internal: Don't mutate syntax trees when preparing proc-macro input #10025
Conversation
| @@ -512,8 +531,11 @@ impl TokenConvertor for Convertor { | |||
| if !&self.range.contains_range(curr.text_range()) { | |||
| return None; | |||
| } | |||
| self.current = curr.next_token(); | |||
|
|
|||
| self.current = match self.censor { | |||
There was a problem hiding this comment.
Oh wow, didn't realise that we already use TextRanges here. Ok I guess, but I'd refeactor that away some day -- I don't see a reason for us to rely on raw ranges here :)
There was a problem hiding this comment.
The if !&self.range.contains_range(curr.text_range()) seems unnecessary in the first place, as we get all tokens from the node that has that range here anyways.
I figured using a TextRange for the censor would be alright given we only want to censor one part in the node so far, which would be simpler to handle than a slice of nodes.
There was a problem hiding this comment.
Yeah, I guess that's ok, although, semantically I'd expect a hash set of nodes, which we can .skip_subtree in Preoreder
There was a problem hiding this comment.
Looks like that check wasnt pointless, makes sense that the token iteartion leaves the node if its not a root node.
Using preorder sounds nice compared to iterating the tokens via next_token, will adapt that to that instead.
There was a problem hiding this comment.
Actually using preorder seems non-trivial since we are interested in the tokens in order but that only gives us the nodes. I'll leave this as is for now.
Veykril
force-pushed
the
non-mut-exp
branch
2 times, most recently
from
1323c00 to
d50f268
Compare
|
bors r+ |
lnicola
changed the title
10289: fix: Only strip derive attributes when preparing macro input r=Veykril a=Veykril Fixes #10246 cc rust-analyzer/rowan#114, follow up to #10025 Co-authored-by: Lukas Wirth <[email protected]>
Fixes #10013