near-runner v1.0.0
near-runner for TypeScript/JavaScript
Write tests once, run them both on NEAR TestNet and a controlled NEAR Sandbox local environment
This software is in early beta and feedback is appreciated.
Quick Start with AVA
near-runner-ava is a thin wrapper around near-runner-js designed to get you up and running as quickly as possible, with minimal configuration and power-boosts like TypeScript. You can install it with one command. You will need NodeJS installed. Then:
npx near-runner-init
This command will:
- Add a
near-runner
directory to the folder where you ran the command. This directory contains all the configuration needed to get you started with near-runner-ava, and a__tests__
subfolder with a well-commented example test file. - Create
test.sh
andtest.bat
scripts in the folder where you ran the command. These can be used to quickly run the tests innear-runner
. Feel free to integrate test-running into your project in a way that makes more sense for you, and then remove these scripts. - Install NPM dependencies using
npm install
. Most of the output you see when running the command comes from this step. You can skip this:npx near-runner-init --no-install
.
If you want to install near-runner-ava manually, see its README.
How It Works
Let's look at some code that focuses on near-runner itself, without any AVA or other testing logic.
Creating a
Runner
.const runner = Runner.create(async ({root}) => { const alice = await root.createAccount('alice'); const contract = await root.createAndDeploy( 'contract-account-name', 'path/to/compiled.wasm' ); return {alice, contract}; });
Let's step through this.
Runner.create
initializes a new NEAR Sandbox node/instance. This is essentially a mini-NEAR blockchain created just for this test. Each of these Sandbox instances gets its own data directory and port, so that tests can run in parallel.root.createAccount
creates a new account with the given name.root.createAndDeploy
creates a new account with the given name, then deploys the specified Wasm file to it.path/to/compiled.wasm
will resolve relative to your project root. That is, the nearest directory with apackage.json
file, or your current working directory if nopackage.json
is found. To construct a path relative to your test file, you can usepath.join(__dirname, '../etc/etc.wasm')
(more info).- After
Runner.create
finishes running the function passed into it, it gracefully shuts down the Sandbox instance it ran in the background. However, it keeps the data directory around. That's what stores the state of the two accounts that were created (alice
andcontract-account-name
with its deployed contract). runner
contains a reference to this data directory, so that multiple tests can use it as a starting point.- The object returned,
{alice, contract}
, will be passed along to subsequent tests.
Writing tests.
near-runner is designed for concurrency (which is why it's a great fit for AVA, which runs tests concurrently by default). Here's a simple way to get concurrent runs using plain JS (for a working example, see near-examples/rust-status-message):
await Promise.all([ runner.run(async ({alice, contract}) => { await alice.call( contract, 'some_update_function', {some_string_argument: 'cool', some_number_argument: 42} ); const result = await contract.view( 'some_view_function', {account_id: alice} ); assert.equal(result, 'whatever'); }), runner.run(async ({alice, contract}) => { const result = await contract.view( 'some_view_function', {account_id: alice} ); assert.equal(result, 'some default'); }); ]);
Let's step through this.
- Like the earlier call to
Runner.create
, each call torunner.run
sets up its own Sandbox instance. Each will copy the data directory set up earlier as the starting point for its tests. Each will use a unique port so that tests can be safely run in parallel. call
syntax mirrors near-cli and either returns the successful return value of the given function or throws the encountered error. If you want to inspect a full transaction and/or avoid thethrow
behavior, you can usecall_raw
instead.- While
call
is invoked on the account doing the call (alice.call(contract, …)
),view
is invoked on the account being viewed (contract.view(…)
). This is because the caller of a view is irrelevant and ignored. - Gotcha: the full account names may or may not match the strings passed to
createAccount
andcreateAndDeploy
, which is why you must writealice.call(contract, …)
rather thanalice.call('contract-account-name', …)
. But! TheAccount
class overridestoJSON
so that you can pass{account_id: alice}
in arguments rather than{account_id: alice.accountId}
. If you need the generated account ID in some other circumstance, remember to usealice.accountId
.
- Like the earlier call to
See the __tests__ directory in this project for more examples.
"Spooning" Contracts from Testnet and Mainnet
Spooning a blockchain is copying the data from one network into a different network. near-runner makes it easy to copy data from Mainnet or Testnet contracts into your local Sandbox environment:
await runner.run(async ({root}) => {
const refFinance = await root.createAccountFrom({
mainnetContract: 'v2.ref-finance.near',
blockId: 50_000_000,
withData: true,
});
});
This would copy the Wasm bytes and contract state from v2.ref-finance.near to your local blockchain as it existed at block 50_000_000
. This makes use of Sandbox's special patch state feature to keep the contract name the same, even though the top level account might not exist locally (note that this means it only works in Sandbox testing mode). You can then interact with the contract in a deterministic way the same way you interact with all other accounts created with near-runner.
Gotcha: withData
will only work out-of-the-box if the contract's data is 50kB or less. This is due to the default configuration of RPC servers; see the "Heads Up" note here. Some teams at NEAR are hard at work giving you an easy way to run your own RPC server, at which point you can point tests at your custom RPC endpoint and get around the 50kB limit.
See an example of spooning contracts at tests/05.spoon-contract-to-sandbox.ava.ts.
