有时我问了以下问题:
我有一个交易日和市场价值的交易清单。每个(交易)日都有新头寸进入列表,但旧头寸永远不会消失(当头寸到期时,价值保持不变)。该列表如下所示:
Deal Trade_Date MktValue Desired_Col Deal1 31.08.2012 10 +10 Deal2 31.08.2012 21 +21 Deal1 03.09.2012 12 +2 Deal2 03.09.2012 19 -2 Deal3 03.09.2012 2 +2
我希望每笔交易都能获得与前一个交易日期的差额(上例中的 Desidered_Col)。
Roland向我提供了以下解决方案:
df <- read.table(text="Deal Trade_Date MktValue Desidered_Col Deal1 31.08.2012 10 +10 Deal2 31.08.2012 21 +21 Deal1 03.09.2012 12 +2 Deal2 03.09.2012 19 -2 Deal3 03.09.2012 2 +2" ,标题=真)
库(data.table)dt <- as.data.table(df)
diff.padded <- function(x) c(x[1],diff(x)) dt[,Desidered_Col2:=diff.padded(MktValue),by=Deal]
Deal Trade_Date MktValue Desired_Col Desired_Col2 1: Deal1 31.08.2012 10 10 10 2: Deal2 31.08.2012 21 21 21 3: Deal1 03.09.2012 12 2 2 4: Deal2 03.09.2012 19 -2 -2 5: Deal3 03.09.2012 2 2 2
该解决方案与 data.table 完美配合。但是考虑到我的桌子的大小,我决定尝试使用 ffdf 对象。因此,我现在将数据保存在 ffdf 文件中,不幸的是,我试图重现相同的解决方案,但没有成功。你有什么建议我可以在 ffdf 中重现它吗?谢谢你的帮助。
这是我正在运行的完整代码:
# Load needed packages
library(RODBC)
library(data.table)
library(ETLUtils)
library(RSQLite)
library(ffbase)
calendar <- read.csv("Trading_Calendar.csv",sep=";",stringsAsFactors=FALSE)
calendar$STICHTAG <- as.Date(calendar$STICHTAG,"%d.%m.%Y")
ST_a=Sys.Date()-2
rd_a=as.Date("13.11.2012","%d.%m.%Y")
ST=paste("'",as.character(format(ST_a,"%d.%m.%Y")),"'",sep="")
rd=paste("'",as.character(format(rd_a,"%d.%m.%Y")),"'",sep="")
gc(TRUE)
st.strom <- calendar[calendar$STICHTAG>=rd_a & calendar$STICHTAG<=ST_a & calendar$BR_Strom==1,"STICHTAG"]
st.strom <- format(st.strom,"%d.%m.%Y")
st.strom.s <- paste("('",do.call(paste, c(as.list(as.character(st.strom)), sep="','")),"')",sep="")
started.at=proc.time()
Sys.sleep(1)
memory.limit(size=4095)
query <- paste("select * from is_bewertung_data where commodity in ('CASH','COAL','CO2','ELEC','GCERT')
and stichtag in ",st.strom.s,sep="")
deals.strom <- read.odbc.ffdf(query = query,odbcConnect.args=list(dsn="dsn",uid="id",pwd="pwd"),
first.rows = 100000, next.rows = 500000, VERBOSE=TRUE)
result <- ffdfdply(deals.strom, deals.strom$DEALID, FUN=function(x){
x <- split(x, x$DEALID)
x <- lapply(x, FUN=function(onlyonedeal){
onlyonedeal$Desidered_Col2 <- c(NA, -diff(onlyonedeal$STICHTAG))
onlyonedeal
})
x <- do.call(rbind, x)
x
})
cat("Finished in",timetaken(started.at),"\n")
这里是 str(deals.strom[1:5,]) 的结果:
'data.frame': 5 obs. of 39 variables:
$ ABBREVIATION : Factor w/ 33553 levels " C 251"," TÜV EE Donaustrom",..: 1893 1892 1894 1895 1896
$ TRADEDATE : POSIXct, format: "2007-06-19" "2007-06-19" "2007-06-19" ...
