Github user afs commented on a diff in the pull request:
https://github.com/apache/jena/pull/391#discussion_r179411949
--- Diff: jena-tdb/src/main/java/org/apache/jena/tdb/lib/NodeLib.java ---
@@ -53,26 +52,32 @@
// Characters in IRIs that are illegal and cause SSE problems, but we
wish to keep.
final private static char MarkerChar = '_' ;
final private static char[] invalidIRIChars = { MarkerChar , ' ' } ;
+ final private static int SIZE = 1024;
+ // Marshalling space.
+ final private static ByteBuffer workspace = ByteBuffer.allocate(SIZE);
--- End diff --
No problem about the review timing - any PR is as far as I'm concerned on
permanent review. The more eyes on concurrency the better.
It is safe. It gets called from `NodeTableNative.writeNodeToTable` which
itself (see its comment) is inside synchronization. Several things need to
happen together - write the node and update the node cache - so the sync is
further out in `NodeTableNative`.
The comment on `encodeStore` where the byte buffer is used applies "Callers
must synchronize to ensure writing is not concurrent."
For decode, the ByteBuffer comes from `ObjectFileStorage.read`.
I can add a comment to the `workspace` declaration. (I'd _like_ scoped
statics inside methods.)
---
