parquet parser for dataframe serialisation

[xiki-tempula]

This PR add a parquet parser, which will load the serialised dataframe and permits the usage of

    from alchemlyb.parsing.parquet import extract_dHdl, extract_u_nk
    import pandas as pd

    u_nk.to_parquet(path='u_nk.parquet', index=True)
    dHdl.to_parquet(path='dHdl.parquet', index=True)

    new_u_nk = extract_u_nk('u_nk.parquet', T=300)
    new_dHdl = extract_dHdl('dHdl.parquet', T=300) 

Fix #316

[codecov]

Codecov Report

Merging #317 (7347a9c) into master (4e590cc) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #317      +/-   ##
==========================================
+ Coverage   98.74%   98.75%   +0.01%     
==========================================
  Files          26       27       +1     
  Lines        1754     1772      +18     
  Branches      382      387       +5     
==========================================
+ Hits         1732     1750      +18     
  Misses          2        2              
  Partials       20       20              

Impacted Files	Coverage Δ
src/alchemlyb/parsing/parquet.py	100.00% <100.00%> (ø)
src/alchemlyb/workflows/abfe.py	99.67% <100.00%> (+<0.01%)	⬆️

[xiki-tempula]

@orbeckst Do you mind have a review of this PR, please? Thank you.

[orbeckst]

Basic functionality looks good but some request for changes

• docs and clearer code (see comments)
• update all build files with new pyarrow dependency: environment.yml, setup.py, + raise issue in https://github.com/conda-forge/alchemlyb-feedstock/ to add pyarrow to the deps of the meta.yml of the conda-forge package (and link to this issue)

[orbeckst]

Suggested change

	Temperature in Kelvin the simulations sampled.
	Temperature in Kelvin of the simulations.

[orbeckst]

State that T is ignored but included for API compatibility

[xiki-tempula]

Sorry but parquet serialisation doesn't preserve the df.attrs. So the dataframe loaded here doesn't contain temperature and the temperature is assigned via this T.

[orbeckst]

Also mention the enhancement to the ABFE workflow

[xiki-tempula]

@orbeckst Thanks for the review. I have change the environment.yml. I will raise an issue with meta.yml when we are doing a release.

So the T problem is a bit tricky in that parquet serialisation doesn't preserve the df.attrs. So the dataframe loaded here doesn't contain temperature and the temperature is assigned via extract function. I will add a not to state that.

[orbeckst]

So the T problem is a bit tricky in that parquet serialisation doesn't preserve the df.attrs. So the dataframe loaded here doesn't contain temperature and the temperature is assigned via extract function. I will add a not to state that.

Ok, I blithely assumed that parquet would save everything — then T is not optional. Definitely document this requirement.

[xiki-tempula]

@orbeckst

Ok, I blithely assumed that parquet would save everything

I'm annoyed by this as well but it seems that parquet is still the best serialiser. This is the only format that preserves index besides to_pickle. Also it faithfully preserve all the data in its original datatype. The to_pickle, though it preserves everything, is slow to read and write and also would not be safe between different versions of pandas.

Definitely document this requirement.

I have added this as a note to both alchemlyb.parsing.parquet.extract_u_nk and alchemlyb.parsing.parquet.extract_dHdl.

[orbeckst]

Let me know when I need to review again.

[xiki-tempula]

@orbeckst I addressed the comments. Do you mind having another review? Thank you.

[orbeckst]

lgtm

[orbeckst]

Please squash-merge when ready.