Implemented WeakKeyIdDict #402
Conversation
Codecov Report
@@ Coverage Diff @@
## master #402 +/- ##
==========================================
- Coverage 98.65% 98.57% -0.09%
==========================================
Files 30 31 +1
Lines 1716 1760 +44
==========================================
+ Hits 1693 1735 +42
- Misses 23 25 +2
Continue to review full report at Codecov.
|
|
Is there any interest in this? Otherwise please close. |
|
Any news? |
|
After the great recent interest in this PR (two people in two month! ;-), I updated it such that it passes again tests and fixed conflicts. Ready to go, as far as I'm concerned. |
|
Thanks. This is quite useful! |
|
Per #479 I guess a case could be made that this PR was open long before then and so should be an exception? |
|
+1 for an exception, since I want to use this too :). Or maybe it can be it's own small package? |
| Base.get(wkh::WeakKeyIdDict{K}, key, default) where {K} = lock(() -> get(wkh.ht, WeakRefForWeakDict(key), default), wkh) | ||
| Base.get(default::Base.Callable, wkh::WeakKeyIdDict{K}, key) where {K} = lock(() -> get(default, wkh.ht, WeakRefForWeakDict(key)), wkh) | ||
| Base.get!(wkh::WeakKeyIdDict{K}, key, default) where {K} = lock(() -> get!(wkh.ht, WeakRefForWeakDict(key), default), wkh) | ||
| Base.get!(default::Base.Callable, wkh::WeakKeyIdDict{K}, key) where {K} = lock(() -> get!(default, wkh.ht, WeakRefForWeakDict(key)), wkh) |
There was a problem hiding this comment.
It would be nice to use Base.@lock here instead of the closure form. It can be more performant and keeps stacktraces shorter; I have a recursive call where this almost doubles the number of stack frames. It is exported on Julia 1.9 and present in many versions of Julia, but not present in 1.0 unfortunately, which DataStructures still supports.
This is what I put together in JuliaLang/julia#28161 (closed PR). I thought it could perhaps live here. Note that a WeakKeyIdDict might go into Base sometime in 1.x: JuliaLang/julia#28182, but who knows when.