Support jump to sub character by letter name

[AndreasArvidsson]

eg:
take glyph air in fine -> Select first instance of character "a" in token "fine"

(@pokey edited to use "glyph" instead of "char" to reserve "char" as a modifier that converts a target to refer to its individual characters rather than by token)

[pokey]

Given that subtoken transformations are expected after the mark, I'd imagine we should follow that convention here? eg "take fine char air". We could think about switching subtoken transformations to be before the mark like they are with other transformations, but either way I think we should be consistent between this transformation and other subtoken transformations

[AndreasArvidsson]

You can actually do both right now
take last word oddand take odd last wordboth works at the moment.

[pokey]

Yep that's correct. But per our current direction I'd argue we pick one to support

[AndreasArvidsson]

I agree

[pokey]

we should think about how this one interacts with the new grammar. I think we might want to remove the "in"

I think we could also support "second char air" or "second air"

Tho we'd want to make sure the grammar isn't ambiguous with #13 in its ordinal form

[josharian]

somewhat related: #1822

both are about how to get at things in the middle of very long tokens efficiently

[pokey]

somewhat related: #1822

both are about how to get at things in the middle of very long tokens efficiently

Agreed, tho fwiw this issue should really be going much broader than that. It's myopic because it's old and was never updated. We should also be able to say things like "next air", "every air line", etc

[pokey]

Fwiw this issue would be trivial to implement now. Would just use nested scope handler with line iteration scope, and yield instances of the char in the line

[josharian]

I was doing some csv processing today and really wanted this to be able to chuck columns (chunk head line second glyph drip every line). So I looked into implementing it, and got stuck. I couldn't figure out how to get the required glyph arg in there.

Maybe we could live-code it together on Sunday? (By which I obviously mean I watch you live code it 😉)

[pokey]

Have a look at eg custom regex scope type or surrounding pair scope type. See how they have fields with the extra data they need

[pokey]

cursorless/packages/common/src/types/command/legacy/PartialTargetDescriptorV3.types.ts

Lines 158 to 174 in f17bdf2

	interface CustomRegexScopeType {
	type: "customRegex";
	regex: string;
	}
	type SurroundingPairDirection = "left" | "right";
	interface SurroundingPairScopeType {
	type: "surroundingPair";
	delimiter: SurroundingPairName;
	forceDirection?: SurroundingPairDirection;
	/**
	* If `true`, then only accept pairs where the pair completely contains the
	* selection, ie without the edges touching.
	*/
	requireStrongContainment?: boolean;
	}