Tested on macOS Mojave / Windows 10 in Parallels Desktop container.
Working in production on Debian 10 / Wine 4.
- About the Project
- Installation
- Documentation
- Usage
- Live data and streaming events
- Error handling
- License
This project was developed to work as a server for the Backtrader Python trading framework. It is based on ZeroMQ sockets and uses JSON format to communicate. But now it has grown to the independent project. You can use it with any programming language that has ZeroMQ binding.
Backtrader Python client is located here: Python Backtrader - Metaquotes MQL5
Thanks to the participation of freedumb2000, the project moved to a new level.
New features:
- support for multiple datastreams in parallel for any combination of symbols and timeframes independently of the timeframe and symbol of the attached chart
- support for tick data
- support for direct download as CSV files
- automatic retry binding to sockets. When running under Wine in Linux, sockets will be blocked for 60 seconds if closed uncleanly. This can happen if the client is still connected while the EA gets reloaded. skip re-initialization on chart timeframe change
In development:
- Devitation
- Stop limit orders
- Install ZeroMQ for MQL5 https://github.com/dingmaotu/mql-zmq
- Put
include/Json.mqh
from this repo to your MetaEditorinclude
directoty. - Download and compile
experts/JsonAPI.mq5
script. - Check if Metatrader 5 automatic trading is allowed.
- Attach the script to a chart in Metatrader 5.
- Allow DLL import in dialog window.
- Check if the ports are free to use. (default:
15555
,15556
,15557
,15558
)
The script uses four ZeroMQ sockets:
System socket
- recives requests from client and replies 'OK'Data socket
- pushes data to client depending on the request via System socket.Live socket
- automatically pushes last candle when it closes.Streaming socket
- automatically pushes last transaction info every time it happens.
The idea is to send requests via System socket
and recieve results/errors via Data socket
. Event handlers should be created for Live socket
and Streaming socket
because the server sends data to theese sockets automatically. See examples in Live data and streaming events section.
System socket
request uses default JSON dictionary:
{
"action": null,
"actionType": null,
"symbol": null,
"chartTF": null,
"fromDate": null,
"toDate": null,
"id": null,
"magic": null,
"volume": null,
"price": null,
"stoploss": null,
"takeprofit": null,
"expiration": null,
"deviation": null,
"comment": null
}
Check out the available combinations of action
and actionType
:
action | actionType | Description |
---|---|---|
CONFIG | null | Set script configuration |
RESET | null | Reset subscribed symbols |
ACCOUNT | null | Get account settings |
BALANCE | null | Get current balance |
POSITIONS | null | Get current open positions |
ORDERS | null | Get current open orders |
HISTORY | DATA | Get data history |
HISTORY | TRADES | Get trades history |
HISTORY | WRITE | Download history data as CSV |
TRADE | ORDER_TYPE_BUY | Buy market |
TRADE | ORDER_TYPE_SELL | Sell market |
TRADE | ORDER_TYPE_BUY_LIMIT | Buy limit |
TRADE | ORDER_TYPE_SELL_LIMIT | Sell limit |
TRADE | ORDER_TYPE_BUY_STOP | Buy stop |
TRADE | ORDER_TYPE_SELL_STOP | Sell stop |
TRADE | POSITION_MODIFY | Position modify |
TRADE | POSITION_PARTIAL | Position close partial |
TRADE | POSITION_CLOSE_ID | Position close by id |
TRADE | POSITION_CLOSE_SYMBOL | Positions close by symbol |
TRADE | ORDER_MODIFY | Order modify |
TRADE | ORDER_CANCEL | Order cancel |
Python 3 API class example:
import zmq
class MTraderAPI:
def __init__(self, host=None):
self.HOST = host or 'localhost'
self.SYS_PORT = 15555 # REP/REQ port
self.DATA_PORT = 15556 # PUSH/PULL port
self.LIVE_PORT = 15557 # PUSH/PULL port
self.EVENTS_PORT = 15558 # PUSH/PULL port
# ZeroMQ timeout in seconds
sys_timeout = 1
data_timeout = 10
# initialise ZMQ context
context = zmq.Context()
# connect to server sockets
try:
self.sys_socket = context.socket(zmq.REQ)
