The fastest and recommended way to develop with near-sdk-js is to create a project with our github template: https://github.com/near/near-sdk-js-template-project.
There are a couple of contract examples in the project:
- Clean contract state
- Doing cross contract call
- Fungible token
- Lockable fungible token
- Non fungible token
- Status message board
The general steps to run these contracts are same. You can also follow their corresponding READMEs to build, test and run the contracts.
- Use near-cli to deploy
jsvm.wasm
from theres
folder to one of account you controlled. For example,jsvm.<your-account>
:
export NEAR_ENV=local
near deploy <jsvm-account> res/jsvm.wasm
cd examples/<example>
yarn && yarn build
to get .base64 file (JS smart-contract).- Deploy .base64 file to
JSVM
account from the previous step.
near js deploy --accountId <your-account> --base64File build/<contract-name>.base64 --deposit 0.1 --jsvm <jsvm-account>
- Interact with your contract using NEAR CLI or
near-api-js
. Encode the parameters and call. If the call cause the state increasement, you also need to attach NEAR to cover the storage deposit for the delta.
near js call <account-that-deployed-js-contract-to-jsvm> <method-name> --accountId <account-performing-call> --args <args> --deposit 0.1 --jsvm <jsvm-account>
- If you want to remove the js contract and withdraw the storage deposit, use:
near js remove --accountId <your-account> --jsvm <jsvm-account>
export NEAR_ENV=testnet
cd examples/<example>
yarn && yarn build
to get .base64 file (JS smart-contract).- Deploy, call and remove JS contract is same as above, except is
jsvm.testnet
. This is also the default value, so you omit--jsvm
.
If you want to indicate an error happened and fail the transaction, just throw an error object in JavaScript. Our JSVM runtime will detect and automatically invoke panic_utf8
with "{error.message}\n:{error.stack}"
. As a result, transaction will fail with "Smart contract panicked: {error.message}\n{error.stack}"
error message. You can also use an error utilities library to organize your errors, such as verror.
When your JS code or library throws an error, uncaught, the transaction will also fail with GuestPanic error, with the error message and stacktrace.
When call host function with inappropriate type, means incorrect number of arguments or arg is not expected type: - if arguments less than params, remaining argument are set as 'undefined' - if arguments more than params, remaining argument are ignored - if argument is different than the required type, it'll be coerced to required type - if argument is different than the required type but cannot be coerced, will throw runtime type error, also with message and stacktrace
We recommend to use near-workspaces to write tests for your smart contracts. See any of the examples for how tests are setup and written.
All NEAR blockchain provided functionality (host functions) are defined in src/api.js
and exported as near
. You can use them by:
import {near} from 'near-sdk-js'
// near.<api doucmented below>. e.g.:
let signer = near.signerAccountId()
To use nightly host functions, such as altBn128G1Sum
, the enclave contract need to be built with make jsvm-nightly
and deployed to a nearcore node that has nightly enabled.
- In arguments,
Uint64: Number | BigInt
. In return,Uint64: BigInt
. Because JavaScript Number cannot hold all Uint64 without losing precision. But as arguments, interger number is also allowed for convinience. Same forUint128
. Bytes
in both arguments and return represent a byte buffer, internally it's a JavaScript String Object. Any binary data0x00-0xff
is stored as the char '\x00-\xff'. This is because QuickJS doesn't have ArrayBuffer in C API. If the bytes happens to have only 1-byte chars, it happens to be same as the the same content string.
function signerAccountId(): String;
function signerAccountPk(): String;
function predecessorAccountId(): String;
function blockIndex(): Uint64;
function blockHeight(): Uint64;
function blockTimestamp(): Uint64;
function epochHeight(): Uint64;
function attachedDeposit(): Uint128;
function prepaidGas(): Uint64;
function usedGas(): Uint64;
function randomSeed(): Bytes;
function sha256(value: Bytes): Bytes;
function keccak256(value: Bytes): Bytes;
function keccak512(value: Bytes): Bytes;
function ripemd160(value: Bytes): Bytes;
function ecrecover(hash: Bytes, sign: Bytes, v: Uint64, malleability_flag: Uint64): Bytes | null;
function panic(msg?: String);
function panicUtf8(msg: Bytes);
function log(msg: String);
function logUtf8(msg: Bytes);
function logUtf16(msg: Bytes);
function storageRead(key: Bytes): Bytes | null;
function storageHasKey(key: Bytes): bool;
function validatorStake(account_id: String): Uint128;
function validatorTotalStake(): Uint128;
function altBn128G1Multiexp(value: Bytes): Bytes;
function altBn128G1Sum(value: Bytes): Bytes;
function altBn128PairingCheck(value: Bytes): bool;
Due to the design of JavaScript VM Contract, some additonal APIs are provided to obtain context, access storage and cross contract call. Since they're not documented at NEAR nomicon. They're explained here.
function jsvmAccountId(): String;
function jsvmJsContractName(): String;
function jsvmMethodName(): String;
function jsvmArgs(): Bytes;
The jsvmAccountId
returns the JavaScript VM's contract account ID.
