Skip to content

Latest commit

 

History

History
120 lines (97 loc) · 5.7 KB

README.md

File metadata and controls

120 lines (97 loc) · 5.7 KB
Raw

pyiqfeed

Reads and parses data from IQFeed (http://www.iqfeed.net). Now supports API version 6.0.

You likely want DTN's API docs handy. They are available at:

https://www.iqfeed.net/dev/api/docs/index.cfm

You need a subscription to IQFeed or this won't work. This library is usually kept up to date to the current version of IQFeed. The variable FeedConn.protocol (available if you import pyiqfeed or just look at conn.py) is the version of the IQFeed protocol currently being used.

Then library depends on Numpy. Much of the data is returned as an numpy array or numpy structured array. If you are using python for trading you are probably using numpy in your trading code anyway so this should not be a big deal. Data that is not numerical market data is returned as a namedtuple. Different namedtuples are defined for different kinds of data.

Most of the code is in the file conn.py. There is a class service.py that can launch the IQFeed service on a windows, OSX or Linux machine (so long as IQFeed has been installed using a recent version of Wine if you aren't on Windows).

If you are installing on OSX, install Wine or Wine-Devel using one of Homebrew, Macports, or Fink and then install IQFeed inside that. The Mac Download of IQFeed is basically just a CodeWeavers Wine "bottled" version of IQFeed, not a native Mac version and because it's been "bottled", passing arguments to IQFeed at startup is complicated. This library assumes that you have installed Wine and then IQFeed and are not using DTN's Mac download. If you choose to use the DTN package, things will still work but you will have to startup IQFeed yourself and pass it parameters like the App Name, login and password using messages from AdminConn, instead of using the command line at startup.

On Ubuntu, use the Wine-Devel from the Wine Development Team's ppa, not the Wine that comes from Canonical.

If you want to run headless you need to have the package xvfb installed. IQFeed.exe is launched under wine and wants to create a GUI. If X isn't running, wine will crash. If you have xvfb, the virtual framebuffer "fake" X server running, this fools wine into believing it has launched the IQFeed.exe GUI. This is also useful if you find the IQFeed.exe gui annoying.

You cannot install IQFeed in a Windows Virtual Machine on the same physical machine. It won't work unless you subscribe to data while running within the same virtual machine. DTN does not allow you to get data on a machine other than the one that IQConnect.exe is running on, even if it's the same physical machine. IQFeed.exe listens on 127.0.0.1. So if you absolutely must run it in a virtual machine and run code that talks to IQFeed on the main machine or vice-versa, there are ways to make it work. But please DO NOT ask me how to do it for you. Hint: Tunneling or run something on the same machine to forward requests.

For an example of how to use it do the following:

  1. Install the package or put it in your search path. python setup.py works.

  2. Create a folder in the same directory as example.py named localconfig and place a file named passwords.py in it. In this file you need 3 lines:

 
 dtn_product_id = "PRODUCT_ID_FROM_DTN"
 dtn_login="Your_IQFEED_LOGIN"
 dtn_password="Your_IQFEED_PASSWORD"
  1. Run example.py using something like python3 ./example.py. You must use python ver >= 3.5.

This exercises many different parts of the library depending on what options you pass it. The best documentation for the library is just reading the source. Public functions are documented and the code has been deliberately kept simple.

It works for me (TM). A diligent effort will be made to squash any bugs reported to me. Bug reports which do not include a short, easy-to-use python script that reproduces the bug and an explanation of the bug will be ignored. Pull requests with bug-fixes and enhancements are greatly encouraged and appreciated.

For all pull requests please ensure the following:

  1. Ensure all code has been run through at least some of the various python code checkers, preferably pylint. If you have IntelliJ or PyCharm, Analyze->Inspect Code does something pretty similar. Python is a duck-typed language. This means many errors can exist in your code and you'll only find them if your tests actually exercise that part of the code. Anywhere near 100% test coverage is basically impossible so please, please run the checkers. Please do not complain about numpy related errors from pylint and how pylint won't work with numpy, just read the pylint docs on how to make it work with external libraries that call into C. Ensure all code is PEP8 compliant. The plan is to put type annotations everywhere and ensure that code passes mypy static type checks.

  2. Please keep external dependencies to a minimum. Ideally only use packages that are built into CPython and numpy. Please do not import other third-party packages, not even pandas, not even in example or test code.

  3. Documentation is good. But gratuitous documentation is bad. The best documentation is well chosen names and simple code so you don't have to read anything to understand what it does. Comments have to be maintained just like other code. No comments is better than comments out of sync with the code. If you are tempted to add a description to a function that says something like "Private function" or "Implementation Detail", consider changing the function's name so it starts with an underscore instead and leave it at that.

  4. Keep your code as simple and straightforward as you can. If you think you need to describe something tricky in a comment, DON'T. Change what you are doing so it's not tricky instead. Concise pydocs which describe how to use functions and interfaces that you expect library users to use is good.