Since JSON handling has moved to json_serialization and web3 now has a test suite of its own, the eth tests here no longer add meaningful coverage. Some of the refactoring post-0.5.4 has also opened up opportunities to simplify the code - better get that done before a release ;) * move pending requests to `RpcConnection` since they're only used for bidirectional transports * fast-path for message processing that avoids async overhead * fix websocket onDisconnect reset * remove `ResponseBatchRx` - the response type (single or batch) must match the request type, in order * fix batch notifications getting repsonses when they shouldn't * add some async waits to tests to check response ordering is maintained * simplify client further |
||
|---|---|---|
| .github/workflows | ||
| json_rpc | ||
| tests | ||
| .gitignore | ||
| config.nims | ||
| json_rpc.nimble | ||
| LICENSE | ||
| nim.cfg | ||
| README.md | ||
json-rpc
json_rpc is a library designed to provide an easy interface for working with remote procedure calls using the JSON RPC standard.
Usage
# In your nimble file
requires "json_rpc"
Introduction
json_rpc is a library for routing JSON 2.0 format remote procedure calls over different transports.
It is designed to automatically generate marshalling and parameter checking code based on the RPC parameter types.
Routing
Remote procedure calls are created using the rpc macro on an instance of RpcRouter.
rpc allows you to provide a list of native Nim type parameters and a return type, generates marshalling to and from json for you, so you can concentrate on using native Nim types for your call.
Routing is then performed by the route procedure.
When an error occurs, the error is populated, otherwise result will be populated.
rpc Parameters
path: The string to match for the method.
body: The parameters and code to execute for this call.
Example
Here's a simple example:
import json_rpc/rpcserver
var router = RpcRouter.init()
router.rpc("hello") do():
%"Hello"
As no return type was specified in this example, result defaults to the JsonNode type from std/json.
Parameters are defined with the do notation, allowing any Nim types to be used.
Here's how to pass a string to an RPC and return a string:
# Specify how strings are converted to JSON
JrpcConv.automaticSerialization(string, true)
# use `do` for the body of the function
router.rpc("hello") do(input: string) -> string:
"Hello " & input
The conversion to and from JSON is done using the JsonConv flavor of nim-json-serialization. For each type used in the API, a serialization declaration tells json_rpc how to convert it to JSON, either using defaults or by overriding readValue and writeValue.
json_rpc will recursively parse the Nim types in order to produce marshalling code.
This marshalling code uses the types to check the incoming JSON fields to ensure they exist and are of the correct kind.
The return type then performs the opposite process, converting Nim types to Json for transport.
Here is a more complex parameter example:
import
json_rpc/rpcserver,
json_rpc/jsonmarshal
type
HeaderKind = enum hkOne, hkTwo, hkThree
Header = object
kind: HeaderKind
size: int64
DataBlob = object
items: seq[byte]
headers: array[3, Header]
MyObject = object
data: DataBlob
name: string
Header.useDefaultSerializationIn JrpcConv
DataBlob.useDefaultSerializationIn JrpcConv
MyObject.useDefaultSerializationIn JrpcConv
router.rpc("updateData") do(myObj: MyObject, newData {.serializedFieldName: "new".}: DataBlob) -> DataBlob:
if myObj.name == "old":
myObj.data
else:
newData
Behind the scenes, all RPC calls take parameters through RequestParamsRx structure.
At runtime, the json is checked to ensure that it contains the correct number and type of your parameters to match the rpc definition.
When named parameters are used, serializedFieldName can be used to customize the name.
Compiling with -d:nimDumpRpcs will show the output code for the RPC call. To see the output of the async generation, add -d:nimDumpAsync.
Special type : Option[T]
Option[T] is a special type indicating that parameter may have value or not.
- If optional parameters located in the middle of parameters list, you set it to
nullto tell the server that it has no value. - If optional parameters located at the end of parameter list and there are no more mandatory parameters after that, those optional parameters can be omitted altogether.
# d can be omitted, b should use null to indicate it has no value
router.rpc("updateData") do(a: int, b: Option[int], c: string, d: Option[T]):
if b.isSome:
# do something
else:
# do something else
- If Option[T] used as return type, it also denotes the returned value might not available.
router.rpc("getData") do(name: string) -> Option[int]:
if name == "monkey":
result = some(4)
- If Option[T] used as field type of an object, it also tell us that field might be present or not, and the rpc mechanism will automatically set it to some value if it available.
type
MyOptional = object
maybeInt: Option[int]
Marshalling
Note that array parameters are explicitly checked for length, and will return an error node if the length differs from their declaration size.
