Refactor UnwrapCastInComparison to remove Expr clones#10115
Conversation
peter-toth
force-pushed
the
refactor-unwrapcastincomparison-to-remove-expr-clones
branch
2 times, most recently
from
cefe8b9 to
295ef58
Compare
peter-toth
force-pushed
the
refactor-unwrapcastincomparison-to-remove-expr-clones
branch
from
295ef58 to
22b55f4
Compare
|
The test failure: seems unrelated. |
| fn f_up(&mut self, expr: Expr) -> Result<Transformed<Expr>> { | ||
| match &expr { | ||
| fn f_up(&mut self, mut expr: Expr) -> Result<Transformed<Expr>> { | ||
| match &mut expr { |
There was a problem hiding this comment.
I was trying a few different patterns to remove Expr clones in this rule. The match &mut expr way seems to be most straightforward as expr is still available after pattern matching so we can return it unchanged if needed (return Ok(Transformed::no(expr));) but we can also mutate some parts of the expr if needed.
| }; | ||
| // unwrap the cast/try_cast for the left expr | ||
| **left = | ||
| mem::replace(left_expr, Expr::Literal(ScalarValue::Null)); |
There was a problem hiding this comment.
Would be nice to have a default Expr so that we could use mem::take().
There was a problem hiding this comment.
Shall I open a follow-up PR that adds default for Expr?
There was a problem hiding this comment.
mem::take is a nice thing to avoid cloning, I think I used it to avoid String cloning other day, and we can have a default Expr with panicking inside I'd say...
There was a problem hiding this comment.
I've opened a minor PR to add default for Expr: #10127, but as Expr is just an enum, I don't know how to panic inside it.
|
CI failure I think is due to #10119, so I retriggered the run |
alamb
left a comment
alamb
left a comment
There was a problem hiding this comment.
Very nice @peter-toth -- I actually tried this msyself last week and I was not able to get it to work. 👌
| // do nothing | ||
| } | ||
| Expr::BinaryExpr(BinaryExpr { left, op, right }) | ||
| if { |
There was a problem hiding this comment.
Today I learned you can add statements in a {} as part of a match clause. Nice!
There was a problem hiding this comment.
+1. I used to wrap another function.
|
Thanks @peter-toth @alamb @comphead |
|
Thanks all for the review! |
4 participants
Which issue does this PR close?
Part of #9637, follow-up to #10087.
Rationale for this change
Eliminates remaining
Exprclones.What changes are included in this PR?
This PR:
UnwrapCastExprRewriterto removeExprclones.BinaryExprcase as if any error occurs this optimizer rule should return the original expression.Are these changes tested?
Yes, with existing UTs.
Are there any user-facing changes?
No.