Sort method inDataTable doesn't support rich.Text types

[msempere]

Coming from #2243.

The sort method inDataTable doesn't support rich.Text types stored in the cells. When that is done a TypeError: '<' not supported between instances of 'Text' and 'Text' error is thrown because rich.Text have no rich comparison operators defined.

cc @rodrigogiraoserrao

[github-actions]

We found the following entries in the FAQ which you may find helpful:

• Does Textual support images?
• Why doesn't Textual support ANSI themes?

Feel free to close this issue if you found an answer in the FAQ. Otherwise, please give us a little time to review.

This is an automated reply, generated by FAQtory

[rodrigogiraoserrao]

My original thought (from the PR linked above):

“Maybe we could add a key parameter to the sort methods, akin to how sorted(..., key=...) works.”

@darrenburns might be better suited to determine if this makes sense.

[willmcgugan]

A key method makes sense. But I have no issue adding comparison methods to Text which just treat it like a string.

[msempere]

A key method makes sense. But I have no issue adding comparison methods to Text which just treat it like a string.

@willmcgugan adding them to Text is probably the best idea. I'll work on that.
Could you assign that issue to me? thanks!

[candidtim]

Only having the ordering functions on the rich.text.Text won't be enough for all use cases (although I entirely support this change too). For example: I want to sort a column with humanized numbers (e.g., file sizes like "42 Bytes" and "2 MiB"). It needs to be sorted based on the raw values themselves, independently of their representation.

Adding a sort_fn to the add_column method would not be helpful in this particular case either, because it implies that it is possible to convert the textual representation back into whatever can be ordered, which in practice is not always the case. But otherwise I think it is nice to have it for more flexibility anyway.

One possible solution to the above problem is to allow conveying the values and their representation separately. Then, either use the raw values for ordering, or also allow a custom function to extract the sort keys.

Meanwhile, for a workaround I extend the rich.text.Text to implement the ordering. For example:

@functools.total_ordering
class Sortable(Text):
    """Like rich.text.Text, but with ordering based on a raw value"""

    def __init__(self, value, *args, **kwargs):
        self.value = value
        super().__init__(*args, **kwargs)

    def __lt__(self, other: "Sortable") -> bool:
        return self.value < other.value

    def __eq__(self, other: "Sortable") -> bool:
        return self.value == other.value

and later use it like so:

    def watch_xyz(self, ...):
        ...
        self.table.add_row(
            ...
            Sortable(file_size, naturalsize(file_size), ...),
            ...
        )

It allows customizing the ordering logic in any way necessary by changing the __lt__ implementation. For example I did things like sticky top/bottom rows, etc.

[TomJGooding]

Adding a sort_fn to the add_column method would not be helpful in this particular case either, because it implies that it is possible to convert the textual representation back into whatever can be ordered, which in practice is not always the case.

I'm assuming you've looked at #3090, could you give an example where this would not work?

[candidtim]

I'm assuming you've looked at #3090, could you give an example where this would not work?

As far as I understand the example from above still holds, but please let me know if I am missing something. If I understand the change well, the new key argument accepts a function which will be called with a row content (key and cell values). This means that if I want to sort based on some value, this value necessarily needs to be present in the row cells in one way or another. What if the value I want to sort on is not directly present in the row? For example, I would show humanized numbers ("1 MiB", "1 GiB") in the column, but would want to sort it based on actual numbers (1 MiB < 1 GiB). Since the actual number is not present in the row, how would the key function obtain it?

[sharkusk]

In the sort routine you can always un-humanize the numbers. You might get some round-off errors if you don't have enough precision in your human readable numbers, but in general works pretty well. (I do this in my app today with a DataTable and a custom sort() function.)

[rodrigogiraoserrao]

How would the key function obtain it?

The key function can do whatever you want with the values it receives, namely “un-humanize” them.
See, for example, the parameter key in the built-in sorted: https://docs.python.org/3/library/functions.html#sorted

[guystreeter]

I would very much like to supply my own callback to the sort function. Simply handing me the row key would be enough, but having all the row data would be better.
An alternative would be enabling a hidden column. I could put a sortable value there that doesn't get displayed.

[TomJGooding]

I stumbled across this open issue looking for something else. You can now sort your DataTable with a custom function (or other callable) in Textual v0.41.0 - see https://textual.textualize.io/widgets/data_table/#sorting

[rodrigogiraoserrao]

Thanks @TomJGooding, looks like we can close this issue now.

[github-actions]

Don't forget to star the repository!

Follow @textualizeio for Textual updates.