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
I would like to be able to select columns directly from the Table tab in the TableReport, to have an easier time writing scripts for acting on a specific set of columns.
The workflow I am thinking of would go like this:
Load the dataframe
Build the TableReport
From the "Table" tab, navigate to one of the cells.
Using the arrows on the keyboard, move left and right to highlight a specific column
From here, I can look at the distribution of values in the column, and make a decision about it (e.g., "this has a lot of unique values", "there are a lot of nulls" etc.).
Depending on the decision, I might want to select and add this column to a growing list of columns (["col_1", "col_2"]) that could be copied like in the Distribution tab.
Afterwards, I'd use the list to drop columns, select a subset and so on: df.drop(["col_1", "col_2"]).
Feature Description
It should be possible to make a selection using only the keyboard: right now, I can navigate in the table report using the arrow keys, which lets me see the distribution of values in each column. I would like to be able to add a column to a list by pressing a key on the keyboard (space, or something else). It should also be possible to remove columns from the selection, if needed.
This feature could also be implemented in the "Stats" tab, rather than in the "Table" tab.
What I find more important is being able to navigate through columns, look at their distribution, then add them to a growing list that I can export easily to a script. Functionally, it would be the same as checking a checkbox, but this should be done using the keyboard rather than the mouse.
Alternative Solutions
This feature is already available in the "Distributions" tab, but at the moment it's necessary to use a mouse and check a checkbox, which is far less convenient.
Additional Context
I am not interested in selecting a range of values in the TableReport, like one would do with A2:B4 in a spreadsheet: I only care about selecting columns.
The text was updated successfully, but these errors were encountered:
thanks @rcap107 I think that's a great idea. it would make an even bigger part of the tablereport's functionality conveniently available from the default tab panel (and the keyboard).
to continue refining the proposal, let's try to answer a few questions:
at the moment the cells can be in several states:
the selected cell (whose content is shown in the bar at the top), it is yellow with an outline
another cell in the column that contains the selected cell (the column whose histogram is shown), it is green without an outline
in a non-selected column, it is gray
moreover the selected cell can have keyboard focus or not, which is shown by having a thick continuous or a thinner dotted outline.
Now, on top of these there would be an additional dimension: a column might be among those selected for the copy-paste list of column names or not. Note this is an independent selection from the existing one: the column whose histogram is shown might not be in the list; columns whose histograms are not shown might be in the list. So we need to decide how that new kind of selection would be represented.
One way would be to have a checkbox next to the column name. It is a different enough indicator that there is no confusion with the already existing use of color and outlines, and it gives a way to select and unselect columns for users who haven't discovered the keyboard shortcut or just prefer to use their mouse (or don't have a keyboard eg on a mobile device, although that's not the main use case). The obvious drawback is that it would take some space for columns whose name is longer than their content. The checkbox could also be on the histogram card (as it is now in the "distributions" tab), and then it wouldn't take space. Any other ideas?
at the moment the use of the keyboard shortcuts is optional, you can also select a cell and copy the contents using the mouse. Any ideas for making it possible to toggle selection with the mouse, eg the checkboxes mentioned above?
the existing keyboard shortcuts are quite standard and obvious: ctr-c to copy, arrows for navigation, escape to blur. This one is more exotic. which key combination do we choose, and (more difficult) how do we make it discoverable?
where should the bar with the copyable list of selected column names be displayed?
Problem Description
I would like to be able to select columns directly from the
Table
tab in the TableReport, to have an easier time writing scripts for acting on a specific set of columns.The workflow I am thinking of would go like this:
["col_1", "col_2"]
) that could be copied like in the Distribution tab.df.drop(["col_1", "col_2"])
.Feature Description
It should be possible to make a selection using only the keyboard: right now, I can navigate in the table report using the arrow keys, which lets me see the distribution of values in each column. I would like to be able to add a column to a list by pressing a key on the keyboard (space, or something else). It should also be possible to remove columns from the selection, if needed.
This feature could also be implemented in the "Stats" tab, rather than in the "Table" tab.
What I find more important is being able to navigate through columns, look at their distribution, then add them to a growing list that I can export easily to a script. Functionally, it would be the same as checking a checkbox, but this should be done using the keyboard rather than the mouse.
Alternative Solutions
This feature is already available in the "Distributions" tab, but at the moment it's necessary to use a mouse and check a checkbox, which is far less convenient.
Additional Context
I am not interested in selecting a range of values in the TableReport, like one would do with
A2:B4
in a spreadsheet: I only care about selecting columns.The text was updated successfully, but these errors were encountered: