[feature]: functional approach to compile data-model artifacts

Old approach

Iroha JavaScript uses the original Iroha repository for a set of reasons:

Run kagami in order to generate data-model schema
Use iroha_data_model crate in order to generate encoded samples for data-model tests
Run iroha binary in end-2-end tests

Previously, iroha_data_model was referenced in packages/data-model-rust-samples/Cargo.toml as:

iroha_data_model = { git = "<git repo>", rev = "<rev>" }

kagami and iroha binaries were built via cargo build:

cargo build --git "<git repo>" --rev "<rev>" --package kagami

This pipeline was encapsulated into @iroha2/dev-iroha-bins package and was semi-automatic - compiled artifacts were produced and committed manually as described at "Manually update reference iroha version" guide.

New approach

Automatic updates of the reference iroha in a single place

@iroha2/iroha-source dev packages comes in instead of @iroha2/dev-iroha-bins. In order to change reference Iroha version, it is only needed to update packages/iroha-source/config.json, and everything else will be re-compiled on the fly.

The configuration schema is pretty straightforward:

{
  "origin": "https://github.com/hyperledger/iroha.git",
  "rev": "52dc18cd81bdc1d1906ffeecb666dd9b2eb27955"
}

These parameters are used to git clone the repo. Thus, origin field might be even a file-system path, and rev - a branch/tag name.

Then, each time when some actor needs to use iroha or kagami binary, the following steps are performed:

import { resolveBinary } from '@iroha2/iroha-source'

const irohaPath = (await resolveBinary('iroha')).path
// 1. If the repo is not cloned, clone it. If the cloned repo is not the same as specified
//    in `config.json`, update it.
// 2. If there is no `target/release/iroha`, run `cargo build --release --package iroha`
// 3. Return path to the `target/release/iroha` binary

Also, data-model-rust-samples crate now references iroha_data_model via path option in Cargo.toml, which:

Removes the need to manually duplicate Iroha git reference as it was before;
Leverages compilation cache in the cloned repo.

Exclude data-model artifacts from Git tracking

Those artifacts aren't tracked by Git anymore:

packages/data-model-schema/src/__schema__.json
packages/data-model-rust-samples/samples.json
packages/data-model/src/__generated__.ts

Their generation now works fast, and it is run at following steps:

Monorepo postinstall script
Before tests and build

Refactored monorepo tasks

I made some changes in the root Jakefile and package.json. Notable ones:

At the postinstall step, data-model artifacts are compiled
pnpm publish-all now runs all necessary steps, including tests and build.

In conclusion, the overall approach is easier to use, maintain and is more robust, because now it is harder to "forget" update data-model schema - it updates automatically before each step that needs it, and it runs fast in dev environment.

However, CI/CD might run longer now, because it will compile 3 Rust crates, but this process is optimized thanks to incremental compilation.

hyperledger / iroha-javascript