labels subcommand broken by default python

[vpipkt]

The labels subcommand runs the stream_filter.py code as a separate process using the system's default python. In my case that was python 2.7 and shapely was not installed so I found this through ImportError.

In general that is confusing and could lead to incorrect behavior or non-deterministic results.

label-maker/label_maker/label.py

Line 79 in df0e10e

run(['python', stream_filter_fpath, json.dumps(bounding_box)],

[drewbo]

@vpipkt good find. It's set up in this slightly rickety manner so that we can pipe from tippecanoe-decode directly to this function. Do you have an idea of how we can do this natively? We might need to call tippecanoe in a different way but I'm not very familiar with writing Python bindings for C++ libraries; it looks like there are a few examples of python-tippecanoe attempts but nothing very well maintained.

[vpipkt]

I have a slightly less rickety idea...

Use sys.executable in place of 'python' in the list to run. That seems to work with pip install label-maker. It recovers the expected interpreter in the shebang line of the label-maker script. Which in my case was the same python used to install pip.

Of interest when I was researching this, I found this footnote:

On Unix, the Python 3.x interpreter is by default not installed with the executable named python, so that it does not conflict with a simultaneously installed Python 2.x executable.

Given that note an alternative may be to use 'python3.6'.

[drewbo]

@vpipkt I did a few quick checks in different environments and the sys.executable seems to work well. I'll put up a PR later today unless you have one ready

[drewbo]

@vpipkt you can install from source right now and get this fix or I'll include it in the next release 0.5.2

[vpipkt]

@drewbo thanks much! I used alternatives to make sure things ran for me... just include this in the next release!