Path: csiph.com!usenet.pasdenom.info!dedibox.gegeweb.org!gegeweb.eu!nntpfeed.proxad.net!proxad.net!feeder1-2.proxad.net!usenet-fr.net!nerim.net!novso.com!newsfeed.xs4all.nl!newsfeed2.news.xs4all.nl!xs4all!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
In-Reply-To: <CALwzidmqKkpuugv2zvu-8RpoVZF899G2Hpdmd_kVn8a33WMWFw@mail.gmail.com>
References: <CAPTjJmqhiyT9zVHFRyKH0UwhdhFgZxA1rcbarChePMyfK7dDTA@mail.gmail.com> <CALwzidmqKkpuugv2zvu-8RpoVZF899G2Hpdmd_kVn8a33WMWFw@mail.gmail.com>
Date: Sun, 24 Nov 2013 19:07:38 +1100
Subject: Re: Importing by file name
From: Chris Angelico <rosuav@gmail.com>
To: python-list@python.org
Content-Type: text/plain; charset=ISO-8859-1
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.3119.1385280468.18130.python-list@python.org>
Lines: 36
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:60354

On Sun, Nov 24, 2013 at 7:00 PM, Ian Kelly <ian.g.kelly@gmail.com> wrote:
> The importer mechanism as far as I know only accepts module names, not
> filesystem paths; I believe this is by design.  You could imitate it by
> doing something like this:
>
> import imp
> import sys
>
> mod = imp.new_module('spam')
> exec(open('/path/to/spam.py').read(), mod.__dict__)
> sys.modules['spam'] = mod

That's a bit better than just working with a bare dictionary as I did,
but it's still getting a bit heavy-handed with the details. How is
__main__ imported? Presumably that one, at least, can be loaded from a
specific file. Or is that pure magic?

> Is 'spam' really the appropriate name for this module?  What if there is
> already a different module with the name 'spam'?  Presumably if the module
> is named 'spam' then it should be importable as 'spam', but then why the
> need for the path-based import?  Alternatively, you might name the module
> something like "</path/to/spam.py>", which surely won't collide with
> anything imported by the normal mechanism, but then what if the spam module
> is also imported by normal means?  You would end up with two copies of the
> same module with different names.

The same problem already exists with __main__ - I was hoping to
replicate that, not necessarily to solve all its problems :) I'm okay
with using a fixed name (either "__main__" or something that doesn't
collide with that), for the purposes of experimentation.

I know the recent Pythons give a lot of import power to the script.
But maybe I'm just asking too much, and some of this stuff really is
magical and implemented in C?

ChrisA