This directory types contains custom proto messages and corresponding go types that make working with null values in a database easier.

In addition, the utility protoc-go-tags can be used to specify additional go struct tags in the proto definitions. This can be used to specify database column names or validation options. It is also possible to replace existing tags, this can be used specifically to overwrite json tags.

Please note that we actually parse the go code and modify the AST (abstract syntax tree) to add your additional go tags. This should be somewhat more reliable than regular expressions alone.

Struct tags must be placed on a separate comment line, but can contain multiple tags.

// proto definitions
message StarfleetShip {
	string name = 1;

	// use a different db column name for the departure time
	// `db:"we_are_leaving_at"`
	dkfbasel.protobuf.Timestamp departure_time = 4;
}

// go struct definition
type StarfleetShip struct {
	Name string `protobuf:"bytes,1,opt,name=name" json:"name,omitempty" `

	// use a different db column name for the departure time
	// `db:"we_are_leaving_at"`
	DepartureTime *dkfbasel_protobuf.Timestamp `protobuf:"bytes,4,opt,name=departure_time,json=departureTime" json:"departure_time,omitempty" db:"we_are_leaving_at"`
}

To add the additional tags to your structs proceed as following - a complete example can be found in the directory example.

1. Add the tag as comment above the respective property. Unfortunately the tag can currently not be placed on the same line, as the protoc grpc plugin will strip out all comments that are on the same line.

2. Run the protoc generator with the grpc plugin.

3. Run the protoc-go-tags utility on the generated files. You need to specify the directory that the utility should process and it will recursively scan all files and transform all go files with the ending .pb.go

4. The additional tags should now be included in your .pb.go files

We are often using sqlx to simplify database access for our programs and found that the matching of named parameters to struct properties would use lowercase name transforms and thus not work with our snakecased column names.

It is however trivial to change the name mapping by simply passing a custom mapper function to the slqx db connection. An example of this can be found in example/main.go.