Repository navigation
add list_dataset filters #369
Description
The OpenML list_datasets call allows to filter datasets by number or instances, features, classes, and missing values, but this is not yet supported by the python API.
https://www.openml.org/api_docs/#!/data/get_data_list_filters
Can we please add this?
Also see openml/OpenML#496
Activity
ArlindKadra commentedon Mar 15, 2018
@joaquinvanschoren are there any other filters following the ones mentioned above ? @mfeurer should we control the filters names and their values ? In case there are only the filters above, I can use regex to not allow illegal values.
mfeurer commentedon Mar 15, 2018
Could you try whether the server returns a helpful error message in case the filter is illegal? If yes, there is no need to handle this in python code.
janvanrijn commentedon Mar 15, 2018
The server does, right?
See, e.g., https://www.openml.org/api/v1/data/list/illegal/123
mfeurer commentedon Mar 15, 2018
Awesome, now @ArlindKadra only needs to figure out whether it prints nicely in python.
ArlindKadra commentedon Mar 15, 2018
@janvanrijn @mfeurer The reply from the server is descriptive and nice. In the case of illegal filter values though on the production server you get a no results error and on the test server you get a database error. Do we want this outcome ?
https://www.openml.org/api/v1/data/list/number_features/ta
https://test.openml.org/api/v1/data/list/number_features/ta
mfeurer commentedon Mar 15, 2018
I think it should be something like illegal argument error. Could you please post an image @ArlindKadra ?
ArlindKadra commentedon Mar 15, 2018
@mfeurer
Feature #369 (#424)
mfeurer commentedon Mar 19, 2018
Added in #424 .