export Base.structdiff

[tpapp]

Base.structdiff was introduced more than 2 years ago in #24795, but not exported.

But since it is useful for efficient and type stable named tuple manipulation, it is now used by some packages working with NamedTuples, eg BangBang.jl, FunctionalTables.jl, NamedTupleTools.jl [these are just examples, I may have missed some packages].

It would be nice to have it exported, or at least officially declared part of the API.

@tkf @JeffreySarnoff

[JeffBezanson]

I'm ok with that. Maybe it needs a more general name, since it could also be extended to Dicts? (E.g. merge works for both dicts and named tuples)

[JeffreySarnoff]

We have setdiff "Construct the set of elements in s but not in any of the iterables in itrs"

setdiff(set, elem[s]togo), structdiff(struct, field[s]togo) and dictdiff(dict, key[s]togo) -->
seems like without(collective, item[s]togo)

a without x without(a, x)

[tkf]

Just brainstorming: keydiff, keysetdiff, antijoin

[JeffreySarnoff]

bump

[JeffreySarnoff]

If naming is all that is in the way of this, perhaps omit or else keysdiff

[goretkin]

Perhaps fleshing out the analogies is helpful

AstractSet	NamedTuple / AbstractDict
identity	keys
union	merge
setdiff	Base.structdiff, NamedTupleTools.delete
intersect	NamedTupleTools.select

(The analogy can be loose because the second argument might be just a collection of keys, and not an association of keys and values)

Are there more?

I think key should be in the name since keys generalizes keys(::AbstractDict) and propertynames(::NamedTuple). Or perhaps pair should be in the name. pairdiff?

[JeffreySarnoff]

For AbstractDict and for NamedTuple, the action in lines 2:4 of your table involves maintaining an association of key[s] with value[s] (even when keeping the association removes some key-value associands from the result or replaces some key-value pair by changing the value). For that reason, pairdiff (or more readably pairsdiff) works for me. (This operation is not limited to key interactions, it seems to me.)

[JeffreySarnoff]

or diffkeys or really anything that this thread finds ameanable

[Eben60]

bump

[vtjnash]

Please do not bump feature request issues, unless you have a PR for itt