It looks like a similar approach -- the parser functions now take the
namespace of the enclosing scope (the Ruby version calls it
"default_namespace" and I call it "parent_namespace"). Plus there's some
extra logic to take into account the fully-qualified name.
Thanks for pointing that out.
Ben
On Tue, 7 May 2013, Martin Kleppmann wrote:
FWIW, I think the Ruby bug was fixed here:
https://issues.apache.org/jira/browse/AVRO-1272 — take a look at the patch,
perhaps you can adopt a similar approach in the C implementation (I don't
know how similar the code is though, so it may be irrelevant).
Martin
On 7 May 2013 02:30, Bruce Mitchener <bruce.mitche...@gmail.com> wrote:
On May 7, 2013, at 5:17 PM, Ben Walsh <ben_w_...@yahoo.co.uk> wrote:
Hi
It seems that the C version of Avro doesn't resolve fully-qualified type
names, and it doesn't understand inherited namespaces.
For example this doesn't work:
{"type": "record", "namespace": "x", "name": "Y", "fields": [
{"name": "e", "type": {"type": "record", "name": "Z", "fields": [
{"name": "f", "type": "x.Z"}
]}}
]}
It can't find the type "x.Z".
(this is similar to a Ruby problem
https://issues.apache.org/jira/browse/AVRO-635).
I have a patch to make this work. Should I submit a JIRA or is this
issue already fixed/not a problem?
Sounds like a patch in JIRA would be ideal.
- Bruce