fix: infinite loop parsing incomplete struct #66
Conversation
The point of the error resilient parsing is to tolerate errors (accumulate them) but then continue parsing the rest of the file |
test/spitfire_test.exs
Outdated
| [ | ||
| {:__aliases__, [last: [line: 1, column: 2], line: 1, column: 2], [:Foo]}, | ||
| {:%{}, [closing: [line: 1, column: 8], line: 1, column: 6], [{:x, [line: 1, column: 8], nil}]} | ||
| ]}, [{[line: 1, column: 8], "missing closing brace for struct"}]} |
There was a problem hiding this comment.
I would be expecting the error message to say something about missing an opening brace for the struct. There might be something else not going correct here. I don't think %Foo is valid syntax by itself.
There was a problem hiding this comment.
Done in efafc6b in a way that also recovers from a bunch of other scenarios, and properly identifies wether the opening or closing braced for the right struct are missing
|
I tried integrating spitfire with elixir_sense and my custom error tolerant lexer (https://github.com/elixir-lsp/elixir_toxic) that returns valid token ranges for all tokens but I found spitfire too buggy. I created some property tests that generate elixir programs and there are just too many cases of spitfire misparsing valid code, crashing or looping that I didn't even started on error recovery. Long story short, I am working on another custom parser. |
Its not worth just fixing the bugs? |
|
Also, that repo is private |
doorgan
force-pushed
the
doorgan/infinite-recursion
branch
from
797a37d to
efafc6b
Compare
3 participants
Split from #65
Fixes
Spitfire.parse("%Foo = x")causing Spitfire to hangWith this fix the parser will continue while reporting an error. I'm not 100% sure if this is the behavior we want, we may want to fail the entire parsing, or report a different error.