Botflow

0.2.0 alpha

• Dataflow programming framework
• Paralleled in coroutines and ThreadPool
• Type- and content-based route function
• Interactive programming with Jupyter Notebook

Requirements

Graphviz.

Installing

Install and update using pip:

pip install -U botflow

Documentation

http://botflow.readthedocs.io

What's dataflow programming?

All functions are connected by pipes (queues) and communicate by data.

When data come in, the function will be called and return the result.

Think about the pipeline operation in unix: ls|grep|sed.

Benefits:

1. Decouple data and functionality
2. Easy to reuse

Botflow provides pipe and route. It makes dataflow programming and powerful data flow processes easier.

Botflow is...

• Simple

Botflow is easy to use and maintain, does not need configuration files, and knows about asyncio and how to parallelize computation.

Here's one of the simple applications you can make:

_Load the price of Bitcoin every 2 seconds. Advantage price aggregator sample can be found here.

from botflow import *


def main():
    Pipe(

        Timer(delay=2),  # send timer data to pipe every 2 seconds
        "http://api.coindesk.com/v1/bpi/currentprice.json",  # send url to pipe when timer trigger
        HttpLoader(),  # read url and load http response
        lambda r: r.json['bpi']['USD']['rate_float'],  # read http response and parse as json
        print,  # print out

    )

    Bot.render('ex_output/simple_bitcoin_price')
    Bot.run()
main()

Or write in chain style

from botflow import *
p_cd_bitcoin=Pipe().Timer(delay=2).Loop("http://api.coindesk.com/v1/bpi/currentprice.json")\
            .HttpLoader().Map(lambda r: r.json['bpi']['USD']['rate_float']).Map(print)

p_cd_bitcoin.run()

• Flow Graph

With render function: Bot.render('bitcoin_arbitrage') Botflow will render the data flow network into a graphviz image. below is the flow graph generated by Botflow.Aggreate 6 exchanges bitcoin price for trading.

• Fast

Nodes will be run in parallel, and they will perform well when processing stream data. :Web Crawle: Botflow is 10x fatter than Scrapy

• Replay-able

With replay mode enabled: config.replay_mode=True when an exception is raised at step N, you don't need to run from setup 1 to N. Botflow will replay the data from nearest completed node, usually step N-1. It will save a lot of time in the development phase.

Release

RoadMap

• Will add Httpserver support(REST,Websocket).
• Will support server machine learning Model online.
• Finshe the api reference doc.
• Rename project to Botflow.?

More about Botflow

Data- programming is typically applied to streams of structured data for filtering, transforming, aggregating (such as computing statistics), or calling other programs.

Botflow has a few basic concepts to implement Dataflow programming .

• Source

  It is feed stream data to the pipe.

  • Timer: It will send a message in the pipe by timer param. delay, max_time until some finished
  • Pipe.run: you can use Pipe.run to trigger the data into pipe. By default it will feed int 0

• Function

  It is callable unit.Any callable function and object can work as Node. It is driven by data. Custom functions work as Map unit. There are some built-in nodes:

  • Fetch: (Alias:HttpLoader) Get a url and return the HTTP response
  • AioFile: for file I/O.
  • SpeedLimit: limit the stream speed limit
  • Delay: delay in special second.
  • Map : Work ad Convert unit.
  • Filter : Drop data from pipe if it does not match some condition
  • Flat : Drop data from pipe if it does not match some condition

• Route

  It will be used to create a complex data flow network, not just one main process. Botflow can nest Routes inside Routes. It is a powerful concept. There are some pre built-in Route:

  • Pipe: It is the main stream process of the program. All units will work inside.
  • Tee : (Alias:Branch) Duplicate data from parent pipe to a child pipe as branch.
  • Zip : Combine multi pipes result to list.
  • Link: (Alias: LinkTo) Route flow to any Node or Route for making loop , circle

All units (Pipe, Node, Route) communicate via queues and perform parallel computation in coroutines. This is abstracted so that Botflow can be used with only limited knowledge of asyncio.

Contributing

Donate

Links