API suggestion: name and import simplification

[gwgundersen]

I want to propose a couple changes to the Ristretto API. I'm happy to try make these changes and create a pull request if the main contributors think this is a good idea.

1. Rather than name methods compute_rx where x is the name of the non-randomized method, what about just rx? For example, rsvd instead of compute_rsvd, rlu instead of compute_rlu. This would match libraries like NumPy, but with an r prefix to indicate it is a randomized version.

2. Rather than importing from modules named ristretto.x where x is the name of the non-randomized method, import directly from ristretto. For example, rather than importing from ristretto.cur, import directly from ristretto.

As an example of combining these two changes, instead of,

>>> from ristretto.svd import compute_rsvd

we would have,

>>> from ristretto import rsvd

[jknox13]

Thank you for the suggestion.

Originally, that was the way the API was designed. However, as a project we are moving to support Scikit-learn compatible estimators (#28) and have such chosen to rename the functions with the compute_ prefix and the compelementary classes as the upper case abbreviation (ie compute_rdmd and RDMD).

[gwgundersen]

Ah, okay, that makes sense.

[erichson]

@gwgundersen thanks a lot for your suggestion. Personally, I favor the API you suggested! But, as @jknox13 said, we try to push the package into a direction so that it can be integrated into standard ML pipelines. Anyway, we always need help and contributions are very welcome. If you have any good ideas or if you like to add new features, feel free to push or to drop as a mail.

Best,
Ben