Huge News!Announcing our $40M Series B led by Abstract Ventures.Learn More
Socket
Sign inDemoInstall
Socket

binance-sdk

Package Overview
Dependencies
Maintainers
1
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

binance-sdk

Binance Python SDK

  • 1.3.4
  • PyPI
  • Socket score

Maintainers
1

binance-sdk

binance-sdk is an another unofficial Binance SDK for python 3.7+, which:

  • Based on Binance Official API Docs v3.
  • Uses Binance's new websocket stream which supports live pub/sub so that we only need ONE websocket connection.
  • Has an optional pandas.DataFrame support. If pandas is installed, columns of all stream data frames are renamed for readability.
  • Based on python async/await
  • Manages the order book for you (handled by OrderBookHandlerBase), so that you need not to worry about websocket reconnection and message losses. For details, see the section OrderBookHandlerBase
  • Supports to change API endpoints, so that we could use faster API hosts.

Install

# Without pandas support
pip install binance-sdk

or

# With pandas support
pip install binance-sdk[pandas]

Basic Usage

#!/usr/bin/env python

import asyncio
from binance import Client

client = Client()

async def main():
    print(await client.get_exchange_info())

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

Handling messages

Binance-sdk provides handler-based APIs to handle all websocket messages, and you are able to not worry about websockets.

#!/usr/bin/env python

from binance import Client, TickerHandlerBase, SubType

client = Client(api_key)

async def main():
    # Implement your own TickerHandler.
    class TickerPrinter(TickerHandlerBase):
        async def receive(self, payload):
            """The function to receive ticker streams.
            The function could either be sync or async

            Args:
                payload (dict): the raw stream payload which is
                message['data'] of the original stream message
            """

            # If binance-sdk is installed with pandas support, then
            #   `ticker` will be a `DataFrame` with columns renamed
            # Otherwise, it is unnecessary to call `super().receive`.
            ticker_df = super().receive(payload)

            # Just print the ticker
            print(ticker_df)

    # Register the handler for `SubType.TICKER`
    client.handler(TickerPrinter())

    # Subscribe to ticker change for symbol BTCUSDT
    await client.subscribe(SubType.TICKER, 'BTCUSDT')

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

# Run the loop forever to keep receiving messages
loop.run_forever()

# It prints a pandas.DataFrame for each message

#    type        event_time     symbol   open            high            low            ...
# 0  24hrTicker  1581597461196  BTCUSDT  10328.26000000  10491.00000000  10080.00000000 ...

# ...(to be continued)

Subscribe to more symbol pairs and types

# This will subscribe to
# - bnbusdt@aggTrade
# - bnbusdt@depth
# - bnbbtc@aggTrade
# - bnbbtc@depth
await client.subscribe(
    # We could also subscribe multiple types
    #   for both `BNBUSDT` and 'BNBBTC'
    [
        SubType.AGG_TRADE,
        SubType.ORDER_BOOK
    ],
    # We could subscribe more than one symbol pairs at a time
    [
        # Which is equivalent to `BNBUSDT`
        'BNB_USDT',
        'BNBBTC'
    ]
)

And since we subscribe to THREE new types of messages, we need to set the handlers each of which should isinstance() of one of

  • TradeHandlerBase
  • AggTradeHandlerBase
  • OrderBookHandlerBase
  • KlineHandlerBase
  • MiniTickerHandlerBase
  • TickerHandlerBase
  • AllMarketMiniTickersHandlerBase
  • AllMarketTickersHandlerBase
  • AccountInfoHandlerBase
  • AccountPositionHandlerBase
  • BalanceUpdateHandlerBase
  • OrderUpdateHandlerBase
  • OrderListStatusHandlerBase
  • HandlerExceptionHandlerBase a special handler to handle stream exceptions
client.handler(MyTradeHandler(), MyOrderBookHandler(), MyKlineHandler())

Subscribe to user streams

# Before subscribe to user stream, you need to provide `api_secret` (and also `api_key`)
client.secret(api_secret)

# Or, you should provide `api_secret` when initialize the client
# ```
# client = Client(api_key, api_secret)
# ```

# binance-sdk will handle user listen key internally without your concern
await client.subscribe(SubType.USER)

Subscribe to handler exceptions

Binance-sdk receives stream messages in background tasks, so sometimes it is difficult to detect the exceptions raised in receive function of user handlers.

Fortunately, we could use HandlerExceptionHandlerBase

from binance import (
    HandlerExceptionHandlerBase,
    KlineHandlerBase
)

class KlineHandler(KlineHandlerBase):
    def receive(self, payload):
        raise RuntimeError('this will ruin my day')

class HandlerExceptionHandler(HandlerExceptionHandlerBase):
    async def receive(self, exception):
        # By calling `super().receive(exception)`,
        # it will print the error stack.
        super().receive(exception)

        await send_to_monitor(exception)

client.handler(KlineHandler())
client.handler(HandlerExceptionHandler())

If you just want to print error stacks, we could:

client.handler(HandlerExceptionHandlerBase())

APIs

Client(**kwargs)

All arguments of the constructor Client are keyworded arguments and all optional.

