You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Possibility to associate metadata to each cell of an array #689
@gdementen One of our team specifically asked for this feature. We need more information about how they want to use these metadata before to start any implementation work on this feature.
To each data? You mean for each cell? I guess this is to contain the "letters" present in some eurostat datasets (provisional etc). Currently we could have an object array next to the main array containing the letters. Or maybe it would be better in this particular case to have several boolean arrays.
We will probably want to add native support for attaching such (an) array(s) to the main array. eg.
arr.cellmeta
or arr.cellmeta.provisional
Yes to each cell and yes, I suppose it is related to some associated "letters".
What you propose is to have something like Numpy masked_array (depending if they need performance or not)? I think they want to make aggregations or manipulations according to subsets defined by these metadata.
Well, we first need to discuss with the members of this team before to think about any implementation.
Yes, that would be somewhat similar to masked_array except there could be several masks. We indeed need more information. As a side note, I think 0.30 is too early for this feature...
changed the title [-](Eurostat data) Find a way to associate metadata to each data of an array[/-][+](Eurostat data) Find a way to associate metadata to each cell of an array[/+]on Jan 9, 2019
changed the title [-](Eurostat data) Find a way to associate metadata to each cell of an array[/-][+]Possibility to associate metadata to each cell of an array[/+]on Jan 9, 2019
Activity
gdementen commentedon Sep 14, 2018
To each data? You mean for each cell? I guess this is to contain the "letters" present in some eurostat datasets (provisional etc). Currently we could have an object array next to the main array containing the letters. Or maybe it would be better in this particular case to have several boolean arrays.
We will probably want to add native support for attaching such (an) array(s) to the main array. eg.
arr.cellmeta
or arr.cellmeta.provisional
alixdamman commentedon Sep 14, 2018
Yes to each cell and yes, I suppose it is related to some associated "letters".
What you propose is to have something like Numpy masked_array (depending if they need performance or not)? I think they want to make aggregations or manipulations according to subsets defined by these metadata.
Well, we first need to discuss with the members of this team before to think about any implementation.
gdementen commentedon Sep 14, 2018
Yes, that would be somewhat similar to masked_array except there could be several masks. We indeed need more information. As a side note, I think 0.30 is too early for this feature...
[-](Eurostat data) Find a way to associate metadata to each data of an array[/-][+](Eurostat data) Find a way to associate metadata to each cell of an array[/+][-](Eurostat data) Find a way to associate metadata to each cell of an array[/-][+]Possibility to associate metadata to each cell of an array[/+]