Silas Davis created AVRO-1640:
---------------------------------
Summary: Does SpecificData.schemaCache actually do anything?
Key: AVRO-1640
URL: https://issues.apache.org/jira/browse/AVRO-1640
Project: Avro
Issue Type: Bug
Components: java
Affects Versions: 1.7.7
Reporter: Silas Davis
Priority: Minor
I've been looking at the source code for SpecificData. On line 182 a
schemaCache is defined as a WeakHashMap.
Having looked into how WeakHashMap works - it stores a WeakReference for the
keys of the map but not the values, I can't see how the schemaCache actually
provides a cache. I think it just stores all Schema objects it has ever seen,
and never expires any. This is because it stores keys of type
java.lang.reflect.Type key, typlically a Class<?>, which typically never fall
out of reference. The only way the Class (a static singleton object) will be
garbage collected is if the ClassLoader that holds a reference to it is garbage
collected. This mostly doesn't happen, so the schemaCache is really just
behaving as a HashMap.
It seems like WeakHashMap is commonly mistakenly used as a cache:
-
http://stackoverflow.com/questions/1802809/javas-weakhashmap-and-caching-why-is-it-referencing-the-keys-not-the-values
-
http://www.codeinstructions.com/2008/09/weakhashmap-is-not-cache-understanding.html
Perhaps I am missing something, like Avro's use of ClassLoaders meaning that
Class objects do get collected, apologies if so.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
