Utility to run a regtest bitcoind process, useful in integration testing environment.
use bitcoincore_rpc::RpcApi;
let exe_path = exe_path().expect("bitcoind executable must be provided in BITCOIND_EXE, or with a feature like '22_0', or be in PATH");
let bitcoind = bitcoind::BitcoinD::new(exe_path).unwrap();
assert_eq!(0, bitcoind.client.get_blockchain_info().unwrap().blocks);
When a feature like 22_0
is selected, the build script will automatically download the bitcoin core version 22.0
, verify the hashes and place it in the build directory for this crate.
Use utility function downloaded_exe_path()
to get the downloaded executable path.
In your project Cargo.toml, activate the following features
[dev-dependencies]
bitcoind = { version = "0.20.0", features = [ "22_0" ] }
Then use it:
let bitcoind = bitcoind::BitcoinD::new(bitcoind::downloaded_exe_path().unwrap()).unwrap();
The MSRV is 1.41.1 for version 0.29.* if no feature is used, otherwise is 1.57
I used integration testing based on external bash script launching needed external processes, there are many issues with this approach like:
- External script may interfere with local development environment https://github.com/rust-bitcoin/rust-bitcoincore-rpc/blob/200fc8247c1896709a673b82a89ca0da5e7aa2ce/integration_test/run.sh#L9
- Use of a single huge test to test everything https://github.com/rust-bitcoin/rust-bitcoincore-rpc/blob/200fc8247c1896709a673b82a89ca0da5e7aa2ce/integration_test/src/main.rs#L122-L203
- If test are separated, a failing test may fail to leave a clean situation, causing other test to fail (because of the initial situation, not a real failure)
- bash script are hard, especially support different OS and versions
- It waits until bitcoind daemon become ready to accept RPC commands
- bitcoind use a temporary directory as datadir. You can specify the root of your temp directories so that you have node's datadir in a RAM disk (eg
/dev/shm
) - Free ports are asked to the OS (a low probability race condition is still possible)
- the process is killed when the struct goes out of scope no matter how the test finishes
- allows easy spawning of dependent process like https://github.com/RCasatta/electrsd
Thanks to these features every #[test]
could easily run isolated with its own environment.