kucoin backend

[jaredgoldman]

This PR includes a brokerd module for Kucoin's api. So far, only code for market data has been implemented. This includes:

• Fetching a users historical trades
• Subscribing to real-time market data via the Kucoin websocket feed
• Multi-chartz support
• Fetching OHLC historical market data
• Kucoin user auth

• Symbol fuzzy search

  Remaining tasks for full support are detailed in https://github.com/pikers/piker/issues/499

[goodboy]

@jaredgoldman this ready for serious review?

[jaredgoldman]

@goodboy I think we're ready for some more testing

[goodboy]

@algorandpa seems like i can only stage order with size=0 but other then that historical and live data seems right 🏄🏼

[goodboy]

@jaredgoldman as mentioned in chat as well we've got some kinda strange thing going on with trade rate calcs:

• there's some kind of trade event being picked up by the fsp system 10/sec
• you can see flat curve for this if you leave it running
• this does not match up with actual vlm / clearing rates

---

yeah so if you look at how we (currently) implement the trade rate, it's basically just counting the number of vlm-containing-trades (aka clears):

• https://github.com/pikers/piker/blob/master/piker/fsp/_volume.py#L318 (where it's consumed and reported from the flows fsp)
• https://github.com/pikers/piker/blob/master/piker/fsp/_volume.py#L193 (where it's calced and written to shm for the dolla_vlm fsp)

my guess is somehow we've got a tick being generated by ur backend code that's fitting into the trade tick type, without it actually being a trade and thus this count is being triggered?

See https://github.com/pikers/piker/blob/master/piker/fsp/_volume.py#L162 for the filter loop.

hopefully that helps to start digging, if not we can resume in chat 👍🏼

[jaredgoldman]

@goodboy merged the binance changes and added anext()

[goodboy]

Looking the code in the ol' vim i'm seeing a lot of these brother:

[goodboy]

Added patch commit to get things working again after your review changes in #498

[jaredgoldman]

@goodboy I refactored the subscribe and stream_messages logic to:

• Only use one ws connection and one ping_task
• Be less nested and more readable

lol, the earlier implementation was wildy inefficient spawning a diff ws connection and ping task for each symbol. In addition I've addressed the comments for which I'm not waiting on a reply from you (no rush). We're also getting to the point where we should consider factoring this into a few different files, especially before adding order control.

[goodboy]

We're also getting to the point where we should consider factoring this into a few different files, especially before adding order control.

yes, if you want to in this PR it's fine by me. currently we something like:

• core backend api client and general utils in .api
• data feed setup, search task and history client in a .feed
• all the trade control stuff (which i don't think you have any of yet) in a .broker

[jaredgoldman]

@goodboy https://github.com/pikers/piker/issues/499

[goodboy]

@jaredgoldman :boom:

excellent work yet again compadre 🏄🏼

nice follow up issue and final linting before getting this in ma dawg!