It's schema definition is weirdly asymmetric, with variants dependent
on another fields' value. Need to decide if we want to either
hand-code a superset or make a more complex decoding, but definitely
not something we'd want the generator to be able to do.
This is likely inherited from bitcoind, and a bit awkward for us, so
we parse it into a classic struct, but serialize it back into the
bitcoind format when talking to the RPC.
We are inferring the field numbers on the fly, which isn't really
compatible with the way GRPC field numbers work, i.e., they must be
stable while the IDL file evolves. So far when a field was added in
the middle of a struct or removed all subsequent fields would get
renumbered, essentially breaking any client that was using the old
scheme.
We now add a meta file `.msggen.json` that keeps track of the numbers
assigned so far, so they can be reused, and new ones can be generated
not to conflict with existing ones. This file is intentionally kept
generic, so other generators can add more information that has to be
managed across runs.
Changelog-None
`listpeers` is a rather deeply nested structure which has a couple of
caveats, namely that we use the same enum multiple times, which causes
naming clashes. So we truncate the state_changes[]. We can later map
them if needed, but it'll get much easier once we have an abstract
model description that isn't JSON schema, which unrolls all types,
causing us to generate those enums multiple times.
There is at least one clash with a built-in for the grpc server trait,
namely `connect` so we add support for renaming a method when
generating the scaffolding
The server doesn't do much more than unwrapping the request from its
grpc envelope, convert it into the matching JSON-RPC binding struct,
initiate the RPC connection (until we have connection pooling), and
then forwards the converted request. The inverse then happens for the
result.
We build an in-memory model of what the API should look like, which
will later be used to generate a variety of bindings. In this PR we
will use the model to build structs corresponding to the requests and
responses for the various methods.
The JSON-RPC schemas serve as ground-truth, however they are missing a
bit of context: methods, and the request-response matching (as well as
a higher level grouping we'll call a Service). I'm tempted to create a
new document that describes this behavior and we could even generate
the rather repetitive JSON schemas from that document. Furthermore
it'd allow us to add some required metadata such as grpc field
numbering once we generate those bindings.
Changelog-Added: JSON-RPC: A new `msggen` library allows easy generation of language bindings for the JSON-RPC from the JSON schemas
joe.miyamoto
Christian Decker