Data types module, collating conversions, mtype inference, checks, register#1201
Data types module, collating conversions, mtype inference, checks, register#1201
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
I guess I win the "maximize number of codeowners who have to review" game with this one... |
TonyBagnall
left a comment
There was a problem hiding this comment.
I've discussed this with my group. This is definitely something that needs doing, and whilst the style is not mine, it fine.
- I would ask you have some explanation of machine types, scitypes, even if just linked from the docstrings.
2)I'm also not 100% sure it needs a full root level module, rather than being in data (scikit style) or utilities (my preference). I also am trying to think of another name, I dont like datatypes.
- I dont like the mix of underscore and camel case naming conventions,
That all said, in order to not hold up transformers, I'm fine with it going in
Yes, I wanted to write a doc page.
Neither am I, I just moved different things from different internal places in one place. Once we come up witih something better, we can deprecate this. Also, "mtypes" or "sciptypes" would be too cryptic as a name...
My linter didn't complain... |
…rs (#1225) This PR further consolidates the datatypes module introduced in #1061 and #1201: * adding example fixtures for most important panel data containers * consolidated checks for "is (some type)" in module, added some missing ones * adding registry constants for panel data containers * renaming leftover "what" variables to "obj" * bugfix in converter from nested to multi-index * adding converters for list-of-data-frames panel type used in the distance module * added tests for checking functionality * adding tests for conversion functionality, testing converters against fixtures * adding docstrings where they were missing
This PR collates data type and conversion related concerns in its own module,
datatypes.The functionality moved to
datatypes:forecasting.base.convertIOutils.data_processing._panelAll references to the old locations have been replaced by references to the new locations.
New features enabled by this move:
convertfunction, for bothSeriesandPanelscitypepanelandseriesNew functionality, partially implemented for extension:
get_examplewhich produces fixtures of the same "scientific content" but a specific mtype, for external tests, and bulk testing of conversion or mtype inference functionality