Frugal is an extension of Apache Thrift which provides additional functionality. Key features include:
- request headers
- request multiplexing
- request interceptors
- per-request timeouts
- thread-safe clients
- code-generated pub/sub APIs
- support for Go, Java, Dart, and Python (2.7 and 3.5)
Frugal is intended to act as a superset of Thrift, meaning it implements the same functionality as Thrift with some additional features. For a more detailed explanation, see the documentation.
brew install frugal
Pre-compiled binaries for OS X and Linux are available from the Github releases tab. Currently, adding these binaries is a manual process. If a downloadable release is missing, notify the messaging team to have it added.
If go is already installed and setup you can also simply:
$ go get github.com/Workiva/frugal
Our usage of godep has been deprecated as we move to glide. Once the deprecation period is over, we will remove both the Godeps/ and vendor/ folder, relying solely on glide for dependency management
-
Clone the frugal repo
$ mkdir -p $GOPATH/src/github.com/Workiva && cd $_ $ git clone [email protected]:Workiva/frugal.git
-
Install the CLI binary
$ cd $GOPATH/src/github.com/Workiva/frugal $ curl https://glide.sh/get | sh # get glide if necessary $ glide install # get dependencies $ go install
When generating go, be aware the frugal go library and the frugal compiler have separate dependencies.
Define your Frugal file which contains your pub/sub interface, or scopes, and Thrift definitions.
# event.frugal
// Anything allowed in a .thrift file is allowed in a .frugal file.
struct Event {
1: i64 ID,
2: string Message
}
// Scopes are a Frugal extension for pub/sub APIs.
scope Events {
EventCreated: Event
}
Generate the code with frugal
. Currently, only Go, Java, Dart, and Python are
supported.
$ frugal -gen=go event.frugal
By default, generated code is placed in a gen-*
directory. This code can then
be used as such:
// publisher.go
func main() {
conn, err := nats.Connect(nats.DefaultURL)
if err != nil {
panic(err)
}
var (
protocolFactory = frugal.NewFProtocolFactory(thrift.NewTBinaryProtocolFactoryDefault())
transportFactory = frugal.NewFNatsScopeTransportFactory(conn)
provider = frugal.NewFScopeProvider(transportFactory, protocolFactory)
publisher = event.NewEventsPublisher(provider)
)
publisher.Open()
defer publisher.Close()
event := &event.Event{ID: 42, Message: "Hello, World!"}
if err := publisher.PublishEventCreated(frugal.NewFContext(""), event); err != nil {
panic(err)
}
}
// subscriber.go
func main() {
conn, err := nats.Connect(nats.DefaultURL)
if err != nil {
panic(err)
}
var (
protocolFactory = frugal.NewFProtocolFactory(thrift.NewTBinaryProtocolFactoryDefault())
transportFactory = frugal.NewFNatsScopeTransportFactory(conn)
provider = frugal.NewFScopeProvider(transportFactory, protocolFactory)
subscriber = event.NewEventsSubscriber(provider)
)
_, err = subscriber.SubscribeEventCreated(func(ctx *frugal.FContext, e *event.Event) {
fmt.Println("Received event:", e.Message)
})
if err != nil {
panic(err)
}
wait := make(chan bool)
log.Println("Subscriber started...")
<-wait
}
By default, Frugal publishes messages on the topic <scope>.<operation>
. For
example, the EventCreated
operation in the following Frugal definition would
be published on Events.EventCreated
:
scope Events {
EventCreated: Event
}
Custom topic prefixes can be defined on a per-scope basis:
scope Events prefix foo.bar {
EventCreated: Event
}
As a result, EventCreated
would be published on foo.bar.Events.EventCreated
.
Prefixes can also define variables which are provided at publish and subscribe time:
scope Events prefix foo.{user} {
EventCreated: Event
}
This variable is then passed to publish and subscribe calls:
var (
event = &event.Event{ID: 42, Message: "hello, world!"}
user = "bill"
)
publisher.PublishEventCreated(frugal.NewFContext(""), event, user)
subscriber.SubscribeEventCreated(user, func(ctx *frugal.FContext, e *event.Event) {
fmt.Printf("Received event for %s: %s\n", user, e.Message)
})
In Thrift, comments of the form /** ... */
are included in generated code. In
Frugal, to include comments in generated code, they should be of the form /**@ ... */
.
/**@
* This comment is included in the generated code because
* it has the @ sign.
*/
struct Foo {}
/**@ This comment is included too. */
service FooService {
/** This comment isn't included because it doesn't have the @ sign. */
Foo getFoo()
}
Annotations are extra directive in the IDL that can alter the way code is generated. Some common annotations are listed below
Annotation | Values | Allowed Places | Description |
---|---|---|---|
vendor | Optional location | Namespaces, Includes | See vendoring includes |
deprecated | Optional description | Service methods, Struct/union/exception fields | Marks a method or field as deprecated (if supported by the language, or in a comment otherwise), and logs a warning if a deprecated method is called. |
Frugal does not generate code for includes by default. The -r
flag is
required to recursively generate includes. If -r
is set, Frugal generates the
entire IDL tree, including code for includes, in the same output directory (as
specified by -out
) by default. Since this can cause problems when using a
library that uses a Frugal-generated object generated with the same IDL in two
or more places, Frugal provides special support for vendoring dependencies
through a vendor
annotation on includes and namespaces.
The vendor
annotation is used on namespace definitions to indicate to any
consumers of the IDL where the generated code is vendored so that consumers can
generate code that points to it. This cannot be used with *
namespaces since
it is language-dependent. Consumers then use the vendor
annotation on
includes they wish to vendor. The value provided on the include-side vendor
annotation, if any, is ignored.
When an include is annotated with vendor
, Frugal will skip generating the
include if use_vendor
language option is set since this flag indicates
intention to use the vendored code as advertised by the vendor
annotation.
If no location is specified by the vendor
annotation, the behavior is defined
by the language generator.
The vendor
annotation is currently only supported by Go, Dart and Java.
The example below illustrates how this works.
bar.frugal ("providing" IDL):
namespace go bar (vendor="github.com/Workiva/my-repo/gen-go/bar")
namespace dart bar (vendor="my-repo/gen-go")
namespace java bar (vendor="com.workiva.bar.custom.pkg")
struct Struct {}
foo.frugal ("consuming" IDL):
include "bar.frugal" (vendor)
service MyService {
bar.Struct getStruct()
}
frugal -r -gen go:package_prefix=github.com/Workiva/my-other-repo/gen-go,use_vendor foo.frugal
When we run the above command to generate foo.frugal
, Frugal will not
generate code for bar.frugal
since use_vendor
is set and the "providing"
IDL has a vendor path set for the Go namespace. Instead, the generated code for
foo.frugal
will reference the vendor path specified in bar.frugal
(github.com/Workiva/my-repo/gen-go/bar).
Frugal is intended to be a superset of Thrift, meaning valid Thrift should be valid Frugal. File an issue if you discover an inconsistency in compatibility with the IDL.