If you wish to support custom types in a particular way, you can provide matching readValue and writeValue procedures.
The custom serializer you write must be using JrpcConv flavor.
readValue
This takes a Json type and returns the Nim type.
Parameters
r: var JsonReader[JrpcConv]: The current JsonReader with JrpcConv flavor.
val: var MyInt: Deserialized value.
Example
proc readValue*(r: var JsonReader[JrpcConv], val: var MyInt)
{.gcsafe, raises: [IOError, JsonReaderError].} =
let intVal = r.parseInt(int)
val = MyInt(intVal)
writeValue
This is the standard way to provide translations from a Nim type to Json.
Parameters
w: var JsonWriter[JrpcConv]: The current JsonWriter with JrpcConv flavor.
val: MyInt: The value you want to convert into Json.
Example
proc writeValue*(w: var JsonWriter[JrpcConv], val: MyInt)
{.gcsafe, raises: [IOError].} =
w.writeValue val.int
JSON Format
The router expects either a Json document with the following structure:
{
"jsonrpc": "2.0",
"method": String,
"params": Array or Object,
"id": Int or String
}
If params is an Array, it is a positional parameters. If it is an Object then the rpc method will be called using named parameters.
Return values use the following node structure:
{
"jsonrpc": "2.0",
"result": Json document,
"error": Json document,
"id": Int Or String
}
Performing a route
To call and RPC through the router, use the route procedure.
There are three variants of route.
Note that once invoked all RPC calls are error trapped and any exceptions raised are passed back with the error message encoded as a Json document.
route by string
This route variant will handle all the conversion of string to Json document and check the format and type of the input data.
Parameters
router: RpcRouter: The router object that contains the RPCs.
data: string: A string ready to be processed into a Json document.
Returns
Future[string]: This will be the stringified JSON response, which can be the JSON RPC result or a JSON wrapped error.
route by Json document
This variant allows simplified processing if you already have a Json document. However if the required fields are not present within data, exceptions will be raised.
Parameters
router: RpcRouter: The router object that contains the RPCs.
req: RequestTx: A pre-processed Json document that matches the expected format as defined above.
Returns
Future[ResponseTx]: The JSON RPC result or a JSON wrapped error.
tryRoute
This route variant allows you to invoke a call if possible, without raising an exception.
Parameters
router: RpcRouter: The router object that contains the RPCs.
data: JsonString: A raw Json document that matches the expected format as defined above.
fut: var Future[JsonString]: The stringified JSON RPC result or a JSON wrapped error.
Returns
Result[void, string] isOk if the method field provided in data matches an available route. Returns isErr when the method cannot be found, or if method or params field cannot be found within data.
To see the result of a call, we need to provide Json in the expected format. Here's an example of how that looks by manually creating the JSON. Later we will see the helper utilities that make this easier.
let call = %*{
"jsonrpc": %"2.0",
"method": %"hello",
"params": %["Terry"],
"id": %1
}
# route the call we defined earlier
let localResult = waitFor router.route(call)
echo localResult
# We should see something like this
# {"jsonrpc":"2.0","result":"Hello Terry","id":1}
Transports
The following transports are available:
- HTTP POST - unidirectional, one request/response pair per call
- Sockets and pipes, via chronos' StreamTransport - bidirectional, persistent connection, custom message framing
Framing.httpHeader-Content-Lengthprefix specifying the length of the payload, compatible with vscode-jsonrpcFraming.lengthHeaderBE32- Big-endian, 32-bit binary prefix - most efficient optionFraming.newLine- "\r\n" suffix
- Websockets - bidirectional, persistent connection
Server
In order to make routing useful, RPCs must be invoked and transmitted over a transport.
The RpcServer type is given as a simple inheritable wrapper/container that simplifies designing your own transport layers using the router field.
Transport specific server need only call the route procedure using a string fetched from the transport in order to invoke the requested RPC.
Server example
This example uses the socket transport defined in socket.nim.
