Handle RuleDelayed in Map over Association expressions

[mmatera]

One slight change over #1692. Also, I add a comment about differences with WMA output in the case that the Association expression does not have the form Association[__(Rule|RuleDelayed)].

[rocky]

LGTM. Thanks for the generalization and clarification.

[Li-Xiang-Ideal]

btw now MapAt maps on values for Association[__Rule] only:

mathics-core/mathics/eval/functional/apply_fns_to_lists.py

Lines 41 to 46 in a60d4f7

	if hasattr(replace_element, "head") and replace_element.head is SymbolRule:
	new_elements[j] = Expression(
	SymbolRule,
	replace_element.elements[0],
	Expression(f, replace_element.elements[1]),
	)

[rocky]

This mistake wasn't caught in the existing tests. So, probably a test needs to be added here for the new features added.

[mmatera]

The question is, what should we test? For the full behavior, I have in mind a bunch of tests, that in the current implementation would not pass. The example in the comment is one of them.
To match with the WMA behavior, this method should be rewritten, and probably it would be also be interesting to define a specific class (like ListExpression) to store valid Association expressions. None of this would be too complicated, but right now I do not have the time to handle it.
What I could do is to list the tests here, or just add the tests to a pytest module, and mark them as xfail.

[rocky]

The question is, what should we test?

So you are saying that nothing in this PR right now benefits a user? There is promise, though, that one day it will?

For the full behavior, I have in mind a bunch of tests, that in the current implementation would not pass. The example in the comment is one of them. To match with the WMA behavior, this method should be rewritten, and probably it would be also be interesting to define a specific class (like ListExpression) to store valid Association expressions. None of this would be too complicated, but right now I do not have the time to handle it. What I could do is to list the tests here, or just add the tests to a pytest module, and mark them as xfail.

Yes, it is better to add xfail or commented-out tests now while the issue is fresh in your mind rather than do it later. Thanks.

[mmatera]

The question is, what should we test?

So you are saying that nothing in this PR right now benefits a user? There is promise, though, that one day it will?

For the full behavior, I have in mind a bunch of tests, that in the current implementation would not pass. The example in the comment is one of them. To match with the WMA behavior, this method should be rewritten, and probably it would be also be interesting to define a specific class (like ListExpression) to store valid Association expressions. None of this would be too complicated, but right now I do not have the time to handle it. What I could do is to list the tests here, or just add the tests to a pytest module, and mark them as xfail.

Yes, it is better to add xfail or commented-out tests now while the issue is fresh in your mind rather than do it later. Thanks.

This PR improves handling of some typical cases, but to cover the most general situation, we need to reimplement this.

[rocky]

happends -> occurs

[mmatera]

Thanks!

[rocky]

happends -> occurs

[rocky]

happends -> occurs

[mmatera]

btw now MapAt maps on values for Association[__Rule] only:

mathics-core/mathics/eval/functional/apply_fns_to_lists.py

Lines 41 to 46 in a60d4f7

	if hasattr(replace_element, "head") and replace_element.head is SymbolRule:
	new_elements[j] = Expression(
	SymbolRule,
	replace_element.elements[0],
	Expression(f, replace_element.elements[1]),
	)

thanks for the observation!