43 port multitemporal to python 3

[ra-tolson]

Fixes #43. I tried to keep the refactoring to a minimum, but I did find it easier to understand the source after I'd renamed a few variables. Also:

• 2.0.0 seemed to make sense; it's a big set of internal changes even if the interfaces aren't any different.
• just to get moving, rebased the docker container onto geographica/gdal2 (we should probably vet that image if we're going to keep using it for stuff)
• python3 port done by futurize; invocations noted in commits. This is why the source looks unconventional for my style in places.
• I didn't run example.py nor testsharedmem.py since they don't look like critical code paths.
• passthrough had some junk it it, might as well take it out
• passthrough could only accept a limited data type (I guess?) so I opened it up to accept anything
• correlate had a NameError, fixed that to make the test run.
• bin/screen.py seemed like it was an accidental copy of screen.pyx so I deleted it.
• otherwise I tried to keep the refactoring to a minimum.

Regarding the test suite:

• the new tests cover all the important code paths I could find in mt.py etc, but the cython code didn't change at all except to make the tests run so there's no attempt to cover cython modules.
• the added pile of tiffs are taken from the tarball; now it's not necessary to untar them every time.
• the tests aren't intended to be good tests of mt functionality so much as they're intended to have broad source coverage, especially of mt.py.

[bhbraswell]

I'm embarrassed to say I just started doing this separately, realized I had told you that I already did it, which is not true, and then realized that you did it yourself and that this PR was sitting here.

[bhbraswell]

The other thing I noticed is my test didn't pass because I was missing test_1.json so I was going to ask about that. Now I'll look at this whole thing and try it out.

[ra-tolson]

Okay, happy to help if you want to post questions etc. Any feedback would be welcome too.

I regret to say I didn't save my test invocation line but it was probably just $ pytest -s -vv from inside a built container; you shouldn't need any ancillary json files I don't think.

[bhbraswell]

Looks like there's possibly a version incompatibility, I'm getting

rob@MBP-18-111:~/repo/multitemporal$ docker run -it --rm multitemporal pytest
=========================================================================================== test session starts ============================================================================================
platform linux -- Python 3.6.7, pytest-5.2.2, py-1.8.0, pluggy-0.13.0
rootdir: /usr/local
plugins: cov-2.8.1
collected 402 items / 3 errors / 399 selected                                                                                                                                                              

================================================================================================== ERRORS ==================================================================================================
_______________________________________________________ ERROR collecting lib/python3.6/dist-packages/Cython/Debugger/Tests/test_libcython_in_gdb.py ________________________________________________________
ImportError while importing test module '/usr/local/lib/python3.6/dist-packages/Cython/Debugger/Tests/test_libcython_in_gdb.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
lib/python3.6/dist-packages/Cython/Debugger/Tests/test_libcython_in_gdb.py:24: in <module>
    import gdb
E   ModuleNotFoundError: No module named 'gdb'
_______________________________________________________ ERROR collecting lib/python3.6/dist-packages/Cython/Debugger/Tests/test_libpython_in_gdb.py ________________________________________________________
ImportError while importing test module '/usr/local/lib/python3.6/dist-packages/Cython/Debugger/Tests/test_libpython_in_gdb.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
lib/python3.6/dist-packages/Cython/Debugger/Tests/test_libpython_in_gdb.py:12: in <module>
    import gdb
E   ModuleNotFoundError: No module named 'gdb'
____________________________________________________________ ERROR collecting lib/python3.6/dist-packages/importlib_metadata/tests/test_zip.py _____________________________________________________________
ImportError while importing test module '/usr/local/lib/python3.6/dist-packages/importlib_metadata/tests/test_zip.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
lib/python3.6/dist-packages/importlib_metadata/tests/test_zip.py:7: in <module>
    from importlib.resources import path
E   ModuleNotFoundError: No module named 'importlib.resources'

During handling of the above exception, another exception occurred:
lib/python3.6/dist-packages/importlib_metadata/tests/test_zip.py:9: in <module>
    from importlib_resources import path
E   ModuleNotFoundError: No module named 'importlib_resources'
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Interrupted: 3 errors during collection !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
============================================================================================= 3 error in 1.14s =============================================================================================

[bhbraswell]

Sorry. I did a fresh checkout and it worked. This is fantastic, thanks.

[ra-tolson]

great, so glad it's working for you!

[bhbraswell]

So unless I'm missing something, pip install -e will create the importable .so files in the source tree which is overwritten if you mount the /multitemporal to the container, and unless the compiled code is already there for some reason, mt won't be able to find the modules. I think this is why the directions at the bottom of the readme don't work for me (unless I drop the -v [stuff]). This raises the question of how do I operate the container when I'm developing. Maybe. Am I just doing it wrong?

[ra-tolson]

@bhbraswell I think you discovered that I forgot a step. I altered the readme; if you want you can try it again.

But by all means develop in whatever way is most comfortable to you; I'm sure the mount isn't required if you're installing an editor in your container.

[ircwaves]

OK, now I see what you're saying. And I think the line added to the readme by Tom's last commit helps with this. If you're going to develop on a mounted (-v /path/to/my/mt/dir:/multitemporal) repo, you do need to build the modules. Does that make sense @bhbraswell? We can meet for a couple minutes to clear this up if it doesn't. I def could be misunderstanding something here.

[bhbraswell]

Sorry, it seems obvious in retrospect, yeah. Thanks. This is looking good, I'm using it now.

[bhbraswell]

I'm just a data scientist, but I approve this message.