// How to create the CVS containing NAV data + deposit?
// IB: PortfolioAnalyst/Reports/CreateCustomReport (SinceInception, Daily, Detailed + AccountOverview/Allocation by Financial Instrument/Deposits). Create in PDF + CSV.
// If it timeouts, run a Custom date for the last 2-5 years. It can be merged together manually as a last resort.
// >DC-IB-MAIN, it seems: 2011-02-02 is the inception date. 2011-02-02, 2011-03-01: didn't work. Timeout. But 2014-12-31 worked. Try at another time.
public void InsertNavAssetFromCsvFile(string p_redisKeyPrefix, string p_csvFullpath)
{
List <DailyNavData> dailyNavData = new List <DailyNavData>();
List <DailyNavData> dailyDepositData = new List <DailyNavData>();
using (StreamReader sr = new StreamReader(p_csvFullpath))
{
int iNavColumn = -1;
int iRow = 0;
string?currentLine;
while ((currentLine = sr.ReadLine()) != null) // currentLine will be null when the StreamReader reaches the end of file
{
iRow++;
if (iRow == 1 && currentLine != @"Introduction,Header,Name,Account,Alias,BaseCurrency,AccountType,AnalysisPeriod,PerformanceMeasure")
{
throw new Exception();
}
// "Allocation by Financial Instrument,Header,Date,ETFs,Options,Stocks,Cash,NAV" or "Allocation by Financial Instrument,Header,Date,ETFs,Options,Stocks,Warrants,Cash,NAV"
if (iRow == 5)
{
if (!currentLine.StartsWith(@"Allocation by Financial Instrument,Header,Date,ETFs,Options,Stocks"))
{
throw new Exception();
}
var navHeaderParts = currentLine.Split(',', StringSplitOptions.RemoveEmptyEntries);
iNavColumn = Array.FindIndex(navHeaderParts, r => r == "NAV");
}
if (currentLine.StartsWith(@"Allocation by Financial Instrument,Data,"))
{
var currentLineParts = currentLine.Split(',', StringSplitOptions.RemoveEmptyEntries); // date is in this format: "20090102" YYYYMMDD
dailyNavData.Add(new DailyNavData()
{
DateStr = currentLineParts[2], ValueStr = currentLineParts[iNavColumn]
});
}
if (currentLine.StartsWith(@"Deposits And Withdrawals,Data,"))
{
var currentLineParts = currentLine.Split(',', StringSplitOptions.RemoveEmptyEntries); // date is in this format: "03/10/09" MM/DD/YY
if (currentLineParts[3] == "Deposit" || currentLineParts[3] == "Incoming Account Transfer" || currentLineParts[3] == "Withdrawal" || currentLineParts[3] == "Outgoing Account Transfer")
{
var monthStr = currentLineParts[2].Substring(0, 2);
var dayStr = currentLineParts[2].Substring(3, 2);
var yearStr = currentLineParts[2].Substring(6, 2); // 09 means 2009, 99 means 1999
var year = Int32.Parse(yearStr);
if (year > 50)
{
yearStr = (year + 1900).ToString();
}
else
{
yearStr = (year + 2000).ToString();
}
dailyDepositData.Add(new DailyNavData()
{
DateStr = yearStr + monthStr + dayStr, ValueStr = currentLineParts[5]
}); // Withdrawals are negative in CSV. Good.
}
}
}
}
// for 3069 days.
// 1. Text data: with fractional NAV values: 70K, with integer NAV values: 47.5K (brotlied: 9.578K), with integer NAV + DateStr-1900years: 44.4K (brotlied: 9.557K, difference is 0.2%, not even 1%. Just forget the -1900). (Wow. 4x compression)
string outputCsv = "D/C," + String.Join(",", dailyNavData.Select(r => {
int year = Int32.Parse(r.DateStr.Substring(0, 4)); // - 1900; // subtracting -1900years only helps about 1/1000. It is not worth it.
