OpenRPC for cursors/ranges rstudioapi shims

[juliasilge]

Intent

Addresses #1312 with support for:

• insertText() / modifyRange()
• setCursorPosition() / setSelectionRanges()

Related to posit-dev/ark#275

Approach

Chugging along with more rstudioapi shims, as before 🚂

QA Notes

Make a new R file with this content:

"hello 🌍, I like crème brûlée!"
"hello 🌷, I like crème brûlée!"
"hello 🤖, I like crème brûlée!"
"hello 🤷‍♀️, I like crème brûlée!"
"hello 😭, I like crème brûlée!"
"hello 💩, I like crème brûlée!"
"hello 🤷‍♀️, I like crème brûlée!"
"hello 🥔, I like crème brûlée!"

rstudioapi::setCursorPosition(rstudioapi::document_position(5, 1))

rgs <- list(
    rstudioapi::document_range(rstudioapi::document_position(2, 3), rstudioapi::document_position(3, 1)),
    rstudioapi::document_range(rstudioapi::document_position(4, 7), rstudioapi::document_position(4, 20)),
    rstudioapi::document_range(rstudioapi::document_position(5, 7), rstudioapi::document_position(5, 20)),
    rstudioapi::document_range(rstudioapi::document_position(6, 7), rstudioapi::document_position(6, 20)),
    rstudioapi::document_range(rstudioapi::document_position(10, 1), rstudioapi::document_position(10, 2))
)
rstudioapi::setSelectionRanges(rgs)

rstudioapi::insertText(lapply(1:5, function(x) c(x, 8, x, 9)), "🤷")
rstudioapi::modifyRange(Map(c, 1:5, 1), paste0("#", sample(letters, 5), " "))

Now run the rstudioapi code, line by line.

Notice that we handle wonky text quite well, with the possible exception, depending on your perspective, of an Emoji ZWJ Sequence. Like in #2132, to do any better, the whole toolchain would need an awareness it does have not of unicode normalization. I say this is good enough for the time being.

[jmcphers]

Should this throw an error?

[juliasilge]

I don't think so (???) because similarly to here we don't know that there is an active text editor currently. Maybe when we address #2571 we should also look at what gets returned when there is no active editor.

If you have no active editor in RStudio, it uses the console, which are not handling at all yet. Dealing with this for ranges is analogous to us eventually handling id here.

[jennybc]

I think we probably have some more work to do here, around multi-byte characters.

Here's my manual test:

1. Make a selection in the source editor.
2. In the Console, via typing / pasting/ or using the up-arrow, execute: x <- rstudioapi::getSourceEditorContext() to capture the current context, including selection data.
3. Now use rstudioapi::setSelectionRanges(x$selection[[1]][["range"]]) to complete the round trip, i.e. re-select the same selection.

We don't pass this test right now if, for example, the selection contains a emoji. RStudio does pass this test (although some of the numbers you see for positions are different). I think we want this sort of self-consistency, i.e. that we handle positions symmetrically for reading and setting.

My hunch is that PR needs to gain this sort of maneuver, but going the other way:

positron/src/vs/workbench/api/common/positron/extHostMethods.ts

Lines 137 to 168 in 7a3d14b

	// The selections in this text editor. The primary selection is always at index 0.
	//
	// The gymnastics here are so that we return character positions with respect to
	// Unicode code points. Otherwise, the native Position type provides offsets with respect to
	// UTF-16 encoded text. That would be confusing for downstream consumers, who probably
	// ultimately receive this text as UTF-8 and want to operate on this text in terms of
	// as user-perceivable "characters". This only matters when the selection's neighborhood
	// includes Unicode characters in the astral plane.
	//
	// Another resource that supports that what I'm doing here is desirable in Jupyter-land:
	// https://jupyter-client.readthedocs.io/en/latest/messaging.html#notes
	const selections = editor.selections.map(selection => {
	const lineTextBeforeActive = editor.document
	.lineAt(selection.active.line)
	.text.substring(0, selection.active.character);
	const unicodePointsBeforeActive = Array.from(lineTextBeforeActive).length;
	const lineTextBeforeStart = editor.document
	.lineAt(selection.start.line)
	.text.substring(0, selection.start.character);
	const unicodePointsBeforeStart = Array.from(lineTextBeforeStart).length;
	const text = editor.document.getText(selection);
	const unicodePointsInSelection = Array.from(text).length;
	return {
	active: { line: selection.active.line, character: unicodePointsBeforeActive },
	start: { line: selection.start.line, character: unicodePointsBeforeStart },
	end: { line: selection.end.line, character: unicodePointsBeforeStart + unicodePointsInSelection },
	text: text
	};
	});

---

Here's the basis of my manual test:

"🌷 b"

x <- rstudioapi::getSourceEditorContext()
f(x$selection[[1]])

rg <- x$selection[[1]][["range"]]
# creating the same document_range explicitly
# rg <- rstudioapi::document_range(
#   rstudioapi::document_position(1, 2),
#   rstudioapi::document_position(1, 5)
# )
rstudioapi::setSelectionRanges(rg)

# manual test cases
# select the text inside the quotes
# "a b"    round trip works
# "ä b"    round trip works
# "🌷 b"   round trip does NOT work, 'b' not part of selection
# "🤷‍♀️ b"   round trip does NOT work, 'b' not part of selection, but
#          for reasons documented elsewhere, we believe this is currently
#          out of scope

f <- function(sel) {
  cli::cli_inform("text: {.q {sel$text}}")
  cli::cli_inform("nchar(text): {.q {nchar(sel$text)}}")
  cli::cli_inform('
    range: [{sel$range$start["row"]}, {sel$range$start["column"]}] --
    [{sel$range$end["row"]}, {sel$range$end["column"]}]')
}

It's a little fiddly to get setup but here's what I do:

1. Define f() (at the end).
2. Copy x <- rstudioapi::getSourceEditorContext() or just execute it once to get it in the history. (I'm using rstudioapi::getSourceEditorContext() instead of rstudioapi::getActiveDocumentContext() to make it easier to run the same code in RStudio, where those functions are actually different.)
3. Then follow the steps above, replacing the text in line 1 with the test cases above. I manually select the text inside the double quotes.