Re-raise exception when NumPy/SciPy imports fail

[mattwthompson]

The check for NumPy/SciPy installs in setup.py prints out a helpful message for the user and exits (so that the rest of the install attempt does not proceed) but raises exit code 0 when it should probably raise 1. This may cause some issues downstream when these packages are not installed properly but this error is only printed to stdout or a logfile but doesn't stop a conda build from happening.

Just to demonstrate this, since it can't easily be added to CI (echo $? prints the exit status of the last command, which I just learned by working on this)

bash-3.2$ python -c "exit()"
bash-3.2$ echo $?
0
bash-3.2$ python -c "raise Exception"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
Exception
bash-3.2$ echo $?
1

And here's what it looks like to the user before and after this patch

(omnia) [forcebalance] python setup.py install                                                                                                                                                                                        master
Error importing numpy and scipy but these are required to install ForceBalance
Please make sure the numpy and scipy modules are installed and try again
(omnia) [forcebalance] echo $?                                                                                                                                                                                                        master
0
(omnia) [forcebalance] git checkout loud-importerror                                                                                                                                                                                  master
Switched to branch 'loud-importerror'
(omnia) [forcebalance] python setup.py install                                                                                                                                                                              loud-importerror
Traceback (most recent call last):
  File "setup.py", line 16, in <module>
    import numpy
ModuleNotFoundError: No module named 'numpy'

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "setup.py", line 23, in <module>
    raise ImportError(msg) from e
ImportError:
Error importing numpy and scipy but these are required to install ForceBalance.
Please make sure the numpy and scipy modules are installed and try again.

(omnia) [forcebalance] echo $?                                                                                                                                                                                              loud-importerror
1

[leeping]

Thank you, Matt! Is there somewhere I can learn more about the as e/from e syntax? One of the tests has a failure that I think is stochastic so I'm rerunning it.

[mattwthompson]

I mostly grabbed the solution from here and the official docs are here. The intention with this approach was to re-raise the same exception. as e catches the exception, and from e makes clear where the raised exception come from. This is my first time chaining exceptions, but I think it's working as intended; the ModuleNotFoundError is printed as the cause of the ImportError (which itself could be a ModuleNotFoundError, I suppose).

This is all a little fluffy, to be honest; simply changing the exit() to an exception would function just about as well.

[leeping]

I can appreciate why one would want the original exception, it could help with debugging. :) Thanks for the explanation!

[mattwthompson]

Ah, but it looks like that's not Python 2.7 compatible. This should work now (works locally, at least)

[leeping]

Thanks. I plan on keeping Python 2.7 support for a while because you never know who is still using it. My guess is there are many research groups that still use it.

[leeping]

LGTM. Thank you!