✨Use parsed float value for useless comparison warnings #5030
Conversation
| float_value: NotNan( | ||
| 1.0, | ||
| ), |
There was a problem hiding this comment.
Isn't this the ONE constant you defined earlier?
| float_value: NotNan( | |
| 1.0, | |
| ), | |
| float_value: NotNan::ONE, |
There was a problem hiding this comment.
This is the snapshot of the parser test, not rust code! 🕵️
There was a problem hiding this comment.
Oh my, sorry for that, I totally missed the context 😆
| 1 │ pub fn main() { 1_0.0 == 10.0 } | ||
| │ ^^^^^^^^^^^^^ This is always `True` |
compiler-core/src/parse.rs
Outdated
| /// A thin f64 wrapper that does not permit NaN. | ||
| /// This allows us to implement `Eq`, which require reflexivity. | ||
| /// | ||
| /// Used for gleam float literals. May contain infinity. |
There was a problem hiding this comment.
If this is used for Gleam float literals it shouldn't contain infinity as well, should it?
There was a problem hiding this comment.
This is true for the Erlang target (and there is a compile time check for it), but on JS you can enter float literals that are too large (e.g. 1.0e500) and they turn into (negative) infinity.
There was a problem hiding this comment.
Aha, now I see, thanks for explaining! Would you mind adding this explanation to the comment? That would be really helpful for future people wondering the same thing
There was a problem hiding this comment.
Good point, done!
fruno-bulax
force-pushed
the
parse-floats
branch
from
9ca2823 to
6fa1918
Compare
fruno-bulax
force-pushed
the
parse-floats
branch
from
6fa1918 to
573bd5d
Compare
2 participants
Closes #5001
Parallel to the implementation for ints, we're now parsing the literal value and passing it along the AST.
This allows us to do smarter useless comparison warnings instead of relying on string comparison.
To maintain the
Eqtrait impl on many types, I created aNotNantype to wrap f64 that beNaNand can therefore properly implementEqandOrd. I wasn't quite sure where to put the type, I'll gladly move it.I noticed that we still do string-based checks in case expressions for both ints and floats, I'll create a follow-up issue after this is merged.
I'm not sure how you'd like the changelog updated when there's an RC out. Do I just add a new
Unreleasedheader at the top?