  • api_key? str=None binance api key
  • api_secret? str=None binance api secret
  • request_params? dict=None global request params for aiohttp
  • stream_retry_policy? Callable[[int, Exception], Tuple[bool, int, bool]] retry policy for websocket stream. For details, see RetryPolicy
  • stream_timeout? int=5 seconds util the stream reach an timeout error
  • api_host? str='https://api.binance.com' to specify another API host for rest API requests. 这个参数的存在意义,使用方法,不累述,你懂的。
  • stream_host? str='wss://stream.binance.com' to specify another stream host for websocket connections.

Create a binance client.

Each API method accepts only keyworded arguments (kwargs) and has verbosed Python doc strings (Google style) which you could check out when you are coding.

The following example shows how to create a new order.

from binance import (
    OrderSide,
    OrderType,
    TimeInForce
)

# All arguments are keyworded arguments.
await client.create_order(
    symbol='BTCUSDT',

    # You could use string `BUY` (NOT recommended) instead of
    # the built-in enum types of Binance-sdk.

    # But it is a good practise to use enums which could help
    # us to avoid spelling mistakes, and save our money.
    side=OrderSide.BUY,
    type=OrderType.LIMIT,
    timeInForce=TimeInForce.GTC,

    # Binance-sdk will not handle Decimals for you,
    # so you'd better to know how to deal with python float precisions.
    # Or you could use string-type quantity.
    quantity=10.,

    # It is better to use string type instead of float.
    # The same as `quantity`
    price='7000.1'
)

client.key(api_key) -> self

Define or change api key. This method is unnecessary if we only request APIs of SecurityType.NONE

client.secret(api_secret) -> self

Define or change api secret, especially when we have not define api secret in Client constructor.

api_secret is not always required for using binance-sdk. See Endpoint security type

await client.get(uri, **kwargs)

await client.post(uri, **kwargs)

await client.put(uri, **kwargs)

await client.delete(uri, **kwargs)

  • uri str the request url
  • security_type? SecurityType endpoint security type. Defaults to SecurityType.NONE.

Send a GET/POST/PUT/DELETE HTTPs request.

await client.subscribe(subtype, *subtype_params) -> None

await client.subscribe(*subscriptions) -> None

  • subtype str subscription type, should be one of SubType.*s. For details, see SubType
  • subtype_params List params for a certain subtype
  • subscriptions List[Tuple] a pack of subscriptions each of which is a tuple of subtype and *subtype_params.

Subscribe to a stream or multiple streams. If no websocket connection is made up, client.subscribe will also create a websocket connection.

from binance import SubType, KlineInterval

await client.subscribe(SubType.TICKER, 'BNBUSDT')

# SubType.ALL_MARKET_MINI_TICKERS with default param
await client.subscribe(SubType.ALL_MARKET_MINI_TICKERS)

# SubType.ALL_MARKET_MINI_TICKERS with update interval 3000ms
await client.subscribe(SubType.ALL_MARKET_MINI_TICKERS, 3000)

# Subcribe to multiple types
await client.subscribe(
    (SubType.KLINE, 'BTC_USDT', KlineInterval.DAY),
    (SubType.TICKER, 'BNBUSDT'),
    (
        [
            SubType.ORDER_BOOK,
            SubType.TRADE
        ],
        ['BNBUSDT', 'BTCUSDT']
    ),
    (SubType.ALL_MARKET_MINI_TICKERS,) # <-- PAY ATTENTION to the `,` here
)

Possible exceptions:

  • InvalidSubParamsException
  • UnsupportedSubTypeException
  • InvalidSubTypeParamException
  • StreamAbandonedException

client.start() -> self

Start receiving streams

client.stop() -> self

Stop receiving streams

await client.close(code=4999) -> None

  • code int=4999 the custom close code for websocket. It should be in the range 4000 - 4999

Close stream connection, clear all stream subscriptions and clear all handlers.

client.handler(*handlers) -> self

  • handlers List[Union[HandlerExceptionHandlerBase,TradeHandlerBase,...]]

Register message handlers for streams. If we've subscribed to a stream of a certain subtype with no corresponding handler provided, the messages of subtype will not be handled.

Except for HandlerExceptionHandlerBase, handlers each of whose name ends with Base should be inherited before use.

Typically, we need to override the def receive(self, payload) method.

class MyTradeHandler(TradeHandlerBase):
    async def receive(self, payload):
        # If pandas is installed, then `payload` is a `pandas.DataFrame`,
        #   otherwise is a dict.
        df = super().receive(payload)

        # If you don't want the `pandas.DataFrame`, use `payload` directly

        await saveTrade(df)

client.handler(MyTradeHandler())

We could also register multiple handlers at one time

client.handler(MyTradeHandler(), MyTickerHandler())

If we register an invalid handler, an InvalidHandlerException exception will be raised.

