资源加载中... loading...

Market quotes collector upgrade again

Author: , Created: 2020-05-26 14:25:15, Updated: 2023-11-02 19:53:34

img

Supporting CSV format file import to provide custom data source

Recently, a trader needs to use his own CSV format file as a data source for FMZ platform backtest system. our platform’s backtest system has many functions and is simple and efficient to use, so that as long as users have their own data, they can perform backtesting according to these data, which is no longer limited to the exchanges and varieties supported by our platform data center.

Design ideas

The design idea is actually very simple. We only need to change it slightly based on the previous market collector. We add a parameter isOnlySupportCSV to the market collector to control whether only the CSV file is used as the data source for the backtest system. The parameter filePathForCSV is used to set the path of the CSV data file placed on the server where the market collector robot runs. At last, it is based on whether the isOnlySupportCSV parameter is set to True to decide which data source to use (collected by yourself or the data in the CSV file), this change is mainly in the do_GET function of the Provider class.

What is a CSV file?

Comma-separated values, also known as CSV, sometimes referred to as character-separated values, because the separator character can also not be a comma. Its file stores the table data (numbers and text) in plain text. Plain text means that the file is a sequence of characters and contains no data that must be interpreted like a binary number. The CSV file consists of any number of records, separated by some newline character; each record is composed of fields, and the separators between fields are other characters or strings, and the most common are commas or tabs. Generally, all records have the exact same sequence of fields. They are usually plain text files. It is recommended to use WORDPAD or Excel to open.

The general standard of the CSV file format does not exist, but there are certain rules, generally one record per line, and the first line is the header. The data in each row is separated by commas.

For example, the CSV file we used for testing is opened with Notepad like this:

img

Observed that the first line of the CSV file is the table header.

,open,high,low,close,vol

We just need to parse and sort out these data, and then construct it into the format required by the custom data source of the backtest system. This code in our previous article has already been processed, and only needs to be modified slightly.

Modified code

import _thread
import pymongo
import json
import math
import csv
from http.server import HTTPServer, BaseHTTPRequestHandler
from urllib.parse import parse_qs, urlparse

def url2Dict(url):
    query = urlparse(url).query  
    params = parse_qs(query)  
    result = {key: params[key][0] for key in params}  
    return result

class Provider(BaseHTTPRequestHandler):
    def do_GET(self):
        global isOnlySupportCSV, filePathForCSV
        try:
            self.send_response(200)
            self.send_header("Content-type", "application/json")
            self.end_headers()

            dictParam = url2Dict(self.path)
            Log("The custom data source service receives the request,self.path:", self.path, "query parameter:", dictParam)
            
            # At present, the backtest system can only select the exchange name from the list. When adding a custom data source, set it to Binance, that is: Binance
            exName = exchange.GetName()                                     
            # Note that period is the bottom K-line period
            tabName = "%s_%s" % ("records", int(int(dictParam["period"]) / 1000))  
            priceRatio = math.pow(10, int(dictParam["round"]))
            amountRatio = math.pow(10, int(dictParam["vround"]))
            fromTS = int(dictParam["from"]) * int(1000)
            toTS = int(dictParam["to"]) * int(1000)

            # Request data
            data = {
                "schema" : ["time", "open", "high", "low", "close", "vol"],
                "data" : []
            }
            
            if isOnlySupportCSV:
                # Handle CSV reading, filePathForCSV path
                listDataSequence = []
                with open(filePathForCSV, "r") as f:
                    reader = csv.reader(f)
                    # Get table header
                    header = next(reader)
                    headerIsNoneCount = 0
                    if len(header) != len(data["schema"]):
                        Log("The CSV file format is wrong, the number of columns is different, please check!", "#FF0000")
                        return 
                    for ele in header:
                        for i in range(len(data["schema"])):
                            if data["schema"][i] == ele or ele == "":
                                if ele == "":
                                    headerIsNoneCount += 1
                                if headerIsNoneCount > 1:
                                    Log("The CSV file format is incorrect, please check!", "#FF0000")
                                    return 
                                listDataSequence.append(i)
                                break
                    
                    # Read content
                    while True:
                        record = next(reader, -1)
                        if record == -1:
                            break
                        index = 0
                        arr = [0, 0, 0, 0, 0, 0]
                        for ele in record:
                            arr[listDataSequence[index]] = int(ele) if listDataSequence[index] == 0 else (int(float(ele) * amountRatio) if listDataSequence[index] == 5 else int(float(ele) * priceRatio))
                            index += 1
                        data["data"].append(arr)
                
                Log("data: ", data, "Respond to backtest system requests.")
                self.wfile.write(json.dumps(data).encode())
                return 
            
            # Connect to the database
            Log("Connect to the database service to obtain data, the database: ", exName, "table: ", tabName)
            myDBClient = pymongo.MongoClient("mongodb://localhost:27017")
            ex_DB = myDBClient[exName]
            exRecords = ex_DB[tabName]
            
            # Construct query conditions: greater than a certain value {'age': {'$ gt': 20}} less than a certain value {'age': {'$lt': 20}}
            dbQuery = {"$and":[{'Time': {'$gt': fromTS}}, {'Time': {'$lt': toTS}}]}
            Log("Query conditions: ", dbQuery, "Number of inquiries: ", exRecords.find(dbQuery).count(), "Total number of databases: ", exRecords.find().count())
            
            for x in exRecords.find(dbQuery).sort("Time"):
                # Need to process data accuracy according to request parameters round and vround
                bar = [x["Time"], int(x["Open"] * priceRatio), int(x["High"] * priceRatio), int(x["Low"] * priceRatio), int(x["Close"] * priceRatio), int(x["Volume"] * amountRatio)]
                data["data"].append(bar)
            
