[Zope3-dev] Re: Some ZPT insights needed

[Martijn Faassen]

Andreas Jung wrote:
> Hi,
> 
> because Zope 2.10.2+ will use unicode as internal representation for
> page templates I am currently checking the options to get rid of
> UnicodeDecodeErrors for Zope 2.11 by providing some more intelligent
> conversion of non-unicode content. Basically I want if the content 
> returned by expressions within a page template if unicode or 
> non-unicode. If it is non-unicode it would be converted based on some 
> policy (to be specified) to unicode. This would avoid any kind of 
> UnicodeDecodeError (hopefully).
> This there a central place in the ZPT code were the evaluated 
> expressions are inserted into the rendered HTML or some internal 
> datastructure used for rending the ZPT? I just need to know where to 
> look at but I have no idea so far :-)

This kind of automagic unicode error defeating logic scares me. With 
Silva, we have a significant Zope 2 application that already takes care 
to output unicode to the page template engine. For a while we were using 
PlacelessTranslationService which has some hacks in place to do 
automagic conversion from unicode to an encoding. Granted this is the 
opposite direction than you are proposing, and my intuition is that your 
direction may be less scary, but I'm still scared. Not having unicode 
errors can make debugging of pure-unicode applications harder. Bugs 
sneak in pretty easily and you end up with non-pure content.

I'd therefore like it if there were a way, application-root specific, to 
turn off any auto-conversion behavior. Do you think that would be possible?

Regards,

Martijn