Commit Graph

2164 Commits

Author SHA1 Message Date
ejona 4d642200e7 [2/2] Swap Operation wrapping from Channel to Transport.
Session is now (properly) implementing transport API, so ChannelImpl has some
testing.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=69074094
2015-01-08 14:42:23 -08:00
ejona 4fd4845481 Remove fluent-style from newtransport Stream.
The fluent-style wasn't being used, yet made it impossible to
simultaneously implement both ClientStream and ServerStream. Having a
single class that switches between client and server behavior based on a
boolean is helpful to some transports.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=69028356
2015-01-08 14:42:23 -08:00
ejona 5e01cea1ef Change newtransport to use c.g.n.stubby.Status.
c.g.n.stubby.Transport.Status isn't intended to be used (at least at
this point).
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=69027874
2015-01-08 14:42:23 -08:00
ejona 35cabd0b82 [1/2] Split Channel and Stub APIs, implement Channel.
Call does not accept InputStream length because Marshaller makes it too hard
for the Stub to determine the length.

SessionCall is updated to implement changes to Call, but it is replaced in the
next CL.

You can see what glue is necessary to join the Transport and Stub APIs. The
ugliest pain is dealing with flow control callback. Some of that code is
unrelated and necessary because the Channel uses a separate thread to run
Call.Listener. Other parts will be necessary because Channel must intervene
during Transport callbacks to handle retry logic.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=69020275
2015-01-08 14:42:23 -08:00
lryan 0cf7d4c353 Bring netty5 up to head to support HTTP2 draft 12 so we can test with GFE
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=69006100
2015-01-08 14:42:22 -08:00
nathanmittler d784765814 Updating the transport Stream API to allow for callbacks for when a message was accepted by flow control.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=68706844
2015-01-08 14:42:22 -08:00
ejona 2af35d9bd3 Improve JavaDoc, including style fixes
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=68555036
2015-01-08 14:42:22 -08:00
lryan f8bbc12468 Remove SPDY support from GRPC
Will re-create the E2E tests through GFE when we can properly initiate HTTP2 with GFE.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=68488736
2015-01-08 14:42:21 -08:00
nathanmittler ef31a5f2ae First cut at the GRPC Transport interfaces for Java.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=68458345
2015-01-08 14:42:21 -08:00
ejona 8be938ee25 Move newstub to third_party and rename to stub.
It was an oversight that newstub was not in third_party when committed.

Renaming of package was preformed via a script.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=68416211
2015-01-08 14:42:20 -08:00
lryan 2ce8446a0d Large refactor to:
- Introduce 'Channel' & 'Call' interfaces
- Unify the surfaces for the prototype generated stubs
- Lighten dependency on MessageLite outside of generated code (see Marshaller)
- Bridge Channel to Session pending Transport interface rewrite
- Update all tests to new interfaces

-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=68407406
2015-01-08 14:42:20 -08:00
zhangkun 347a22dde0 Add toString() to Status.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=67721763
2015-01-08 14:42:20 -08:00
ejona 07d3f6a202 Move gRPC core to third_party
Half our tests still need to be moved, but that will be for a later
time.
-------------
Created by MOE: http://code.google.com/p/moe-java
MOE_MIGRATED_REVID=67023748
2015-01-08 14:42:19 -08:00
ejona de5413f6c2 Initial commit 2015-01-08 14:42:02 -08:00