[Zope3-dev] Big import blocks

[Guido van Rossum]

> Guido van Rossum writes:
>  > So I decided to count.  There are currently 4904 "from ... import ..."
>  > lines in the entire Zope3 source tree (including tests and
>  > zopeproducts).  Because many of these are of the form "from X import
> 
> I'd actually prefer that imports in the tests be ignored for this
> purpose; a separate "API helper" module can be created for use from
> test modules if it turns out to be needed.

I agree that the four imports from unitest shouldn't be included in
zapi.  OTOH I think that the other dead chickens found in the test
modules are representative of code using the Zope 3 API, so I do want
to count those.  Here are the unittest imports:

 166 unittest.makeSuite
 162 unittest.TestCase
 156 unittest.TestSuite
 154 unittest.main

This means 4 names account for 638 imports.  Subtracting these from
the totals, 114 names out of 1724 account for 2492 imports out of a
total of 5995, or 6.6% of the names account for 42% of imports.

>  > Assuming this data is roughly correct, we can cut the number of this
>  > particular kind of dead chickens roughly in half by creating a module
>  > with a short name (e.g. "zapi") that imports the 120 most popular
> 
> I think this is a useful analysis, and would love to see something
> like this done instead of continuing the current import hell.  A
> script to perform the analysis would be good; additions could be made
> to the API module when something becomes popular enough.

I have a script to perform the analysis; it's just too crummy to
post. :-)

--Guido van Rossum (home page: http://www.python.org/~guido/)