Add WeakRef class #3276
Add WeakRef class #3276
Conversation
bcardiff
force-pushed
the
feature/weakref
branch
from
2f128b0 to
7a0c700
Compare
| @@ -0,0 +1,70 @@ | |||
| require "spec" | |||
|
|
|||
| class State | |||
There was a problem hiding this comment.
State and Foo might eventually conflict with other specs. Maybe use names like WeakRefState and WeakRefFoo?
This is one of the reasons I want to implement file-private types.
There was a problem hiding this comment.
I would prefer a spec-specific solution, of somehow being able to define helper classes and functions in the namespace of a describe Foo statement. This seems possible with macros, and I much prefer it as a solution than having file-private classes and methods.
There was a problem hiding this comment.
file-private classes are a general solution, I don't understand why we should go with a less general solution. I also want to use it here, and this is not a spec. Sometimes there are some classes that only makes sense in a file, just like some functions that only make sense in a single file.
bcardiff
force-pushed
the
feature/weakref
branch
from
7a0c700 to
5436fd6
Compare
| end | ||
|
|
||
| def self.allocate | ||
| GC.malloc_atomic(sizeof(WeakRef(T))).as(WeakRef(T)) |
There was a problem hiding this comment.
Does using self instead of WeakRef(T) work? (shorter and maybe more clear)
bcardiff
force-pushed
the
feature/weakref
branch
from
5436fd6 to
03a2dd3
Compare
|
I didn't update gc/null.cr to match gc/boehm.cr. Should I? They already differ on |
|
I don't think |
|
It could be a pro to keep gc/null up to date - to maintain perspective, not unnecessarily naming/implementing funcs that are too tied to Boehm specifically (in this case, primarily the name), if a more general interface could easily be defined. For the future. |
Weak references are implemented by boehm's register_disappearing_link
bcardiff
force-pushed
the
feature/weakref
branch
from
03a2dd3 to
efead56
Compare
|
@ozra , any gc-agnostic terms you would suggest? Let's handle gc/null or rewording of the methods in a separate issue, if needed. |
|
Yes, I guess when the time comes, a few refactors in stdlib code won't be too much handle (not much user code should rely on GC-funcs anyway), so ignore my bike-shedding :-) I started on a PoC-GC-experiment for Crystal a few months ago, but it's far down on my urgent list, otherwise I wouldn't even have blinked ;-) Until the parallelism model is settled, there's not much point making a hot rod anyway. |
Weak references are implemented by boehm's register_disappearing_link