[Bug #21513] Raise on converting endless range to set #13902
Conversation
range.c
Outdated
| range_to_set(VALUE range) | ||
| { | ||
| if (NIL_P(RANGE_END(range))) { | ||
| rb_raise(rb_eRangeError, "cannot convert endless range to a set"); |
There was a problem hiding this comment.
I believe that endless range error should be the same error type as beginless range error. Calling to_set on a beginless range currently raises TypeError. This method should probably have beginless range raise RangeError instead, and include a test and an example for that.
There was a problem hiding this comment.
Makes sense. Should we also tweak Range#to_a to raise RangeError in both cases?
./ruby -e '(..1).to_a'
-e:1:in 'Range#each': can't iterate from NilClass (TypeError)There was a problem hiding this comment.
Makes sense. Should we also tweak
Range#to_ato raiseRangeErrorin both cases?
@nobu what do you think?
There was a problem hiding this comment.
It feels clearer for me.
@knu Any thoughts?
If we will change it too, maybe ruby/spec also needs to be updated?
There was a problem hiding this comment.
We should perhaps check the size of a given Enumerable in Set.new and raise an error when it is infinity.
There was a problem hiding this comment.
We should perhaps check the size of a given Enumerable in Set.new and raise an error when it is infinity.
Seems good to me. I'll try to follow this path
There was a problem hiding this comment.
I tried out this idea and it seems to work fine (we now raise ArgumentError error for any endless enumerable, e.g. 1.upto(Float::INFINITY).
$ ./ruby -e 'Set.new((1..nil))'
-e:1:in 'Set#initialize': cannot initialize Set from a enumerable with infinite size (ArgumentError)
from -e:1:in '<main>'
$ ./ruby -e '(1..nil).to_set'
-e:1:in 'Enumerable#to_set': cannot initialize Set from a enumerable with infinite size (ArgumentError)
from -e:1:in '<main>'
4b9f122 to
31a3861
Compare
|
There's a SEGV in the ZJIT test suite: https://github.com/ruby/ruby/actions/runs/16305765996/job/46051383888?pr=13902 Not sure if it's related to my changes |
31a3861 to
a5a7db6
Compare
ZJIT error seems entirely unrelated to your change from a quick glance |
|
Looks like a weird GC issue related to marking though so paging in @k0kubun for good measure |
a5a7db6 to
e1b6f53
Compare
e1b6f53 to
ca35402
Compare
❌ Tests Failed✖️no tests failed ✔️62191 tests passed(1 flake) |
|
I agree that it's probably not related to the changes. We're looking into GC issues with ZJIT. |
7753557 to
2de1bf0
Compare
|
Could someone please take a look? |
2de1bf0 to
36e3cee
Compare
set.c
Outdated
| @@ -494,6 +494,14 @@ set_i_initialize(int argc, VALUE *argv, VALUE set) | |||
| rb_check_arity(argc, 0, 1); | |||
|
|
|||
| if (argc > 0 && (other = argv[0]) != Qnil) { | |||
| ID id_size = rb_intern("size"); | |||
There was a problem hiding this comment.
Should move this to the following else block, so it won't be called for arrays, as arrays never have infinite size.
set.c
Outdated
| if (rb_respond_to(other, id_size)) { | ||
| VALUE size = rb_funcall(other, id_size, 0); | ||
| if (RB_TYPE_P(size, T_FLOAT) && RFLOAT_VALUE(size) == INFINITY) { | ||
| rb_raise(rb_eArgError, "cannot initialize Set from a enumerable with infinite size"); |
There was a problem hiding this comment.
You aren't checking that the argument is enumerable, only that the size is infinite. It would be more correct to say "an object" instead of "a enumerable". Even if you keep the current wording, it should be "an enumerable" instead of "a enumerable".
ref: https://bugs.ruby-lang.org/issues/21513 Before this patch, trying to convert endless range (e.g. `(1..)`) to set (using `to_set`) would hang
36e3cee to
f9dce06
Compare
|
@jeremyevans thank you very much for helping me with that! ❤️ |
Before this patch, trying to convert endless range (e.g.
(1..)or any other inifinte enumerable) to set (usingto_set) would hang