The jsvmJsContractName
, when called, returns the JavaScript contract name that are called at the moment.
The jsvmJsContractName
returns the method name being called.
The jsvmArgs
return the arguments passed to the method.
function jsvmStorageWrite(key: Bytes, value: Bytes): bool;
function jsvmStorageRead(key: Bytes): Bytes | null;
function jsvmStorageRemove(key: Bytes): bool;
function jsvmStorageHasKey(key: Bytes): bool;
function storageGetEvicted(): Bytes;
These are equivalent to storage*
but access limit to the substate of current JS contract. The jsvmStorageWrite
and jsvmStorageRemove
require and refund deposit to cover the storage delta. jsvmStorage*
access the substate of current JS contract by prefix the key of current JS contract name (deployer's account id). You can use storageRead
and storageHasKey
to get code and state of other JS contracts. More specifically: code of contractA
is stored under the key contractA/code
. state of contractA
is stored under contractA/state/
concat with developer specifid key. And:
jsvmStorageRead(k)
// equvalent to
storageRead(jsvmJsContractName() + '/state/' + k)
When jsvmStorageWrite
write to a key that already exists, the old value would be saved and can be obtained by storageGetEvicted()
. In this case, jsvmStorageWrite returns true
. If key doesn't exist before, returns false
.
When jsvmStroageRemove
remove a key that exists, the old value would be saved and can be obtained by storageGetEvicted()
. In this case, jsvmStroageRemove returns true
. If key doesn't exist before, nothing is removed and returns false
.
function jsvmValueReturn(value: Bytes);
function jsvmCall(contract_name: String, method: String, args: Bytes): any;
function jsvmCallRaw(contract_name: String, method: String, args: Bytes): Bytes;
The jsvmValueReturn
is the version of valueReturn
that should be used in all JavaScript contracts. It play well with jsvmCall
.
The jsvmCall
invoke a synchronous cross contract call, to the given JavaScript contract_name
, method
with args
. And returned the return value parsed as JSON into a JS object.
The jsvmCallRaw
is similar to jsvmCall
, but return the raw, unparsed Bytes.
A few useful on-chain persistent collections are provided. All keys, values and elements are of type Bytes
.
Vector is an iterable implementation of vector that stores its content on the trie. Usage:
import {Vector} from 'near-sdk-js'
// in contract class constructor:
constructor() {
super()
this.v = new Vector('my_prefix_')
}
// Override the deserializer to load vector from chain
deserialize() {
super.deserialize()
this.v = Object.assign(new Vector, this.v)
}
someMethod() {
// insert
this.v.push('abc')
this.v.push('def')
this.v.push('ghi')
// batch insert, extend:
this.v.extend(['xyz', '123'])
// get
let first = this.v.get(0)
// remove, move the last element to the given index
this.v.swapRemove(0)
// replace
this.v.replace(1, 'jkl')
// remove the last
this.v.pop()
// len, isEmpty
let len = this.v.len()
let isEmpty = this.v.isEnpty()
// iterate
for (let element of this.v) {
near.log(element)
}
// toArray, convert to JavaScript Array
let a = this.v.toArray()
// clear
ths.v.clear()
}
LookupMap is an non-iterable implementation of a map that stores its content directly on the trie. It's like a big hash map, but on trie. Usage:
import {LookupMap} from 'near-sdk-js'
// in contract class constructor:
constructor() {
super()
this.m = new LookupMap('prefix_a')
}
// Override the deserializer to load vector from chain
deserialize() {
super.deserialize()
this.m = Object.assign(new LookupMap, this.m)
}
someMethod() {
// insert
this.m.set('abc', 'aaa')
this.m.set('def', 'bbb')
this.m.set('ghi', 'ccc')
// batch insert, extend:
this.m.extend([['xyz', '123'], ['key2', 'value2']])
// check exist
let exist = this.m.containsKey('abc')
// get
let value = this.m.get('abc')
// remove
this.m.remove('def')
// replace
this.m.set('ghi', 'ddd')
}
LookupSet is an non-iterable implementation of a set that stores its content directly on the trie. It's like LookupMap, but it only stores whether the value presents. Usage:
import {LookupSet} from 'near-sdk-js'
// in contract class constructor:
constructor() {
super()
this.s = new LookupSet('prefix_b')
}
// Override the deserializer to load vector from chain
deserialize() {
super.deserialize()
this.s = Object.assign(new LookupSet, this.s)
}
someMethod() {
// insert
this.s.set('abc')
this.s.set('def')
this.s.set('ghi')
// batch insert, extend:
this.s.extend(['xyz', '123'])
// check exist
let exist = this.s.contains('abc')
// remove
this.s.remove('def')
}
UnorderedMap is an iterable implementation of a map that stores its content directly on the trie. Usage:
import {UnorderedMap} from 'near-sdk-js'
// in contract class constructor:
constructor() {
super()
this.m = new UnorderedMap('prefix_c')
}
// Override the deserializer to load vector from chain
deserialize() {
super.deserialize()
this.m.keys = Object.assign(new Vector, this.m.keys)
this.m.values = Object.assign(new Vector, this.m.values)
this.m = Object.assign(new UnorderedMap, this.m)
