UNABLE TO GET IDLE TO RUN

[Chris Angelico]

On Mon, Nov 2, 2015 at 12:19 PM, Steven D'Aprano <steve at pearwood.info> wrote:
> - Python as a whole should move "" from the start of sys.path to the end (or
> at least the middle, after the stdlib) so as to avoid accidental shadowing.
>
> - Even if Python doesn't do this, IDLE could do it, and could do it
> immediately, without waiting for a new point release. IDLE is an
> application, an IDE, not the Python interpreter, and like IPython or any
> other IDE, it is perfectly entitled to behave differently from the vanilla
> Python interpreter.

IDLE consists of two things: its own code, and a means of running user
code. Conceptually, they're entirely separate - it's almost a
coincidence that they happen to be using the same language. If the IDE
wants to protect itself against running unexpected userland modules,
it is absolutely allowed to do that, same as any other application is;
it can remove "" from sys.path, or reposition it, or add traps to see
where some_module.__file__ is, or anything at all.

But when it's running user code, it should be semantically as close as
possible to the command-line interpreter as it can be. If I type
"import foo" into IDLE's interactive mode, or if I type "import foo"
into IDLE's editor and hit Run, or if I type "import foo" into the
standard command-line interactive interpreter, I expect that they'll
all find the same module, and if they don't, it's an extra debugging
hassle. When I'm debugging a script, I'll often try some parts of it
in an interactive interpreter, and it should not matter whether I pick
"python", or Idle, or ipython in a terminal window, or the in-browser
ipython notebook, etc, etc, etc. I want to be able to keep the
differences *in my head*. There are already a few; interactive
interpreters...

* Can't have blank lines in class/function definitions;
* Automatically print non-None expression results;
* Recover from exceptions by returning to the main loop, rather than
bombing the whole module load
* Warnings are enabled by default

AFAIK, all of these are consistent across _all_ interactive Python
interpreters. (The second one is configurable, but it's always the
default.) I do *not* want to have to add "Local modules do not shadow
stdlib modules" as an additional difference, particularly not in one
specific interface and not others.

So yes, technically IDLE is allowed to behave differently. But I would
hope this is extremely rare and VERY solidly justified (eg "Python 3.6
will do it this way, and IDLE is a bit ahead of the curve").

ChrisA