Unneccessary ast optimization

[Eclips4]

Bug report

Bug description:

AST optimizer is trying to optimize target of the following statements:

• del statement (del a) - target is a target_list
• Assign statement (a = b) target is a target_list
• AugAssign statement (a += b) target is a augtarget
• AnnAssign statement (a: int or a: int = 1) target is a augtarget
• TypeAlias statement (type A = int) target is a identifier
• For statement (for a in b: pass) target is a target_list
• Async for statement (async for a in b: pass) target is a target_list

None of these targets contain anything that can be optimized.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-123403: Remove unnecessary "target" ast optimizations #123404

[JelleZijlstra]

Many (probably all) of these can in fact be optimized.

>>> ast.dump(ast.parse("del [1, 2][0 + 1]"))
'Module(body=[Delete(targets=[Subscript(value=List(elts=[Constant(value=1), Constant(value=2)], ctx=Load()), slice=BinOp(left=Constant(value=0), op=Add(), right=Constant(value=1)), ctx=Del())])])'
>>> ast.dump(ast.parse("del [1, 2][0 + 1]", optimize=1))
'Module(body=[Delete(targets=[Subscript(value=List(elts=[Constant(value=1), Constant(value=2)], ctx=Load()), slice=Constant(value=1), ctx=Del())])])'

[JelleZijlstra]

I think the only exception is the name of a type alias; that really can only be a name. But even then I'm not sure it's worth skipping the AST optimizer; it's useful to have it recurse into everything so we don't have to think about whether we're missing something.

[Eclips4]

I think the only exception is the name of a type alias; that really can only be a name. But even then I'm not sure it's worth skipping the AST optimizer; it's useful to have it recurse into everything so we don't have to think about whether we're missing something.

I'm agree with you in general, but I'd like to skip AST optimization on a type alias :) Not a big deal, though.

Also, these optimizations related to assignments (as you have shown above with a del statement) haven't been tested. Do you think they're worth testing?

[JelleZijlstra]

Do you think they're worth testing?

No strong opinion. I guess in theory we could test every AST node in the optimizer tests but that feels a bit repetitive.

[Eclips4]

I'm going to close this issue; I may open another issue about testing ast optimizations.