Architecture

Mt4 and Python Architecture

Mt4 runs an expert attached to a chart in a thread. Python in a Mt4 expert starts a main thread that is unique to the application, and starts another worker thread. The main thread is blocked until Python is unloaded, and everything takes place in the worker thread. The worker thread can start more threads.

You can send commands from Mt4 to Python, but Python can't send commands to Mt4. So what you can do is queue up things for Mt4 to do, and periodically get Mt4 to ask Python if it has anything it wants Mt4 to do. It's easiest to think of registering an EventTimer with your expert, and to have OnTimer function periodically ask Python if it has any work. (Remember that EventTimer events fire even if the market is closed or you have no connection, unlike OnTick events.)

The problem is that Mt4 does not have an Eval command, so if Python wants Mt4 to do something, like OrderSend, we can't just pass a string to Mt4 that says OrderSend(...) and expect Mt4 to eval it: there is no eval. So what we have done is write a simplistic replacement to what should be an Eval command in Mt4, that will parse as string and execute it. See that function zMt4LibProcessCmd in MQL4/Libraries/OTMql4/OTLibMt4ProcessCmd.mq4 of the OTMql4Lib package (https://github.com/OpenTrading/OTMql4Lib/). See also OTLibProcessCmd.mq4 for an example of how to extend zMt4LibProcessCmd to process your own functions.

This is not a proper replacement to an Eval function. It only executes commands that it knows about, and has no ability to decipher complex commands. Still, it allows Python to make decisions and ask Mt4 to do things, which is enough for order and account processing. You can use this for bi-directional interaction with a program outside of Mt4. For example, you could use a call from OnTick into Python to send tick and bar info to another program, and then use a call from OnTimer into Python to ask if there are any requests back from the program that it wants Mt4 to evaluate.

You may also want to use threads and queues within Python to make sure that any command that you ask Python to execute is accomplished within the time it takes to get the next tick: programming with Python iterators is probably a good idea. It is untested as to whether the Python code dispatched from OnTick can block the chart or not.

In our demo experts, we have set the timer interval to be quite high: 10 seconds. What you attach the expert to the chart, you can set it to what you want. It's set high for testing; obviously if you have a lot of charts with OTMqlPy and set the interval low, it may add a lot of work to Mt4.

---

Parent: Home