Add "access" scope type

[pokey]

• Implements Create scope type for link in a chain of field accesses #707 for Typescript
• Implements Create scope type for link in a chain of field accesses #707 for Python

Questions

• Unfortunately, both the queries and tests are almost identically duplicated between Typescript and Python 😕.
  • I wonder if we should support some kind of templating language, eg handlebars, along with a meta-updater-style approach where we keep the generated files in source control as well
  • For the test cases, we could maybe support a list of language ids in the recorded test yaml 🤷‍♂️
• Also, should the . be considered part of the content range, or just the removal range?
• Should a function call on its own be considered "access"? eg foo(). What about an identifier on its own, eg foo?

Checklist

• Make it so eg bar or foo() on its own is an access? That might cover a lot of things 🤔. Eg what happens with something in parens? That could end up as part of access, eg (a + b).c
• What does happen with (a + b).c?
• I have added tests
• I have updated the docs and cheatsheet
• I have not broken the cheatsheet

[pokey]

wasn't sure what to call nodes like this where we're just using them to allow us to write a predicate that rejects the pattern. We should probably standardize on something that allows multiple per pattern, because in the future I'd like to be strict about allowable capture names to detect typos in capture names, which today just result in scopes silently not working

[AndreasArvidsson]

I don't think the dot should be part of the content range. I view this as a delimiter.

I definitely understand what you mean about the test cases. On fixture for multiple languages might be reasonable.

[pokey]

yeah I'm on the fence, because it is a bit like a delimiter, but not quite. Comparing it to "arg", in that case, an arg can exist without the ,, whereas something that is after the first element in the chain can't exist without the .. Ie you can't take it and make it a single-element access chain without a . the way you can with an arg. So the . really feels more like it is a part of the access. That being said, I do think you'd probably expect "change" to leave the . 🤔

To make it a delimiter, we'd need to:

1. support insertion delimiters for our tree-sitter queries, and
2. Ensure that a single function call / identifier / access is considered an "access"

Which are not unreasonable anyway imo

[pokey]

I guess one argument to treat it as a delimiter is that the decision to use ?. instead of . is determined by the preceding identifier rather than the following, so arguably it's not wholly owned by the trailing access

[pokey]

aligned on having it be delimiter, so need the extra things above

[pokey]

Update from meet-up (see #1923 (comment) for more context):

• Make it private by not having it appear in csvs, but don't throw error if we manually add it to the csv. Also call it private.access
• Ship it as is where it includes the . in the content range. If it's annoying we'll revisit

[pokey]

Ok it's now a private scope type. You'll need to add a line like the following to your modifier_scope_types.csv:

# WARNING: Please do not copy this into your own CSVs; it is for internal
# experimentation and may break at any time without warning.
# See https://github.com/cursorless-dev/cursorless/pull/1519#issuecomment-1745459484
access, private.fieldAccess

@AndreasArvidsson it's worth giving this a whirl to make sure that:

1. You can't use it without adding the above line, and there's no error if you don't
2. It works if you do add the above line

NOTE to anyone stumbling across this PR: if you do not heed the warning above, and do paste the above lines in your CSV, please at least subscribe to notifications for this PR; we'll leave a message here when we deactivate this feature in the future 😊

[pokey]

Keep in mind that someone could still copy one of our settings csvs and get it that way, but hopefully the comment helps 🤷

[AndreasArvidsson]

Tested and works as expected

This is also one of those scopes where I would like to have a text based fall back :D

[pokey]

Yeah that would be nice. Could prob build something using the pair scopes once those are migrated

[pokey]

@AndreasArvidsson I think this one is good to go when you get a minute

[AndreasArvidsson]

There are a lot of unfinished tasks on this pr and, but you're still happy to merge?

[pokey]

yeah I should prob check them, but I left them because this PR is really kind of unfinished but we want to merge it as a private feature to get some feedback from real use