Don't rewrite ~ @foo to ~@foo

[frenchy64]

The problem is ~ @foo is being reformatted to ~@foo, a different expression. ~,@foo is used as the canonical form because it seems to have more emphasis and stability than ~ @foo (just by my taste, lmk).

This decision has knock-on effects if there are are other forms between ~ and @ which are demonstrated in the tests.

The implementation uses transform twice, but it should ideally just be one tree pass. Please let me know if there's a preferred way to do this.

[weavejester]

Thanks for the bug report, and the PR. Can you explain why you used a volatile? I don't understand why it's necessary. Also, why is ~ @foo rewritten to ~,@foo? Why not keep it as ~ @foo?

[frenchy64]

Can you explain why you used a volatile? I don't understand why it's necessary.

Starting with the premise of using transform twice, I used a volatile as a performance optimization to only perform the second walk if absolutely necessary.

It would be better to have just one walk but I didn't know the best way to do this.

Also, why is ~ @foo rewritten to ~,@foo? Why not keep it as ~ @foo?

This part is up to you.

My thought was: it's a really strange example of nested syntax that requires a space between the container and the child. I can't think of any other examples like it.

My idea with ~,@foo was to make the disambiguation more explicit to draw attention to itself.

[weavejester]

I think I'd like to know why cljfmt is removing the space in the first place. In the PR, code is changed for insert-missing-whitespace, presumably to re-insert the space that's removed by some other mechanism, but why not fix the code that causes it to remove the space in the first place?

[frenchy64]

Ah, I misinterpreted insert-missing-whitespace as reinserting essential whitespace that was removed by cljfmt. I'll look into your suggestion.

[frenchy64]

@weavejester seems to work!

[weavejester]

Much better! I don't think I have any more notes. Could you squash your commits down?

[frenchy64]

Done.