[SQLObject] Re: MySQL connection encoding problem

[Robert S. <rsz...@mu...>]

Oleg Broytmann wrote:
> On Thu, Apr 27, 2006 at 08:32:58AM +0200, Robert Szefler wrote:
> 
>>>  You can set the encoding in .my.cnf. Thus you don't need to patch
>>>SQLObject.
>>
>>I want SQLObject's (which is a high-level interface) behavior to be 
>>deterministic and not dependent on site configuration. Again, if it can 
>>be, why won't it? Can't we just fix the f*d up behavior?
> 
>    The behavior is not f*d up. ;)

The libmysql's is IMHO, I'm not (yet) complaining about SQLObject - it's 
still in the making after all. Depending on arbitrary host settings when 
you want it to just work will bite you sooner or later. Like testing on 
a host with encoding=utf8 in my.cnf and deploying where there's koi8-r 
or whatever. You don't want the my.cnf to be a part of your project's 
config taking into account that it might break a lot of other thigs. 
That's the f*d up design I talk about. The connection encoding should be 
hardwired for most, if not all, purposes. UTF8 is OK because it covers 
exactly the whole unicode range.

>>>>    cursor.execute('SET NAMES utf8')  # ADDED
> 
> 
>    I don't want to put such an arbitrary constant ("utf8") into the code.
> Why not "koi8-r" - that's what I use?

Recoding would take care of this. You already need it when you define 
UnicodeCols. Second one certainly wouldn't hurt and it would give the 
system a well-defined behavior.

I admit my "patch" is a bit of a dirty hack, but this should be solved.