Have code run conditionally on load() vs. command line

[pao]

To make interactive testing of Julia scripts easier and improve the ability to reuse components, it would be nice if there were an equivalent to Python's:

if __name__ == "__main__":
    #things

or Ruby's

if $0 == __FILE__
    # things
end

[Keno]

Could we just have a global variable that is set by the REPL to indicate whether it's the REPL or not?

[pao]

The Python and Ruby approaches are variations on that (with __name__ and __FILE__ being the special variables). @raggi is working on something already.

[JeffBezanson]

One way to tell is if ARGS[1] is the name of the source file.

[pao]

That's the Ruby approach, but unless we provide a special variable which returns the name of the file you are in (or if that already exists, in which case this just becomes documentation) you have to hardcode the file name. That would be sufficient to make this not painful.

[StefanKarpinski]

One way to tell is if ARGS[1] is the name of the source file.

Actually, that's a bug. ARGS should only be the arguments to the program, not the program name. That should be provided via another global variable.

I'm not an expert on the Python thing, but isn't this generally considered to be a misfeature that's just been rampantly abused? I've never really wanted this in Perl or Ruby — a script is a script, a module is a module.

[pao]

And interactive debugging from the REPL or a test harness requires that you be able to do both. Let's say I have a script foo.jl with a bar() function--I might like to test bar() independently of any side effects of running other parts of julia foo.jl.

[StefanKarpinski]

Doesn't that seem to indicate that you should move the definition of bar into a separate file from the script that uses it? I mean, if you have a test suite, I think you're project is big enough to make that separation, and it's kind of good practice, no? I'm not saying there can't be any way to check this, but I'm just unsure that it's good practice. The idea that something is both a module and a script has always struck me as bizarre and kind of wrong.

[pao]

I have a script. That script has repeated code which I want to make a function out of, but which is still quite specific to that script. I really don't want to put that in a separate file just so I can test that out from the REPL.

Also "bigness" is a relative term in technical computing. I've written a lot of throwaway MATLAB that I still need to quickly check before using--forcing me into a heavyweight, multiple-file workflow for that isn't going to make me happy.

[StefanKarpinski]

Ok, fair enough. We can add something like this. Not sure what it should look like. Maybe reflective access to the options that julia was invoked with?

[JeffBezanson]

I can see the case of wanting to load a script in the repl to test its functions without calling main. So maybe just a way to test if we are in interactive mode, like @loladiro suggested?

[pao]

Sounds good to me.