feat(df): Make .input_cell_selection() return a consistent type shape; Change row order to data view row order

[schloerke]

Should be rebased after #1374 is merged!

TODO

• Merge feat(data frame): Expose .data_view_rows(), .input_sort(), .input_filter(), .update_sort(), and .update_filter() #1374
• Rebase to main
• Change merge branch to main… (or rows) tuple containing all column (or row) numbers

** Update merging into #1374 to get changes

---

Features for .input_cell_selection():

• Removed None return type
• Added missing cols (or rows) when the input object sent across the wire does not contain the values.
• Row order is now ordered as if it is being subsetted from Data View row order

Note:
input.<ID>_cell_selection()'s shape did not change. It does not contain missing cols (or rows) as that information is always unnecessary.

(However, this leads to an awkward user interaction with the code, so <ID>.input_cell_selection() upgrades the return value to support the missing keys.)

---

Previous return type for .input_cell_selection():

class CellSelectionNone(TypedDict):
    type: Literal["none"]

class CellSelectionRow(TypedDict):
    type: Literal["row"]
    rows: ListOrTuple[int]

class CellSelectionCol(TypedDict):
    type: Literal["col"]
    cols: ListOrTuple[int]

class CellSelectionRect(TypedDict):
    type: Literal["rect"]
    # Attempt to type the list of size 2, but it is not type enforced
    rows: tuple[int, int] | Annotated[list[int], 2]
    cols: tuple[int, int] | Annotated[list[int], 2]

CellSelection = Union[
    CellSelectionRow,
    CellSelectionCol,
    CellSelectionRect,
    CellSelectionNone,
]

def .input_cell_selection() -> CellSelection | None:

New return type for .input_cell_selection():

class CellSelection(TypedDict):
    type: Literal["none", "row", "col", "rect"]
    rows: ListOrTuple[int]
    cols: ListOrTuple[int]

def .input_cell_selection() -> CellSelection:

• If type = "none", rows and cols are empty tuples.
• If type = "row", cols is all column numbers of the data.
• If type = "col", rows is all row numbers of the data view rows
• If type = "rect", all rows and cols are provided (no longer just min/max values).

---

Previous user interaction:

@render.data_frame
def grid(): ...

selected_rows = (grid.input_cell_selection() or {}).get("rows", ())

New user interaction:

selected_rows = grid.input_cell_selection()["rows"]

[wch]

Note that after we release this, we'll need to update the examples in py-shiny-site. The changes would be a follow-up to posit-dev/py-shiny-site#143.

[schloerke]

From #1374 (comment)

• Also, input_cell_selection rows should be ordered the way the user sees it.

• In the meeting, we also came to the agreement that the None return type should be reinstated so that it is handled by the user

[schloerke]

@jcheng5 None support within .input_cell_selection() feels wonky. Should it instead throw a req(False)?

[schloerke]

Squash merging into base branch for easier PR interactions as main is frozen for v0.10.0 release