Running on Testnet
near-runner is set up so that you can write tests once and run them against a local Sandbox node (the default behavior) or against NEAR TestNet. Some reasons this might be helpful:
- Gives higher confidence that your contracts work as expected
- You can test against deployed testnet contracts
- If something seems off in Sandbox mode, you can compare it to testnet
- Until we have a full-featured dev environment that includes Explorer, Wallet, etc, you can write full end-to-end tests using a tool like Cypress
You can run in testnet mode in two ways.
When creating your Runner, pass a config object as the first argument:
const runner = Runner.create( {network: 'testnet'}, async ({root}) => { … } )
Set the
NEAR_RUNNER_NETWORK
environment variable when running your tests:NEAR_RUNNER_NETWORK=testnet node test.js
If you set both, the config object takes precedence.
Stepping through a testnet example
Let's revisit a shortened version of the example from How It Works above, describing what will happen in Testnet.
Create a
Runner
.const runner = Runner.create(async ({root}) => { await root.createAccount('alice'); await root.createAndDeploy( 'contract-account-name', 'path/to/compiled.wasm' ); });
Let's step through this.
Runner.create
does not interact with Testnet at all yet. Instead, the function runs at the beginning of each subsequent call torunner.run
. This matches the semantics of the sandbox that all subsequent calls torun
have the same starting point, however, testnet requires that each call have its own account. In factRunner.create
creates a unique testnet account and each test is a unique sub-account.
Write tests.
await Promise.all([ runner.run(async ({alice, contract}) => { await alice.call( contract, 'some_update_function', {some_string_argument: 'cool', some_number_argument: 42} ); const result = await contract.view( 'some_view_function', {account_id: alice} ); assert.equal(result, 'whatever'); }), runner.run(async ({alice, contract}) => { const result = await contract.view( 'some_view_function', {account_id: alice} ); assert.equal(result, 'some default'); }); ]);
Each call to
runner.run
will:- Get or create its own sub-account on testnet account, e.g.
t.rdsq0289478
. If creating the account the keys will be stored at$PWD/.near-credentials/runner/testnet/t.rdsq0289478.json
. - Run the
initFn
passed toRunner.create
- Create sub-accounts for each
createAccount
andcreateAndDeploy
, such asalice.t.rdsq0289478
- If the test account runs out of funds to create accounts it will request a transfer from the root account.
- After the test is finished each account created is deleted and the funds sent back to the test account.
- Get or create its own sub-account on testnet account, e.g.
Note: Since the testnet accounts are cached, if account creation rate limits are reached simply wait a little while and try again.
Skipping Sandbox-specific tests
If some of your runs take advantage of Sandbox-specific features, you can skip these on testnet in a couple ways:
runSandbox
: Instead ofrunner.run
, you can userunner.runSandbox
:await Promise.all([ runner.run(async ({…}) => { // runs on any network, sandbox or testnet }), runner.runSandbox(async ({…}) => { // only runs on sandbox network }); ]);
Runner.networkIsSandbox
: You can also skip entire sections of your files by checkingRunner.networkIsSandbox
(Runner.networkIsTestnet
andRunner.getNetworkFromEnv
are also available).let runner = Runner.create(async ({root}) => ({ // note the implicit return contract: await root.createAndDeploy( 'contract-account-name', 'path/to/compiled.wasm' ) })); runner.run('thing that makes sense on any network', async ({…}) => { // logic using basic contract & account interactions }); if (Runner.networkIsSandbox) { runner.run('thing that only makes sense with sandbox', async ({…}) => { // logic using patch-state, fast-forwarding, etc }); }
Patch State on the Fly
In Sandbox-mode tests, you can add or modify any contract state, contract code, account or access key with patchState
.
You cannot perform arbitrary mutation on contract state with transactions since transactions can only include contract calls that mutate state in a contract-programmed way. For example, with an NFT contract, you can perform some operation with NFTs you have ownership of, but you cannot manipulate NFTs that are owned by other accounts since the smart contract is coded with checks to reject that. This is the expected behavior of the NFT contract. However, you may want to change another person's NFT for a test setup. This is called "arbitrary mutation on contract state" and can be done with patchState
. Alternatively you can stop the node, dump state at genesis, edit genesis, and restart the node. The later approach is more complicated to do and also cannot be performed without restarting the node.
It is true that you can alter contract code, accounts, and access keys using normal transactions via the DeployContract
, CreateAccount
, and AddKey
actions. But this limits you to altering your own account or sub-account. patchState
allows you to perform these operations on any account.
To see an example of how to do this, see the patch-state test.
near-runner will support expanded patchState-based functionality in the future:
- Allow bootstrapping sandbox environment from testnet/mainnet contracts & state
- Allow replaying all transactions from testnet/mainnet
- Support time-travel / fast-forwarding
Pro Tips
NEAR_RUNNER_DEBUG=true
– run tests with this environment variable set to get copious debug output and a full log file for each Sandbox instance.Runner.create
config – you can pass a config object as the first argument toRunner.create
. This lets you do things like:skip initialization if specified data directory already exists
Runner.create( { init: false, homeDir: './test-data/alice-owns-an-nft' }, async ({root}) => { … } )
always recreate such data directory instead (the default behavior)
specify which port to run on
and more!
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago