Multi-column sorting for Tables

[runrc]

To enhance the usability of Tables, it would be useful to allow multi-column sorting via selection of multiple column headers. Clicking on a column header would allow up, down or no-sorting. Subsequent columns could be added to the sorting criteria by clicking. Columns can be removed by clicking the column header to select no-sorting.

[kkoreilly]

So the first column you click on is the primary sorting criteria, and then each additional column you click on is used as a secondary sorting criteria to sort within each resultant sorting group of the first column? Also, currently when you click on a column already used for sorting, it switches back and forth between sorting ascending and descending; how do you remove columns from sorting then? Is it a third option that you click through? I can envision clicking through multiple options being somewhat confusing, so maybe we would put the option to stop sorting in a context menu on the header? Regardless, this seems like a reasonable possibility. Thank you for the suggestion!

[runrc]

OK great. We're dealing with hundreds of thousands of rows of data. Thus, sorting via multiple criteria would be very useful.

[kkoreilly]

OK. Feel free to ping us later when you need it if it hasn't been implemented yet.

[rcoreilly]

@runrc you might want to look into the tensor/table data structure given the scale of your data -- it has a very efficient index-based sorting algorithm and built-in support for multi-column sorting. It uses the same basic gui to display the table. Good thing we made the table only use the visible rows and scroll only the data :)

[rcoreilly]

also has interactive filtering (which we can improve now that we have yaegi integrated)

[kkoreilly]

Just to be clear, it is tensor/tensorcore.Table. If you need any help with that API, you can ask; it is not as well documented yet.