Add Set candidate for SetHash's STORE

[MasterDuke17]

So set metaops like ∖= will DWIM. A possible final fix for GH #1203. perl6 -e 'my %days := SetHash.new: Date.today … Date.new: "2018-06-02"; say %days.elems; %days ∖= %days.grep: *.key.day-of-week > 5; say %days.elems' now prints out 14 and 10, instead of 14 and 1.

Rakudo builds ok and passes make m-test m-spectest (will add some new test to roast if this is merged).

[lizmat]

FWIW, I'm -1 on this fix atm on the grounds that it doesn't feel right: will need some time to look into the problem deeper, which will be tomorrow.

[MasterDuke17]

To add some additional data, the docs say that the set operators on SetHashes return immutable sets (https://docs.perl6.org/type/SetHash#Operators), which does seem to be specced in roast (e.g., https://github.com/perl6/roast/blob/master/S03-operators/set_symmetric_difference.t#L43).

[MasterDuke17]

@lizmat ping. Any further thoughts?

[lizmat]

Looking at this again now :-)

[MasterDuke17]

ISTR @jnthn had some comments on IRC, I'll see if I can find them and add them here.

[MasterDuke17]

I haven't found the chat yet, but I think it was something about how it should be done as a new candidate for METAOP_ASSIGN (https://github.com/rakudo/rakudo/blob/master/src/core/metaops.pm6#L2-L4), but since that just takes an op and returns a sub, it would be a bit tricky.

[MasterDuke17]

Ah, here it was, http://colabti.org/irclogger/irclogger_log/perl6-dev?date=2018-06-10#l180

[lizmat]

Sorry, rejecting this PR on the same grounds as #3013