[reportlab-users] Pypi package incorrect

[Marius Gedminas]

On Fri, Mar 05, 2010 at 03:05:12PM +0000, Andy Robinson wrote:

> On 5 March 2010 14:52, Mike Driscoll <mdriscoll at co.marshall.ia.us> wrote:

> > Some people seem to think that "distribute" is the new thing:

> >  http://pypi.python.org/pypi/distribute

> 

> Yes, and I have NO intention of putting in ever-growing amounts of

> staff time trying to cater to all of them  ;-)


I understand and sympathise with that position.


> I'be just been talking to my colleagues on what we did last time.  To

> date we have prepared packages with 'svn export', tar, and a script

> which applies some python-version-specific packages.  For the PyPi

> distro we manually created the extra file it needs and ensured it was

> in place in the tar file; it then all appeared to work fine for

> easy_install.


It does work with easy_install.  Easy_install follows the "be liberal in
what you accept", "try to work with everything that already exists out
there" and "let me guess what you mean" route.  This kind of guessing
sometimes leads to problems that are extremely hard to debug (the Zen of
Python has "In the face of ambiguity, refuse the temptation to guess."
for a reason).  AFAIU pip was born as a reaction against this policy; it
tries to avoid magic and is therefore more strict in what it accepts.

If you could change the location of the extra file (PKG-INFO) and put it
in the ReportLab_2_4 subdirectory instead of the root, I'm sure both pip
and easy_install would work (and I volunteer to test a tarball with
both, if you need testers).  You'd also avoid the hassle of having to
keep MANIFEST.in up-to-date.


> If it makes things better for other tools, we are happy to switch over

> to (a) using 'setup.py sdist' and (b) changing the naming convention

> of the archive and the top-level directory inside it.


That's nice to hear.

My personal view here is: do whatever's the simplest thing for you.  An
simple fix today is better than a perfect fix next month (assuming
the simple fix isn't a hack that introduces a maintenance burden, of
course).

Marius Gedminas
-- 
I suspect the main reason chemical rockets are always so much worse than
alternative schemes is that chemical rockets are real.
        -- Matt McIrvin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
Url : <http://two.pairlist.net/pipermail/reportlab-users/attachments/20100305/1225d1d4/attachment.pgp>