Decompressor/Compressor pool #34
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #34 +/- ##
==========================================
- Coverage 94.55% 93.30% -1.25%
==========================================
Files 3 4 +1
Lines 202 269 +67
==========================================
+ Hits 191 251 +60
- Misses 11 18 +7 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| // keep finishing stream until we don't get a buffer overflow | ||
| while true { | ||
| do { | ||
| try lastBuffer.compressStream(to: &self.window, with: self.compressor, flush: .finish) | ||
| try lastBuffer.compressStream(to: &self.window, with: compressor, flush: .finish) | ||
| try await self.parentWriter.write(self.window) |
There was a problem hiding this comment.
If this write fails, the allocation isn't freed. Can we use a non-copyable with a deinit to handle this instead? Or a class or something similar.
There was a problem hiding this comment.
Using AllocatedValue class to hold these.
|
We can also lend an allocation using |
Unfortunately the way ResponseBodyWriters work doesn't lend themselves to |
Yes I understand, but we can't pool these without managing the allocation of them. |
Add a pool of decompressors and compressors for both gzip and deflate. When middleware needs either it will use one in the pool if it exists otherwise it will allocate a new one. When freeing the compressor/decompressor it will pass it back to the pool if the pool hasn't already reached its maximum size.
This should reduce the amount of allocations made by zlib considerably