[Grok-dev] Re: Grok 1.0 and beyond

Martijn Faassen faassen at startifact.com
Fri Jan 4 20:39:18 EST 2008


On Jan 5, 2008 2:20 AM, Wichert Akkerman <wichert at wiggy.net> wrote:
> Previously Martijn Faassen wrote:
> > I know there are people who are in favor of easy_install over buildout,
> > but we'd need to see some really compelling arguments before we consider
> > throwing buildout overboard or trying to mix the two in some way, given
> > that buildout is already a working solution right now (with a lot of
> > recipes available).
> Ease of use. Telling someone to run 'easy_install xyz' is a lot simpler
> than explaining that he needs to edit src/<appname>/setup.py, being very
> careful to not introduce any syntax errors, and run bin/buildout which
> can take a long time.

That isn't very compelling:

setup.py is not in 'src/appname', it's directly in the top-level directory.

Grok's bin/buildout does not run a long time at all (after the first
time you run grokproject). buildout.cfg is configured so it doesn't
needlessly check the network, after all.

You can still argue people might get the install_requires section
wrong. easy_install is more easy there, but on the other hand people
will need to know setup.py's install_requires section anyway for even
moderate use cases - people do need to check in their package's
dependencies, after all.

If we really need more ease of use, we could a recipe that creates a
recipe bin/install, and then you could type:

bin/install foo

to get foo added to setup.py, and to run bin/buildout. It's be pretty
easy to create.

The main argument for easy_install is familiarity. It's a good
argument, but by itself it isn't compelling enough to switch.

Beyond this use case, buildout does more things than just installing
dependencies for us; it's extensible by recipes of which there are
quite a few available already.



More information about the Grok-dev mailing list