Skip to content

callbacks.md

Latest commit

 

History

History
53 lines (39 loc) · 2.67 KB

callbacks.md

File metadata and controls

53 lines (39 loc) · 2.67 KB

Using Callbacks

Cryptofeed is a library that uses asyncio to handle asynchronous events. When fh.run() is called, the main program thread of execution will block until an exception is hit or the user terminates the program (There is a slight exception to this, if run is called with the kwarg start_loop=False the feedhandler will not be started, the user can add more tasks/coroutines, and will then be responsible for starting the event loop later). Because the program is effectively blocked on the event loop, the user needs to define callbacks that will handle data from cryptofeed. Only data you register for will be delivered via these callbacks.

Callback Types

There are two types of callbacks supported in cryptofeed, raw and backend. The raw callbacks deliver the data directly to the specified function. Backend callbacks take the data and do something else with it (typically store or send). Some examples of the backend callbacks are Redis, Postgres and TCP. You might use the Redis or Postgres callbacks to store the data, and you could use the TCP callback to send data to another application for processing.

The raw callbacks are defined here. They are:

  • Trade
  • Ticker
  • Book
  • Open Interest
  • Funding
  • Liquidation
  • Candles
  • Index
  • L1Book (aka Top of Book)
  • Order Info
  • User Fills
  • Transactions
  • Balances

It's important to note that if your choose to use the raw callbacks and your callbacks are async functions, you do not need to use these wrappers (like is commonly shown in the example code). You can use your callback functions without wrapping them in TradeCallback, TickerCallback, etc.

Every callback has the same signature, two positional arguments, the data object and the receipt timestamp. The data object differs by data type. The data objects are defined in types.pyx

Backends

The backends are defined here. Currently the following are supported:

  • Arctic
  • ElasticSearch
  • GCP Pub/Sub
  • InfluxDB
  • Kafka
  • MongoDB
  • Postgres
  • QuestDB
  • RabbitMQ
  • Redis
  • Redis Streams
  • TCP/UDP/UDS sockets
  • VictoriaMetrics
  • ZMQ

There are also a handful of wrappers defined here that can be used in conjunction with these and raw callbacks to convert data to OHLCV, throttle data, etc.

Performance Considerations

Do not do anything computationally intensive in your callbacks, or this will greatly impact the performance of cryptofeed. Data should be quickly processed and passed along to another process/application/etc or a backend callback should be used to forward the data elsewhere. If possible, use async libraries in your callbacks!
⚡