0.2.1

• Better serialization for the callable we submit to the Dask scheduler

0.2.0

• Update method to get the context body. Hopefully more reliable 🤞
• with afar.run: now gives a useful error message (missing location!) instead of running the code
• Add run.context_body attribute to hold a list of lines of the context body
• Add later location, such as with run, later:
  • This is useful to get the context body, but not execute it
  • later doesn't do much yet. I'll handle it later.
• Test running in a notebook via nbconvert
• Test PyPy with conda

0.1.3

• Allow data to be passed to afar.run(data=data)
• Expose run.data, but the singleton afar.run doesn't keep .data around
• Probably fix a bug
• For warm fuzzies, test on Windows, OS X, and Linux, and with pip and conda

0.1.2

• Change names of functions we pass to client.submit to afar_run and afar_get.
• Set up CI, which required changing how we do some magic, so maybe fix a bug?
• Tests pass for Python 3.7, 3.8, 3.9, and PyPy--hooray!

0.1.1

• Keyword arguments to remotely(**kwargs) will be passed to client.submit
• Added afar.get, which is like afar.run, but gathers data locally so calling .result() isn't necessary

0.1.0

• Minimal example working with dask.distributed!

0.0.1

Look what you just made me do...