Stream: compiler development

Topic: ✔ whitespace sensitivity - match expressions

Luke Boswell (Jul 01 2025 at 04:16):

I was getting confused about some parsing errors while working on match expressions. Claude thinks this is the issue.

Looking at these two snapshots, the key difference is indentation:

basic_tag_union.md (works correctly):

match color {
    Red => 1
    Blue => 2
    Green => 3
}

boolean_patterns.md (fails to parse):

match isReady {
    True => "ready to go!"
    False => "not ready yet"
}

The working version uses tabs for indentation, while the broken version uses 4 spaces. You can see this in the token positions:

• basic_tag_union: UpperIdent(2:2-2:5) - pattern starts at column 2 (tab indentation)
• boolean_patterns: UpperIdent(2:5-2:9) - pattern starts at column 5 (4-space indentation)

Luke Boswell (Jul 01 2025 at 04:18):

I'm going to investigate a bit further and try and fix. @Anthony Bullard FYI

Luke Boswell (Jul 01 2025 at 04:24):

Ignore the above... on deeper inspection I can see the issue is with parsing string literals instead

Luke Boswell (Jul 01 2025 at 04:29):

We don't expect any comma's between match branches right? ... I've modified the parser to use newlines and it seems to be happy now.

Notification Bot (Jul 01 2025 at 05:44):

Luke Boswell has marked this topic as resolved.

Anton (Jul 01 2025 at 09:58):

Luke Boswell said:

We don't expect any comma's between match branches right? ... I've modified the parser to use newlines and it seems to be happy now.

Hmm, I seem to remember something about that, I think it was in a conversation with @Anthony Bullard

Anthony Bullard (Jul 02 2025 at 22:46):

i do too, but can't remember where it was or what the outcome was

Richard Feldman (Jul 02 2025 at 23:27):

I think ideally we don't need them

Last updated: Jul 06 2025 at 12:14 UTC