And then commit
> the data into the database. The whole thing sounds complicated but can
> be treated as a pattern and let a framework to take care of those and
> the code can be as clean as a single language app. I actually have the
> framework that works well for me with mysql database.

It would be easier to normalize the tables and have a referral.
Also person names arent usually localized - at least this would
be very uncommon. If you want you can transcribe them when you
output it (e.g. replace umlauts with their dual char ascii
representation)

> I'm trying to port the app to pgsql database but got trouble doing so. I
> can read and set a row with multiple languages (in some paricular
> cases), but cannot do queries like  (gb_first_name = 'A' and
> b5_first_name = 'B') with A a gb2312 string, B a big5 string. The tables
> are of unicode encoded, and the dbclient encoding is set to GBK. The
> application's char set are selectable by user among iso-8859-1,gb2312
> and big5.

the language and their encoding are more or less orthogonal. So if
you settle for one common encoding which covers all the language
you want to use, then its easy to have all the language words
side by side. General unicode (in its utf-8 representation) is
used in postgres for such a general case.


> I didn't do anything about language encoding in mysql database, it just
> worked for me. At least with english, gb2312 and big5 altogether in a
> table like table Person above. I noticed that (english, gb2312, big5,
> Jp) cannot work together even in mysql database.  My approach seems fine
> with most western languages

Well it might work but it might not work as expected. Beside the
encoding, you also have a collating order to obey. This is very
depending on the language (and the portion of the charset it
uses) so this would not even work in mySQL or other databases
w/o any specification.

In PG, assuming you would have a table with the texts, it would
carry the language_id too and you could provide a functional index
which would take care of the sorting of the individual language
based on the language id.


> So after all such experimental work, I still don't know how to make a
> real multi language web app such that the languages are switchable
> within the same session.

I'd highly recommend restructuring the tables to be more flexible
and to have a faster and cleaner approach. Joins do not hurt so
much in postgres so use them to your advantage.

Regards
Tino