Update serialization extension to support custom cache-size metric #52
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #52 +/- ##
==========================================
+ Coverage 81.92% 81.97% +0.05%
==========================================
Files 3 3
Lines 354 355 +1
==========================================
+ Hits 290 291 +1
Misses 64 64 ☔ View full report in Codecov by Sentry. |
ext/SerializationExt.jl
Outdated
| @@ -29,20 +29,21 @@ function Serialization.deserialize(s::AbstractSerializer, ::Type{LRU{K, V}}) whe | |||
| lock = Serialization.deserialize(s) | |||
| by = Serialization.deserialize(s) | |||
| finalizer = Serialization.deserialize(s) | |||
| n_items = Serialization.deserialize(s) | |||
There was a problem hiding this comment.
is there a way to detect there are no bytes left to deserialize and fallback to n_items = currentsize here? to continue supporting de-serializing old serialized files
There was a problem hiding this comment.
(if not, I'm not sure this should be a patch release, it seems like it could unexpectedly break stuff. Not quite sure what folks use LRU serialization for though, is it for IPC only or for persistence?)
There was a problem hiding this comment.
That's a good point; this will break deserializing old serialized data. I think we can make this backward compatible, and that would be nicer in case there are some persisted LRUs out there.
Any existing file would have n_items in the currentsize position, so we can continue storing n_items there, and always recompute currentsize.
There was a problem hiding this comment.
I updated the PR. Thanks for pointing this out!
|
This seems good to me as well. Thanks for the great changes! |
The current Serialization extension assumes the cache size is always the number of items in the cache. This PR allows serializing LRUs with custom
byfunctions, wherelru.currentsize != length(lru).byfunctions #51