# set port timeout
self.sys_socket.RCVTIMEO = sys_timeout * 1000
self.sys_socket.connect('tcp://{}:{}'.format(self.HOST, self.SYS_PORT))
self.data_socket = context.socket(zmq.PULL)
# set port timeout
self.data_socket.RCVTIMEO = data_timeout * 1000
self.data_socket.connect('tcp://{}:{}'.format(self.HOST, self.DATA_PORT))
except zmq.ZMQError:
raise zmq.ZMQBindError("Binding ports ERROR")
def _send_request(self, data: dict) -> None:
"""Send request to server via ZeroMQ System socket"""
try:
self.sys_socket.send_json(data)
msg = self.sys_socket.recv_string()
# terminal received the request
assert msg == 'OK', 'Something wrong on server side'
except AssertionError as err:
raise zmq.NotDone(err)
except zmq.ZMQError:
raise zmq.NotDone("Sending request ERROR")
def _pull_reply(self):
"""Get reply from server via Data socket with timeout"""
try:
msg = self.data_socket.recv_json()
except zmq.ZMQError:
raise zmq.NotDone('Data socket timeout ERROR')
return msg
def live_socket(self, context=None):
"""Connect to socket in a ZMQ context"""
try:
context = context or zmq.Context.instance()
socket = context.socket(zmq.PULL)
socket.connect('tcp://{}:{}'.format(self.HOST, self.LIVE_PORT))
except zmq.ZMQError:
raise zmq.ZMQBindError("Live port connection ERROR")
return socket
def streaming_socket(self, context=None):
"""Connect to socket in a ZMQ context"""
try:
context = context or zmq.Context.instance()
socket = context.socket(zmq.PULL)
socket.connect('tcp://{}:{}'.format(self.HOST, self.EVENTS_PORT))
except zmq.ZMQError:
raise zmq.ZMQBindError("Data port connection ERROR")
return socket
def construct_and_send(self, **kwargs) -> dict:
"""Construct a request dictionary from default and send it to server"""
# default dictionary
request = {
"action": None,
"actionType": None,
"symbol": None,
"chartTF": None,
"fromDate": None,
"toDate": None,
"id": None,
"magic": None,
"volume": None,
"price": None,
"stoploss": None,
"takeprofit": None,
"expiration": None,
"deviation": None,
"comment": None
}
# update dict values if exist
for key, value in kwargs.items():
if key in request:
request[key] = value
else:
raise KeyError('Unknown key in **kwargs ERROR')
# send dict to server
self._send_request(request)
# return server reply
return self._pull_reply()
All examples will be on Python 3. Lets create an instance of MetaTrader API class:
api = MTraderAPI()
First of all we should configure the script symbol
and timeframe
. Live data stream will be configured to the same params. You can use any number of symbols
and timeframes
. The server subscribes to these sembols and will transmit them through the Live data
socket
print(api.construct_and_send(action="CONFIG", symbol="EURUSD", chartTF="M5"))
print(api.construct_and_send(action="CONFIG", symbol="AUDUSD", chartTF="M1"))
...
There is also tick
data. You can subscribe for tick
and candle
data at the same symbol
.
print(api.construct_and_send(action="CONFIG", symbol="EURUSD", chartTF="TICK"))
print(api.construct_and_send(action="CONFIG", symbol="EURUSD", chartTF="M1"))
If you want to stop Live data
, you should reset server subscriptions.
rep = api.construct_and_send(action="RESET")
print(rep)
Get information about the trading account.
rep = api.construct_and_send(action="ACCOUNT")
print(rep)
Get historical data. fromDate
should be in timestamp format. The data will be loaded to the last candle if toDate
is None
. Notice, that the script sends the last unclosed candle too. You should delete it manually.
There are some issues:
- MetaTrader keeps historical data in cache. But when you make a request for the first time, MetaTrader downloads the data from a broker. This operation can exceed
Data socket
timeout. It depends on your broker. Second request will be handeled quickly. - It takes 6-7 seconds to process
50000
M1 candles. It was tested on Windows 10 in Parallels Desktop container with 4 cores and 4GB RAM. So if you need more data there are three ways to handle it. 1) IncreaseData socket
timeout. 2) You can load data partially usingfromDate
andtoDate
. 3) You can use more powerfull hardware.