$ BOOK : Factor w/ 30 levels "CR_RIR_RISKRED",..: 10 10 10 10 10
$ CONTRACT : Factor w/ 20 levels "Base","DNULL",..: 1 5 5 1 1
$ BUYSELL : Factor w/ 2 levels "BUY","SELL": 2 1 2 1 1
$ RATE : num 54.2 57.2 57.3 54.2 55.1
$ AMOUNT : num 474792 501072 501773 474792 964476
$ CUR : Factor w/ 2 levels "EUR","USD": 1 1 1 1 1
$ VOLUME : num 8760 8760 8760 8760 17520
$ UNIT : Factor w/ 2 levels "MWH","t": 1 1 1 1 1
$ STARTDATE : POSIXct, format: "2010-01-01" "2010-01-01" "2010-01-01" ...
$ ENDDATE : POSIXct, format: "2011-01-01" "2011-01-01" "2011-01-01" ...
$ BROKERAGE : num 0 0 0 0 175
$ DV : num 85078 -98218 98919 -85078 -185048
$ REALIZED : num 85078 -98218 98919 -85078 -185048
$ PV : num 0 0 0 0 0
$ DV_DAY : num 0 0 0 0 0
$ DV_MONTH : num 0 0 0 0 0
$ DV_YEAR : num 0 0 0 0 0
$ TRADER : Factor w/ 16 levels "Adolf Plentz",..: 7 7 7 7 12
$ ACTIVE : Factor w/ 2 levels "LONGTERM","SHORTTERM": 2 2 2 2 2
$ STATUS : Factor w/ 2 levels "GCPTY","INT": 1 1 2 2 1
$ PV_MIN : num 0 0 0 0 0
$ PV_PLUS : num 0 0 0 0 0
$ VERTRAGSPARTY : Factor w/ 21 levels "EDL_G059","EDL_G097",..: 10 10 3 3 10
$ GESELLSCHAFT : Factor w/ 1 level "24/7 Trading": 1 1 1 1 1
$ COMMODITY : Factor w/ 5 levels "CASH","CO2","COAL",..: 4 4 4 4 4
$ TO_BE_DELIVERED: num 0 0 0 0 0
$ ACCOUNT : Factor w/ 8 levels "CR_RISKRED","HO_COAL",..: 5 5 5 5 5
$ VERW_PREIS : num 0 0 0 0 0
$ PV_ND : num 0 0 0 0 0
$ BILANZIERUNG : Factor w/ 2 levels "JA","NEIN": 1 1 1 1 1
$ MOTIV : Factor w/ 8 levels "Emissionszertifikate",..: 4 4 4 4 4
$ STICHTAG : POSIXct, format: "2012-11-13" "2012-11-13" "2012-11-13" ...
$ DEALID : Factor w/ 59704 levels "FUX.E.EEX.K.20090622.002",..: 7175 7103 12584 12500 17985
$ COUNTERPARTY : Factor w/ 174 levels "24sieben GmbH",..: 171 171 53 53 141
$ COMMODITY2 : Factor w/ 8 levels "CASH","CER","COAL",..: 4 4 4 4 4
$ MARKTGEBIET : Factor w/ 3 levels "Kohle","Strom",..: 2 2 2 2 2
$ INSTRUMENT : Factor w/ 88 levels "-","Elektrizität FUX EEX Base Apr11 EEXFUT",..: 1 1 1 1 1
Jan提示后我的解决方案不起作用:
test <- as.ffdf(deals.strom[,c("DEALID","STICHTAG","PV")])
test <- transform(test,chg=c(NA,diff(PV)),chg2=c(NA,-diff(PV)))
fdd <- as.ff(!duplicated(test$DEALID))
test[fdd,c("chg","chg2")] <- test[fdd,"PV"]
我收到以下错误消息:error: is.null(rownames(x)) is not TRUE。不知何故,我无法将 ffdf 子集化。