[Pyrex] array setting and math function annoyance

[Jim Kleckner]

Robert Bradshaw wrote:
> On Mar 13, 2008, at 12:14 AM, Arc Riley wrote:
[snip]
>> We'd be using Cython if it were not for a few pesky bugs we've yet  
>> to work around, namely our source is arranged in a manner which  
>> Cython cannot handle;
>>
>> src/<extension>/soy.<extension>.pyx which includes src/<extension>/ 
>> <EachClass>.pxi
>>
>> The resulting compile, as you could likely guess, raises errors  
>> such as:
>>   AttributeError: 'src.colors.soy.colors.Color' object has no  
>> attribute '_a'
> 
> 
> Yes, we handle submodule naming based on the directory hierarchy to  
> be consistent with Python. Sorry this doesn't work for you.

See this thread for a detailed discussion of this topic:
  http://thread.gmane.org/gmane.comp.python.pyrex/3063

See this wiki page for some notes from David McNab on how
to address this issue:
  http://wiki.cython.org/PackageHierarchy
You might want to add comments there.

Like you, Arc, I would prefer that I could use one
set of file names that would work for both Pyrex and Cython.
I ended up creating a Makefile that can switch between
cython and pyrex and a small helper script to link the pyx
and pxd files under both naming conventions.

The Makefile passes into setup.py a parameter to tell it
which names and paths to use depending on the compiler.
Yes, it is a hack but it allows me to switch back and forth
for testing.  (The Makefile runs "python setup.py usecython"
from the parent directory to make this all work).