SubType

In this section, we will note the parameters for each subtypes

SubType with parameters symbol and interval

  • SubType.KLINE

And interval should be one of the KlineInterval enumerables

SubTypes with a param symbol

  • SubType.TRADE
  • SubType.AGG_TRADE
  • SubType.MINI_TICKER
  • SubType.TICKER
  • SubType.ORDER_BOOK

SubTypes with an optional param updateInterval=1000 (ms)

  • SubType.ALL_MARKET_MINI_TICKERS
  • SubType.ALL_MARKET_TICKERS

Subtype with no param

  • SubType.USER

RetryPolicy

Retry policy is used by binance-sdk to determine what to do next after the client fails to do some certain thing.

abandon, delay = stream_retry_policy(info)

# `info.fails` is the counter number of
#   how many times has the stream encountered the connection failure.
# If the stream is disconnected just now for the first time, `info.fails` will be `1`

# `info.exception` is the exception that raised which caused the failure

# If abandon is `True`, then the client will give up reconnecting.
# Otherwise:
# - The client will asyncio.sleep `delay` seconds before reconnecting.

OrderBookHandlerBase(**kwargs)

  • kwargs
    • limit? int=100 the limit of the depth snapshot
    • retry_policy? RetryPolicy=

By default, binance-sdk maintains the orderbook for you according to the rules of the official documentation.

Specifically, OrderBookHandlerBase does the job.

We could get the managed OrderBook object by method handler.orderbook(symbol).

async def main():
    client = Client(api_key)

    # Unlike other handlers, we usually do not need to inherit `OrderBookHandlerBase`,
    #   unless we need to receive the raw payload of 'depthUpdate' message
    handler = OrderBookHandlerBase()

    client.handler(handler)
    await client.subscribe(SubType.ORDER_BOOK, 'BTCUSDT')

    # Get the reference of OrderBook object for 'BTCUSDT'
    orderbook = handler.orderbook('BTCUSDT')

    while True:
        # If the `retry_policy` never abandon a retry,
        #   the 'try' block could be emitted
        try:
            await orderbook.updated()
        except Exception as e:
            print('exception occurred')
        else:
            await doSomethingWith(orderbook.asks, orderbook.bids)

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

loop.run_forever()

OrderBook(symbol, **kwargs)

  • symbol str the symbol name
  • kwargs
    • limit? int=100 limit of the orderbook
    • client Client=None the instance of binance.Client
    • retry_policy? Callable[[int, Exception], (bool, int, bool)] retry policy for depth snapshot which has the same mechanism as Client::stream_retry_policy

OrderBook is another public class that we could import from binance-sdk and you could also construct your own OrderBook instance.

async def main():
    # PAY attention that `orderbook` should be run in an event loop
    orderbook = OrderBook('BTCUSDT', client=client)

    await orderbook.updated()

    print(orderbook.asks)

orderbook.set_client(client) -> None

  • client Client the instance of binance.Client

Set the client. If client is not specified in the constructor, then executing this method will make the orderbook to fetch the snapshot for the first time.

orderbook.set_limit(limit) -> None

  • limit int

Set depth limit which is used by binance reset api.

orderbook.set_retry_policy(retry_policy) -> None

  • retry_policy RetryPolicy

Set retry policy of the certain orderbook

property orderbook.ready -> bool

There is a property getter in orderbook to detect whether the asks and bids are updated in the orderbook.

If there is a network malfunction of the stream which causing the gap between two depth update messages, orderbook will fetch a new snapshot from the server, and during that time and before we merge the snapshot, orderbook.ready is False.

property orderbook.asks -> list

property orderbook.bids -> list

Get asks and bids in ascending order.

orderbook.update(payload) -> bool

  • payload dict the data payload of the depthUpdate stream message

Returns True if the payload is valid and is updated to the orderbook, otherwise False

If the return value is False, the orderbook will automatically start fetching the snapshot

await orderbook.fetch() -> None

Manually fetch the snapshot.

For most scenarios, you need NOT to call this method because once there is an invalid payload, the orderbook will fetch the snapshot itself.

await orderbook.updated() -> None

Wait for the next update of the orderbook.

We could also await orderbook.updated() to make sure the orderbook is ready.

If the orderbook fails to fetch depth snapshot for so many times which means the fetching is abanboned by the retry_policy, an aiohttp exception will be raised.

Listen to the updates of orderbook
async def start_listening_updates(orderbook):
    # This is an infinite loop
    while True:
        await orderbook.updated()
        # do something

def start():
    return asyncio.create_task(start_listening_updates(orderbook))

task = start()

# If we want to stop listening
task.cancel()

License

MIT

Keywords

FAQs


Did you know?

Socket

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Install

Related posts

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

Packages

npm

Stay in touch

Get open source security insights delivered straight into your inbox.


  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc