Teach you to write a K-line synthesis function in the Python version

Teach you to write a K-line synthesis function in the Python version

·

4 min read

When writing and using strategies, we often use some rarely used K-line period data. However, exchanges and data sources do not provide data on these periods. It can only be synthesized by using data with an existing period. The synthesized algorithm already has a JavaScript version (link). In fact, it is easy to transplant a piece of JavaScript code to Python. Next, let's write a Python version of the K-line synthesis algorithm.

JavaScript version

function GetNewCycleRecords (sourceRecords, targetCycle) {    // K-line synthesis function
      var ret = []

      // Obtain the period of the source K-line data first
      if (!sourceRecords || sourceRecords.length < 2) {
          return null
      }
      var sourceLen = sourceRecords.length
      var sourceCycle = sourceRecords[sourceLen - 1].Time - sourceRecords[sourceLen - 2].Time

      if (targetCycle % sourceCycle != 0) {
          Log("targetCycle:", targetCycle)
          Log("sourceCycle:", sourceCycle)
          throw "targetCycle is not an integral multiple of sourceCycle."
      }

      if ((1000 * 60 * 60) % targetCycle != 0 && (1000 * 60 * 60 * 24) % targetCycle != 0) {
          Log("targetCycle:", targetCycle)
          Log("sourceCycle:", sourceCycle)
          Log((1000 * 60 * 60) % targetCycle, (1000 * 60 * 60 * 24) % targetCycle)
          throw "targetCycle cannot complete the cycle."
      }

      var multiple = targetCycle / sourceCycle


      var isBegin = false 
      var count = 0
      var high = 0 
      var low = 0 
      var open = 0
      var close = 0 
      var time = 0
      var vol = 0
      for (var i = 0 ; i < sourceLen ; i++) {
          // Get the time zone offset value
          var d = new Date()
          var n = d.getTimezoneOffset()

          if (((1000 * 60 * 60 * 24) - sourceRecords[i].Time % (1000 * 60 * 60 * 24) + (n * 1000 * 60)) % targetCycle == 0) {
              isBegin = true
          }

          if (isBegin) {
              if (count == 0) {
                  high = sourceRecords[i].High
                  low = sourceRecords[i].Low
                  open = sourceRecords[i].Open
                  close = sourceRecords[i].Close
                  time = sourceRecords[i].Time
                  vol = sourceRecords[i].Volume

                  count++
              } else if (count < multiple) {
                  high = Math.max(high, sourceRecords[i].High)
                  low = Math.min(low, sourceRecords[i].Low)
                  close = sourceRecords[i].Close
                  vol += sourceRecords[i].Volume

                  count++
              }

              if (count == multiple || i == sourceLen - 1) {
                  ret.push({
                      High : high,
                      Low : low,
                      Open : open,
                      Close : close,
                      Time : time,
                      Volume : vol,
                  })
                  count = 0
              }
          }
      }

      return ret 
  }

There are JavaScript algorithms. Python can be translated and transplanted line by line. If you encounter JavaScript built-in functions or inherent methods, you can go to Python to find the corresponding methods. Therefore, the migration is easy.
The algorithm logic is exactly the same, except that JavaScript function calls var n=d.getTimezoneOffset(). When migrating to Python,n=time.altzone in Python's time library is used instead. Other differences are only in terms of language grammar (such as the use of for loops, Boolean values, logical AND, logical NOT, logical OR, etc.).

Migrated Python code:

import time

def GetNewCycleRecords(sourceRecords, targetCycle):
    ret = []

    # Obtain the period of the source K-line data first
    if not sourceRecords or len(sourceRecords) < 2 : 
        return None

    sourceLen = len(sourceRecords)
    sourceCycle = sourceRecords[-1]["Time"] - sourceRecords[-2]["Time"]

    if targetCycle % sourceCycle != 0 :
        Log("targetCycle:", targetCycle)
        Log("sourceCycle:", sourceCycle)
        raise "targetCycle is not an integral multiple of sourceCycle."

    if (1000 * 60 * 60) % targetCycle != 0 and (1000 * 60 * 60 * 24) % targetCycle != 0 : 
        Log("targetCycle:", targetCycle)
        Log("sourceCycle:", sourceCycle)
        Log((1000 * 60 * 60) % targetCycle, (1000 * 60 * 60 * 24) % targetCycle)
        raise "targetCycle cannot complete the cycle."

    multiple = targetCycle / sourceCycle

    isBegin = False
    count = 0 
    barHigh = 0 
    barLow = 0 
    barOpen = 0
    barClose = 0 
    barTime = 0 
    barVol = 0 

    for i in range(sourceLen) : 
        # Get the time zone offset value
        n = time.altzone        

        if ((1000 * 60 * 60 * 24) - (sourceRecords[i]["Time"] * 1000) % (1000 * 60 * 60 * 24) + (n * 1000)) % targetCycle == 0 :
            isBegin = True

        if isBegin : 
            if count == 0 : 
                barHigh = sourceRecords[i]["High"]
                barLow = sourceRecords[i]["Low"]
                barOpen = sourceRecords[i]["Open"]
                barClose = sourceRecords[i]["Close"]
                barTime = sourceRecords[i]["Time"]
                barVol = sourceRecords[i]["Volume"]
                count += 1
            elif count < multiple : 
                barHigh = max(barHigh, sourceRecords[i]["High"])
                barLow = min(barLow, sourceRecords[i]["Low"])
                barClose = sourceRecords[i]["Close"]
                barVol += sourceRecords[i]["Volume"]
                count += 1

            if count == multiple or i == sourceLen - 1 :
                ret.append({
                    "High" : barHigh,
                    "Low" : barLow,
                    "Open" : barOpen,
                    "Close" : barClose,
                    "Time" : barTime,
                    "Volume" : barVol,
                })
                count = 0

    return ret 

# Test
def main():
    while True:
        r = exchange.GetRecords()
        r2 = GetNewCycleRecords(r, 1000 * 60 * 60 * 4)      

        ext.PlotRecords(r2, "r2")                                 
        Sleep(1000)

Test

Huobi market chart

4-hour chart of backtest synthesis

The above code is for reference only. If it is used in specific strategies, please modify and test according to the specific requirements.
If there is a bug or improvement suggestion, please leave a message. Thank you very much. o^_^ o

From: https://blog.mathquant.com/2022/12/26/teach-you-to-write-a-k-line-synthesis-function-in-the-python-version.html