Webware for Python

Andreas Poisel schrieb:
> Importing from eggs in a servlet doesn't work when AutoReload is
> switched on.  Is this a known issue?  I tried Webware 0.9 with
> simplejson-1.1-py2.4.egg and json_py-3.4-py2.4.egg under Linux (Gentoo
> and Ubuntu).

I checked this and it seems you are right. I cannot even use KidKit if 
Kid is installed as an egg, and AutoReload is on.

I think (with my limited expertise) this is because AutoReload uses 
ImportSpy, ImportSpy uses ihooks, and ihooks does not support eggs.
At least, if I do the following

import ihooks
ihooks.install(ihooks.ModuleImporter())

then I cannot import any python eggs afterwards.

Maybe Ian Bicking or others with more expertise can confirm this and 
propose a solution (replace ihooks with something different)?

-- Christoph

Christoph Zwerschke wrote:
> I checked this and it seems you are right. I cannot even use KidKit if 
> Kid is installed as an egg, and AutoReload is on.
> 
> I think (with my limited expertise) this is because AutoReload uses 
> ImportSpy, ImportSpy uses ihooks, and ihooks does not support eggs.
> At least, if I do the following
> 
> import ihooks
> ihooks.install(ihooks.ModuleImporter())
> 
> then I cannot import any python eggs afterwards.
> 
> Maybe Ian Bicking or others with more expertise can confirm this and 
> propose a solution (replace ihooks with something different)?

One thing that's been discussed is simply dropping the ihooks part -- 
polling sys.modules for changed modules works without any tricks.

(Of course, you can also turn off installing of zip files, and install 
the unpacked egg -- this gives you better tracebacks, for instance)

-- 
Ian Bicking  /  ianb@...  /  http://blog.ianbicking.org

Ian Bicking wrote:
> One thing that's been discussed is simply dropping the ihooks part -- 
> polling sys.modules for changed modules works without any tricks.
> 
> (Of course, you can also turn off installing of zip files, and install 
> the unpacked egg -- this gives you better tracebacks, for instance)

Yes, but it would also be good to make ihooks work with Eggs. For 
instance, Kid (kid.lesscode.org) suffers from the same problem with 
ihooks. Do you think there is a chance to make ihooks compatible with 
Eggs? I already asked on c.l.python but nobody answered.

-- Christoph

Christoph Zwerschke wrote:
> Ian Bicking wrote:
> 
>> One thing that's been discussed is simply dropping the ihooks part -- 
>> polling sys.modules for changed modules works without any tricks.
>>
>> (Of course, you can also turn off installing of zip files, and install 
>> the unpacked egg -- this gives you better tracebacks, for instance)
> 
> 
> Yes, but it would also be good to make ihooks work with Eggs. For 
> instance, Kid (kid.lesscode.org) suffers from the same problem with 
> ihooks. Do you think there is a chance to make ihooks compatible with 
> Eggs? I already asked on c.l.python but nobody answered.

In theory it should work; loading from zips is implemented with ihooks. 
  But it requires that any added ihook defer to existing ihooks, and if 
it is doing file lookup it needs to use the PEP-specified technique.

-- 
Ian Bicking  /  ianb@...  /  http://blog.ianbicking.org

Ian Bicking wrote:
> In theory it should work; loading from zips is implemented with ihooks. 
>  But it requires that any added ihook defer to existing ihooks, and if 
> it is doing file lookup it needs to use the PEP-specified technique.

Are you refering to PEP 302? As far as I understand, importing from zips 
is implemented with "new style" hooks which are incompatible with 
ihooks. See this passage at the end:

"Existing __import__ hooks will not invoke new-style hooks by magic, 
unless they call the original __import__ function as a fallback. For 
example, ihooks.py, iu.py and imputil.py are in this sense not forward 
compatible with this PEP."

Maybe this is the reason why it does not work? Instead of ihooks, should 
one use the new style hooks of PEP 302? I must admit I don't see through 
this whole importing mess.

-- Christoph

Christoph Zwerschke wrote:
> Ian Bicking wrote:
> 
>> In theory it should work; loading from zips is implemented with 
>> ihooks.  But it requires that any added ihook defer to existing 
>> ihooks, and if it is doing file lookup it needs to use the 
>> PEP-specified technique.
> 
> 
> Are you refering to PEP 302? As far as I understand, importing from zips 
> is implemented with "new style" hooks which are incompatible with 
> ihooks. See this passage at the end:
> 
> "Existing __import__ hooks will not invoke new-style hooks by magic, 
> unless they call the original __import__ function as a fallback. For 
> example, ihooks.py, iu.py and imputil.py are in this sense not forward 
> compatible with this PEP."
> 
> Maybe this is the reason why it does not work? Instead of ihooks, should 
> one use the new style hooks of PEP 302? I must admit I don't see through 
> this whole importing mess.

Mmmm... import hooks have scared/scarred me enough that I haven't really 
looked into the details ;)  But what you are saying sounds right.


-- 
Ian Bicking  /  ianb@...  /  http://blog.ianbicking.org