feat!: add HybridResolver #62
Conversation
| message MyMessageSet { | ||
| option message_set_wire_format = true; | ||
| extensions 100 to max; | ||
| } |
There was a problem hiding this comment.
It was necessary to delete this plus all references to it because message sets are a protov1 feature unsupported by the official protobuf and so registering this file descriptor was failing
| @@ -596,13 +603,52 @@ func MessageType(name string) reflect.Type { | |||
|
|
|||
| // A registry of all linked proto files. | |||
| var ( | |||
| protoFiles = make(map[string][]byte) // file name => fileDescriptor | |||
| protoFiles = make(map[string][]byte) // file name => fileDescriptor | |||
| gogoProtoRegistry = new(protoregistry.Files) | |||
There was a problem hiding this comment.
Maybe this can actually be made public like *GlobalFiles and then other checks can be performed by the caller if needed?
There was a problem hiding this comment.
Exported this as GogoResolver. Let me know if you see any issue with that
There was a problem hiding this comment.
I think it should be okay exported.
|
@mark-rushakoff |
|
Also this is technically a breaking change to a public API. I'm pretty sure this only being used by the SDK so it may be safe to consider the previous API as pre-release, but it wasn't tagged that way. Any suggestions on how to deal with this? I can do a rewrite to keep the previous API and add a new one, but not sure it's worth the effort. We could also retract the previous tag maybe... |
|
I'm okay with a tag retraction in go.mod, as the recent tag is only a couple weeks old and should only have been in use in the SDK. |
|
What do you think is the best way to approach the retraction? Delete the tag and create a new one? Or do we actually need a retract directive in the go.mod? |
|
Also, really confused by the CI errors here. Seems like there's a lot of stuff failing unrelated to any changes in this PR. |
|
I think just a retract directive in go.mod is fine. That might be a slight misuse of retraction, but I think it's okay given the narrow scope here -- "don't use this tag as it was a breaking API change, which we undid in the following tag". I understand the CI error message -- we are enforcing the leading slash somewhere as an expectation against the cosmos protos we produce. But I am not immediately seeing why that just started happening on this PR. |
…delete tests with pathological import
|
I added a retract directive for 1.4.8 The test failures seem due to registration errors in the
Going to move forward with merging this unless anyone sees any issues with this. |
|
See #63 for the CHANGELOG |
This provides a quick way to get accessed to a merged protodesc.Resolver by building a
*protoregistry.Filesfor all the gogo files and then implementing a resolver on top. The merged registry functionality can then only be used when validation of conflicts is needed.