            Log("data: ", data, "Respond to backtest system requests.")
            # Write data response
            self.wfile.write(json.dumps(data).encode())
        except BaseException as e:
            Log("Provider do_GET error, e:", e)


def createServer(host):
    try:
        server = HTTPServer(host, Provider)
        Log("Starting server, listen at: %s:%s" % host)
        server.serve_forever()
    except BaseException as e:
        Log("createServer error, e:", e)
        raise Exception("stop")

def main():
    LogReset(1)
    if (isOnlySupportCSV):
        try:
        # _thread.start_new_thread(createServer, (("localhost", 9090), ))         # local test
            _thread.start_new_thread(createServer, (("0.0.0.0", 9090), ))         # Test on VPS server
            Log("Start the custom data source service thread, and the data is provided by the CSV file. ", "#FF0000")
        except BaseException as e:
            Log("Failed to start the custom data source service!")
            Log("Error message: ", e)
            raise Exception("stop")
        while True:
            LogStatus(_D(), "Only start the custom data source service, do not collect data!")
            Sleep(2000)
    
    exName = exchange.GetName()
    period = exchange.GetPeriod()
    Log("collect", exName, "Exchange K-line data,", "K line cycle:", period, "Second")
    
    # Connect to the database service, service address mongodb: //127.0.0.1: 27017 See the settings of mongodb installed on the server
    Log("Connect to the mongodb service of the hosting device, mongodb://localhost:27017")
    myDBClient = pymongo.MongoClient("mongodb://localhost:27017")   
    # Create a database
    ex_DB = myDBClient[exName]
    
    # Print the current database table
    collist = ex_DB.list_collection_names()
    Log("mongodb", exName, "collist:", collist)
    
    # Check if the table is deleted
    arrDropNames = json.loads(dropNames)
    if isinstance(arrDropNames, list):
        for i in range(len(arrDropNames)):
            dropName = arrDropNames[i]
            if isinstance(dropName, str):
                if not dropName in collist:
                    continue
                tab = ex_DB[dropName]
                Log("dropName:", dropName, "delete:", dropName)
                ret = tab.drop()
                collist = ex_DB.list_collection_names()
                if dropName in collist:
                    Log(dropName, "failed to delete")
                else :
                    Log(dropName, "successfully deleted")
    
    # Start a thread to provide a custom data source service
    try:
        # _thread.start_new_thread(createServer, (("localhost", 9090), ))     # local test
        _thread.start_new_thread(createServer, (("0.0.0.0", 9090), ))         # Test on VPS server
        Log("Open the custom data source service thread", "#FF0000")
    except BaseException as e:
        Log("Failed to start the custom data source service!")
        Log("Error message:", e)
        raise Exception("stop")
    
    # Create the records table
    ex_DB_Records = ex_DB["%s_%d" % ("records", period)]
    Log("Start collecting", exName, "K-line data", "cycle:", period, "Open (create) the database table:", "%s_%d" % ("records", period), "#FF0000")
    preBarTime = 0
    index = 1
    while True:
        r = _C(exchange.GetRecords)
        if len(r) < 2:
            Sleep(1000)
            continue
        if preBarTime == 0:
            # Write all BAR data for the first time
            for i in range(len(r) - 1):
                bar = r[i]
                # Write root by root, you need to determine whether the data already exists in the current database table, based on timestamp detection, if there is the data, then skip, if not write
                retQuery = ex_DB_Records.find({"Time": bar["Time"]})
                if retQuery.count() > 0:
                    continue
                
                # Write bar to the database table
                ex_DB_Records.insert_one({"High": bar["High"], "Low": bar["Low"], "Open": bar["Open"], "Close": bar["Close"], "Time": bar["Time"], "Volume": bar["Volume"]})                
                index += 1
            preBarTime = r[-1]["Time"]
        elif preBarTime != r[-1]["Time"]:
            bar = r[-2]
            # Check before writing data, whether the data already exists, based on time stamp detection
            retQuery = ex_DB_Records.find({"Time": bar["Time"]})
            if retQuery.count() > 0:
                continue
            
            ex_DB_Records.insert_one({"High": bar["High"], "Low": bar["Low"], "Open": bar["Open"], "Close": bar["Close"], "Time": bar["Time"], "Volume": bar["Volume"]})
            index += 1
            preBarTime = r[-1]["Time"]
        LogStatus(_D(), "preBarTime:", preBarTime, "_D(preBarTime):", _D(preBarTime/1000), "index:", index)
        # Increase drawing display
        ext.PlotRecords(r, "%s_%d" % ("records", period))
        Sleep(10000)

Run test

First, we start the market collector robot. We add an exchange to the robot and let the robot run.

Parameter configuration:

img img

Then we create a test strategy:

function main() {
    Log(exchange.GetRecords())
    Log(exchange.GetRecords())
    Log(exchange.GetRecords())
}

The strategy is very simple, only obtain and print K-line data three times.

On backtest page, set the data source of the backtest system as a custom data source, and fill in the address of the server where the market collector robot runs. Since the data in our CSV file is a 1-minute K line. So when backtest, we set the K-line period to 1 minute.

img

Click to start the backtest, and the market collector robot receives the data request:

img

After the execution strategy of the backtest system is completed, a K-line chart is generated based on the K-line data in the data source.

img

Compare the data in the file:

img img

Market quotes collector upgrade again


Related

More