New Features

This release includes the following:

• #71
• #82

Audit trail for GetChanges and GetFiles calls

The SDK library now provides a mechanism to keep log fields during the message exchange between lookoutd and your analyzer.

An example of the logs fields that are kept in the message exchange would be:

{
    "repository": "github.com/src-d/lookout-sdk",
    "github.pr": 17
}

In practice this means that your analyzer will be able to easily add relevant information to its logs, and lookoutd will have a context for each call you make. A sort of audit trail for each GetChanges or GetFiles call.

This mechanism is implemented differently in the Go SDK and in the Python SDK

Go SDK

The above mentioned mechanism is implemented by using gRPC interceptors:

• pb.CtxlogUnaryClientInterceptor,
• pb.CtxlogStreamClientInterceptor,
• pb.CtxlogUnaryServerInterceptor,
• pb.CtxlogStreamServerInterceptor.

These interceptors are added automatically using pb.DialContext and pb.NewServer. There are also pb.DialContextWithInterceptors and pb.NewServerWithInterceptors respectively in case you want to add other interceptors. See examples/language-analyzer.go.

In your analyzer you can then use the pb.GetLogFields and pb.AddLogFields to read and add log fields to the context. The log fields that you add will be available to lookoutd.

Python SDK

The above mentioned mechanism is implemented by passing a wrapped context to each analyzer call (Notify{Review,Push}Event), and by changing the base class of your analyzer:

• in version v0.5.0: lookout.sdk.pb.AnalyzerServicer -> lookout.sdk.service_analyzer_pb2_grpc.AnalyzerServicer,
  in version v0.6.1: lookout.sdk.pb.AnalyzerServicer -> lookout.sdk.service_analyzer.AnalyzerServicer.

The wrapped context is a proxy to the gRPC one. It additionally provides functionalities for adding and reading log fields that will be available to lookoutd, once you "activate" the audit trail (see later). See lookout.sdk.grpc.log_fields.LogFieldsContext.

This change is not a breaking change. Without updating anything the code is expected to work as before the update. In order to "activate" the audit trail you have to follow some steps. See the migration guide.

Logger interceptors

The SDK library now provides gRPC interceptors both for Go SDK and Python SDK for logging client calls and server request. See examples/language-analyzer.go and examples/language-analyzer.py to see how to use them.

Python SDK

An extra note must be added for the Python SDK. At time of writing gRPC for Python doesn't provide an api to add server interceptors for diffenret method types, so an api on top of that has been added to provide the user the capability to also add custom server interceptors. See lookout.sdk.grpc.interceptors.base.

[Python SDK] Default message size for gRPC server

A new create_server function creates a server with default values for gRPC max send and max receive message length.

Migration Guide: v0.5.0 => v0.6.1

You might want to check out the changes made to examples/language-analyzer.go and examples/language-analyzer.py in the comparison view.

Go SDK

Except for using pb.DialContext and pb.NewServer or their corresponding pb.DialContextWithInterceptors and pb.NewServerWithInterceptors, you don't have anything else to do in order to activate the audit trail.

Python SDK

Follow these steps to activate the audit trail.

1. Rename the following:

• NotifyReviewEvent -> notify_review_event,
• NotifyPushEvent -> notify_push_event.

2. Use create_server.
3. Use DataStub imported from lookout.sdk.service_data and not from pb. The returned stub has a more pythonic api. You have to rename the following calls:
   • GetChanges -> get_changes,
   • GetFiles -> get_files.
     Both methods needs as first positional argument the context.