Once executed, the "hello" RPC will be available to a socket based client.
import json_rpc/rpcserver
# Create a socket server for transport
var srv = newRpcSocketServer("localhost", Port(8585))
# Specify how strings are converted to JSON
JrpcConv.automaticSerialization(string, true)
# srv.rpc is a shortcut for srv.router.rpc
srv.rpc("hello") do(input: string) -> string:
result = "Hello " & input
srv.start()
runForever()
Client
json_rpc also comes with a client implementation, built to provide a framework for transports to work with.
To simplify demonstration, we will use the socket transport defined in socketclient.nim.
Below is the most basic way to use a remote call on the client. Here we manually supply the name and json parameters for the call.
The call procedure takes care of the basic format of the JSON to send to the server.
However you still need to provide params as a JsonNode or RequestParamsTx, which must exactly match the parameters defined in the equivalent rpc definition.
import json_rpc/[rpcclient, rpcserver], chronos, json
var
server = newRpcSocketServer("localhost", Port(8545))
client = newRpcSocketClient()
server.start
server.rpc("hello") do(input: string) -> string:
"Hello " & input
waitFor client.connect("localhost", Port(8545))
let response = waitFor client.call("hello", %[%"Daisy"])
# the call returns a `JsonString` type which contains the result
echo response
createRpcSigs
To make things more readable and allow better static checking client side, json_rpc supports generating wrappers for client RPCs using createRpcSigs.
This macro takes a type name and the path of a file containing forward declarations of procedures that you wish to convert to client RPCs. The transformation generates procedures that match the forward declarations provided, plus a client parameter in the specified type.
Because the signatures are parsed at compile time, the file will be error checked and you can use import to share common types between your client and server.
Parameters
clientType: This is the type you want to pass to your generated calls. Usually this would be a transport specific descendant from RpcClient.
path: The path to the Nim module that contains the RPC header signatures.
Variants of createRpcSigs
createRpcSigsFromString, generate rpc wrapper from string instead load it from file.createSingleRpcSig, generate rpc wrapper from single Nim proc signature, with alias. e.g. calling same rpc method using different return type.createRpcSigsFromNim, generate rpc wrapper from a list Nim proc signature, without loading any file.
Example
For example, to support this remote call:
server.rpc("bmi") do(height, weight: float) -> float:
result = (height * height) / weight
You can have the following in your rpc signature file:
proc bmi(height, weight: float): float
When parsed through createRpcSigs, you can call the RPC as if it were a normal procedure.
So instead of this:
let bmiIndex = await client.call("bmi", %[%120.5, %12.0])
You can use:
let bmiIndex = await client.bmi(120.5, 12.0)
Or you can use batch call to send multiple request at once to the server.
let batch = client.prepareBatch()
batch.bmi(120.5, 12.0)
batch.bmi(120.5, 13.0)
batch.bmi(120.5, 14.0)
let res = await batch.send()
# But you need to manually process the response e.g. decode from JSON to
# your expected type because you can mix various rpc method call in one batch
# with various return type.
This allows you to leverage Nim's static type checking whilst also aiding readability and providing a unified location to declare client side RPC definitions.
Working with client transports
Transport clients should provide a type that is inherited from RpcClient where they can store any transport related information.
Additionally, the following two procedures are useful:
-
Callself: a descendant ofRpcClientname: string: the method to be calledparams: JsonNode: The parameters to the RPC call ReturningFuture[JsonString]: A wrapper for the resultJson documentand a flag to indicate if this contains an error.
Note: Although call isn't necessary for a client to function, it allows RPC signatures to be used by the createRpcSigs.
-
Connectclient: a descendant ofRpcClientReturningFutureBase: The base future returned when a procedure is annoted with{.async.}
processMessage
To simplify and unify processing within the client, the processMessage procedure can be used to perform conversion and error checking from the received string originating from the transport to the Json document representation that is passed to the RPC.
After a RPC returns, this procedure then completes the futures set by call invocations using the id field of the processed Json document from line.
Parameters
self: a client type descended from RpcClient
line: string: a string that contains the JSON to be processed
Contributing
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Please make sure to update tests as appropriate.
License
Licensed and distributed under either of
- MIT license: LICENSE-MIT or http://opensource.org/licenses/MIT
or
- Apache License, Version 2.0, (LICENSE-APACHEv2 or http://www.apache.org/licenses/LICENSE-2.0)
at your option. This file may not be copied, modified, or distributed except according to those terms.