}
someMethod() {
// insert
this.m.set('abc', 'aaa')
this.m.set('def', 'bbb')
this.m.set('ghi', 'ccc')
// batch insert, extend:
this.m.extend([['xyz', '123'], ['key2', 'value2']])
// get
let value = this.m.get('abc')
// remove
this.m.remove('def')
// replace
this.m.set('ghi', 'ddd')
// len, isEmpty
let len = this.m.len()
let isEmpty = this.m.isEnpty()
// iterate
for (let [k, v] of this.m) {
near.log(k+v)
}
// toArray, convert to JavaScript Array
let a = this.m.toArray()
// clear
this.m.clear()
}
UnorderedSet is an iterable implementation of a set that stores its content directly on the trie. It's like UnorderedMap but it only stores whether the value presents. Usage:
import {UnorderedSet} from 'near-sdk-js'
// in contract class constructor:
constructor() {
super()
this.s = new UnorderedSet('prefix_d')
}
// Override the deserializer to load vector from chain
deserialize() {
super.deserialize()
this.s.elements = Object.assign(new Vector, this.s.elements)
this.s = Object.assign(new UnorderedSet, this.s)
}
someMethod() {
// insert
this.s.set('abc')
this.s.set('def')
this.s.set('ghi')
// batch insert, extend:
this.s.extend(['xyz', '123'])
// check exist
let exist = this.s.contains('abc')
// remove
this.s.remove('def')
// len, isEmpty
let len = this.s.len()
let isEmpty = this.s.isEnpty()
// iterate
for (let e of this.s) {
near.log(e)
}
// toArray, convert to JavaScript Array
let a = this.s.toArray()
// clear
this.s.clear()
}
Due to the architecture of the JSVM, some NEAR host functions, part of Standalone SDK or Rust SDK, are not revelant or being replaced by above JSVM specific APIs. Those unavailable APIs are explained here.
-
The
current_account_id
would always puts the account id of the JavaScript VM contract account in given register. The namingcurrent_account_id
is therefore confusing and not as helpful as a Rust contract. In some case, developer may want to get JavaScript VM contract account name, for example, determines whether it's running on testnet or mainnet, and behaves differently. So we expose this functionality underjsvm_account_id()
. -
The
input
puts the argument passed to call the contract in given register. In JavaScript VM, this is encoded as"js_contract_name\0method_name\0args...
. This format isn't very convinient to developer, therefore, separate APIjsvm_js_contract_name
,jsvm_method_name
andjsvm_args
are provided. -
The
storage_usage
return the storage bytes used by JavaScript VM contract. User doesn't care about the storage usage of the JSVM. Instead, users care about storage usage of a given JavaScript contract. This can be obtained bystorage_read
and count the sum ofregister_len
. -
The
account_balance
andaccount_locked_balance
returns balance and locked_balance of JavaScript VM. Those are also not cared by users. -
The
value_return
is a NEAR primitive that puts the value to return in a receipt. However we would want to access it as a JavaScript return value in a cross contract call. So we have a new APIjsvmValueReturn
, which does return the value in receipt and also as a JavaScript value returned byjsvm_call
. ThejsvmValueReturn
should be used whenever you needvalue_return
. -
abort
is intended to mark error location (line number). A full stacktrace with line numbers is provided by QuickJS, available when you throw a JS Error. So this API isn't needed. -
Promise APIs act on the JSVM contract and could create subaccount, use the balance from JSVM account.JSVM would be a common VM used by the community instead of a Rust contract owned by the deployer. Therefore, promise APIs are not allowed.
-
The
storage_write
andstorage_remove
have access to all JavaScript contract codes and states deployed on JSVM. User can only write to their account owned code and state, as a substate of the JSVM. Therefor these two APIs are disallowed. Usejsvm_storage_write
andjsvm_storage_remove
instead. Read to other people owned code and state is allowed, as they're public as part of the blockchain anyway.
You can also layout your project by install the npm package manually:
yarn add near-sdk-js
# or
npm install near-sdk-js
It is tested on Ubuntu 20.04, Intel Mac and M1 Mac. Other linux should also work but they're not tested.
- Make sure you have
wget
,make
,cmake
andnodejs
. On Linux, also make sure you havegcc
. - Run
make
to get platform specificqjsc
andjsvm
contract inres
folder.
See https://github.com/near/near-sdk-js/tree/master/tests
near js
subcommand in near-cli is a recent feature. Under the hood, it is encoding a special function call to jsvm contract.
The equivalent raw command is:
near call <jsvm-account> deploy_js_contract --accountId <your-account> --args $(cat <contract-name>.base64) --base64 --deposit 0.1
The equivalent raw command is:
near call <jsvm-account> call_js_contract --accountId <your-account> --args <encoded-args> --base64
# where `<encoded-args>` can be obtained by:
node scripts/encode_call.js <your-account> <method-name> <args>
The equivalent raw command is:
near call <jsvm-account> remove_js_contract --accountId <your-account>