// casting Double to Int will just remove the fractionals, but not round it to the nearest integer.
int nearestIntValue = (int)Math.Round(Double.Parse(r.ValueStr), MidpointRounding.AwayFromZero); // 0.5 is rounded to 1, -0.5 is rounded to -1. Good.
return(year.ToString("D3") + r.DateStr.Substring(4) + "/" + nearestIntValue.ToString());
}));
var outputCsvBrotli = Utils.Str2BrotliBin(outputCsv);
// 2.1 Bin data: 3069*6=18.4K. Brotlied: 15.268K (less compression if date + float are mixed)
var dailyStructsBin = dailyNavData.Select(r => {
DateOnly dateOnly = new DateOnly(Int32.Parse(r.DateStr.Substring(0, 4)), Int32.Parse(r.DateStr.Substring(4, 2)), Int32.Parse(r.DateStr.Substring(6, 2)));
float navValue = (float)Double.Parse(r.ValueStr);
return(new DailyNavDataBin()
{
dateOnly = dateOnly, floatValue = navValue
});
}).ToArray();
var outputBin1 = dailyStructsBin.SelectMany(r => {
byte[] dateBytes = BitConverter.GetBytes(r.dateOnly.ToBinary());
byte[] navBytes = BitConverter.GetBytes(r.floatValue);
IEnumerable <byte> dailyBytes = dateBytes.Concat(navBytes);
return(dailyBytes);
}).ToArray();
var outputBin1Brotli = Utils.Bin2BrotliBin(outputBin1);
// 2.2 Bin data: 3069*6=18.4K. Brotlied: 13.359K (more compression if date array is separate and float array is separate)
var outputBin2 = new byte[dailyNavData.Count * 6];
var dateOnlyArr = dailyNavData.Select(r =>
{
DateOnly dateOnly = new DateOnly(Int32.Parse(r.DateStr.Substring(0, 4)), Int32.Parse(r.DateStr.Substring(4, 2)), Int32.Parse(r.DateStr.Substring(6, 2)));
return(dateOnly.ToBinary());
}).ToArray();
Buffer.BlockCopy(dateOnlyArr, 0, outputBin2, 0, dateOnlyArr.Length * 2); // 'Object must be an array of primitives.'
var navOnlyArr = dailyNavData.Select(r =>
{
float navValue = (float)Double.Parse(r.ValueStr);
return(navValue);
}).ToArray();
Buffer.BlockCopy(navOnlyArr, 0, outputBin2, dailyNavData.Count * 2, navOnlyArr.Length * 4); // 'Object must be an array of primitives.'
var outputBin2Brotli = Utils.Bin2BrotliBin(outputBin2);
string depositCsv = String.Join(",", dailyDepositData.Select(r =>
{
int nearestIntValue = (int)Math.Round(Double.Parse(r.ValueStr), MidpointRounding.AwayFromZero); // 0.5 is rounded to 1, -0.5 is rounded to -1. Good.
return(r.DateStr + "/" + nearestIntValue.ToString());
}));
var depositCsvBrotli = Utils.Str2BrotliBin(depositCsv); // 479 bytes compressed to 179 bytes. For very long term, we can save 1KB per user if compressing. Do it.
var redisConnString = Program.gConfiguration.GetConnectionString("RedisDefault");
IDatabase db = DbCommon.RedisManager.GetDb(redisConnString, 0);
string redisKey = p_redisKeyPrefix + ".brotli";
db.HashSet("assetQuoteRaw", redisKey, RedisValue.CreateFrom(new System.IO.MemoryStream(outputCsvBrotli)));
db.HashSet("assetBrokerNavDeposit", redisKey, RedisValue.CreateFrom(new System.IO.MemoryStream(depositCsvBrotli)));
Console.WriteLine("InsertNavAssetFromCsvFile() Ended. Conclusion: For 11 years of daily Date/float data. Without compression binary (18.4K) is smaller then CSV text (47.5K). But with Brotli binary (13.4K) is 41% bigger then brotli CSV text (9.5K). Because there is not much repeatable pattern in Float data, while a lot of repeatable comma and digits in text data. And Brotli is attacking the limits of theoretical compression possibilites. Conclusion: USE CSV text data with Brotli. Tested: Brotlied CSV is 30% smaller than brotlied binary.");
}