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
Maintaining selected variables when moving between pages #51
There is no memory of the selected variables when moving between Past Performance / Analysis (and Future Change, once PR #50 is merged). This can be cumbersome when the user wants to look at the Past Performance and compare the Future Change for the same model and situation. It would be nice to find a way that users can easily do this, without repeating selections.
Finally, a question:
When moving between areas, the selection is maintained, but this is not (should not be?) always possible. For example, when selecting Precipitation/Winter/ETHZ/C
It is not possible to make this same selection with the N area, because ETHZ does not appear in the dropdown. However, if the selection is made beforehand, and then a different area is selected, then figures appear after all:
There is no memory of the selected variables when moving between Past Performance / Analysis (and Future Change, once PR #50 is merged). This can be cumbersome when the user wants to look at the Past Performance and compare the Future Change for the same model and situation. It would be nice to find a way that users can easily do this, without repeating selections.
Agree. Two things:
If we combine past performance and future change into one page, this will be less of a problem
It is relatively straightforward to maintain the selection by moving the selections to the parent (pages/cpm.vue) and passing them into the children, just like we do with the map for the region selection.
When moving between areas, the selection is maintained, but this is not (should not be?) always possible.
Good point! Question is what to do instead? Again, two options:
Don't display anything or just a message ("no maps available for this region/model group")
Automatically change the selection of the region to the first one that is available
Good point! Question is what to do instead? Again, two options:
Don't display anything or just a message ("no maps available for this region/model group")
Automatically change the selection of the region to the first one that is available
I would opt for the first, as it can be easy to change one part the selection and then assume that nothing else has changed and the resulting images concern the selection that a user thinks they made.
However, my confusion is mostly the fact that the combination does yield data figures (at least for RCM and GCM), which seems to fit the selection made (precipitation/winter/ETHZ/N), even though this selection can not be made if one would start from the region. Making the selection in another region and then moving to N gives us the pictures.
My reasoning is that either those images are not correct, or that selection should in fact be possible. Either way, it seems off.
These figures are not wrong and should still not be possible. The focus is on CPM, and the regional and global models are only there to show the forcing data. If a CPM model is not available, (RCM and) GCM info for that region might still be available, but it's irrelevant.
Activity
Peter9192 commentedon Mar 29, 2022
Agree. Two things:
Good point! Question is what to do instead? Again, two options:
bvreede commentedon Mar 30, 2022
I would opt for the first, as it can be easy to change one part the selection and then assume that nothing else has changed and the resulting images concern the selection that a user thinks they made.
However, my confusion is mostly the fact that the combination does yield data figures (at least for RCM and GCM), which seems to fit the selection made (precipitation/winter/ETHZ/N), even though this selection can not be made if one would start from the region. Making the selection in another region and then moving to N gives us the pictures.
My reasoning is that either those images are not correct, or that selection should in fact be possible. Either way, it seems off.
Peter9192 commentedon Mar 30, 2022
These figures are not wrong and should still not be possible. The focus is on CPM, and the regional and global models are only there to show the forcing data. If a CPM model is not available, (RCM and) GCM info for that region might still be available, but it's irrelevant.