Re: [Pypgsql-users] __unicodeConvert and sub-classing unicode

[Nicola L. <ni...@te...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[Disclaimer: not an official response, just some bystander who happens to have 
delved deep into the code of pyPgSQL, while resisting the urge to fork 
it. ;-) ]


> I've experimented with hacking PgSQL.py to use isinstance() in the 
> various places instead.  ie. Instead of:
> 
>  if type(obj) is UnicodeType:
>   do_something()

This way of checking types is deprecated, exactly because of the problem you 
encountered.


> I've changed it to:
> 
>  if isinstance(obj, UnicodeType):
>   do_something()

This is how it's supposed to be done, at this moment.


> In my case at least, this appears to fix things, but I haven't tested it 
> intensively.
> 
> I realise that sub-classing 'str' and 'unicode' isn't a normal thing for 
> most people to do.

You are definitely supposed to be able to do that.


> Is that the only reason that the 'is' operator was used, or is it more
> complicated than that? 

Just historical reasons, probably.


> If someone who better understands the inner workings of pyPgSQL could 
> clarify it for me, I'd appreciate it.  Otherwise I'll go ahead and 
> report it as a bug.

I'm not claiming to be an expert, but I didn't see anything in the code that 
needs the original unicode type, and nothing else. So yes, you should go 
ahead, IMO.


- -- 
Nicola Larosa - ni...@te...

Unchecked police and military power undermine personal security,
and increase the risk of terrorism attacks. Unchecked terrorism only
increases the risk of terrorism attacks. Statistically, if both are
possible, I am more at risk from the first, and am likely to lose more.
 -- Daniel Staal, August 2004

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)

iD8DBQFBfz/VXv0hgDImBm4RAhi5AKCAYlHCdAdirtQJRcjhGRlTYxtfFACgg5Ay
748fHMKPiUtq50djdy7gtbI=
=4piy
-----END PGP SIGNATURE-----