rep = api.construct_and_send(action="HISTORY", actionType="DATA", symbol="EURUSD", chartTF="M5", fromDate=1555555555)
print(rep)
History data reply example:
{'data': [[1560782340, 1.12271, 1.12288, 1.12269, 1.12277, 46.0],[1560782400, 1.12278, 1.12299, 1.12276, 1.12297, 43.0],[1560782460, 1.12296, 1.12302, 1.12293, 1.123, 23.0]]}
Buy market order.
rep = api.construct_and_send(action="TRADE", actionType="ORDER_TYPE_BUY", symbol="EURUSD", "volume"=0.1, "stoploss"=1.1, "takeprofit"=1.3)
print(rep)
Sell limit order. Remember to switch SL/TP depending on BUY/SELL, or you will get invalid stops
error.
- BUY: SL < price < TP
- SELL: SL > price > TP
rep = api.construct_and_send(action="TRADE", actionType="ORDER_TYPE_SELL_LIMIT", symbol="EURUSD", "volume"=0.1, "price"=1.2, "stoploss"=1.3, "takeprofit"=1.1)
print(rep)
All pending orders are set to Good till cancel
by default. If you want to set an expiration date, pass the date in timestamp format to expiration
param.
rep = api.construct_and_send(action="TRADE", actionType="ORDER_TYPE_SELL_LIMIT", symbol="EURUSD", "volume"=0.1, "price"=1.2, "expiration"=1560782460)
print(rep)
Event handler example for Live socket
and Data socket
.
import zmq
import threading
api = MTraderAPI()
def _t_livedata():
socket = api.live_socket()
while True:
try:
last_candle = socket.recv_json()
except zmq.ZMQError:
raise zmq.NotDone("Live data ERROR")
print(last_candle)
def _t_streaming_events():
socket = api.streaming_socket()
while True:
try:
trans = socket.recv_json()
request, reply = trans.values()
except zmq.ZMQError:
raise zmq.NotDone("Streaming data ERROR")
print(request)
print(reply)
t = threading.Thread(target=_t_livedata, daemon=True)
t.start()
t = threading.Thread(target=_t_streaming_events, daemon=True)
t.start()
while True:
pass
There are only two variants of Live socket
data. When everything is ok, the script sends subscribed data on new even. You can divide streams by symbol and timeframe names:
{"status":"CONNECTED","symbol":"EURUSD","timeframe":"TICK","data":[1581611172734,1.08515,1.08521]}
{"status":"CONNECTED","symbol":"EURUSD","timeframe":"M1","data":[1581611100,1.08525,1.08525,1.08520,1.08520,10.00000]}
If the terminal has lost connection to the market:
{"status":"DISCONNECTED"}
When the terminal reconnects to the market, it sends the last closed candle again. So you should update your historical data. Make the action="HISTORY"
request with fromDate
equal to your last candle timestamp before disconnect.
OnTradeTransaction
function is called when a trade transaction event occurs. Streaming socket
sends TRADE_TRANSACTION_REQUEST
data every time it happens. Try to create and modify orders in the MQL5 terminal manually and check the expert logging tab for better understanding. Also see MQL5 docs.
TRADE_TRANSACTION_REQUEST
request data:
{
'action': 'TRADE_ACTION_DEAL',
'order': 501700843,
'symbol': 'EURUSD',
'volume': 0.1,
'price': 1.12181,
'stoplimit': 0.0,
'sl': 1.1,
'tp': 1.13,
'deviation': 10,
'type': 'ORDER_TYPE_BUY',
'type_filling': 'ORDER_FILLING_FOK',
'type_time': 'ORDER_TIME_GTC',
'expiration': 0,
'comment': None,
'position': 0,
'position_by': 0
}
TRADE_TRANSACTION_REQUEST
result data:
{
'retcode': 10009,
'result': 'TRADE_RETCODE_DONE',
'deal': 501700843,
'order': 501700843,
'volume': 0.1,
'price': 1.12181,
'comment': None,
'request_id': 8,
'retcode_external': 0
}
First of all, when you send a command via System socket
, you should always receive back "OK"
message via System socket
. It means that your command was received and deserialized.
All data that come through Data socket
have an error
param. This param will have true
key if somethng goes wrong. Also, there will be description
and function
params. They will hold information about error and the name of the function with error.
This information also applies to the trade commannds. See MQL5 docs for possible server answers.
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See LICENSE
for more information.