Question

I am looking for some ideas/functions to improve a task I coded in a quite inefficient way.

My originial data frame look like:

        CUSIP       Date        Pclose  TRI
1       30161N101   2011-01-03  38.8581 2011-01-03
2       06738G878   2011-01-03  48.4040 2011-01-03
3       74339G101   2011-01-03  24.0880 2011-01-03
4       74348A590   2011-01-03  81.7200 2011-01-03
5       26922W109   2011-01-03  87.8700 2011-01-03
...

The column ‘TRI’ contains the Date in a date format.

And what I want to obtain look like:

    Date        233052109   126650100   251566105   149123101 ...
1   2011-01-03  22.8031     34.3034     11.2645      91.6178
2   2011-01-04  22.6843     34.2740     11.1862      91.1897
3   2011-01-05  22.7933     34.6362     10.9948      91.9779
4   2011-01-06  22.8034     34.2838     11.0470      91.0242
5   2011-01-07  22.6248     34.3034     11.0644      91.2091
.
.
.

In the second data frame each column (except the date one) has the name of a CUSIP from the first data frame and is filled with data from Pclose.

I am making my second data frame with regular loop but I am sure there is a way with compilled function do to way better (perhaps subset)

my functions are:

To build the second data frame:

function(cusippresent,cusiplist){
    workinglist=list()
    workinglist[1]=as.character('Date')
    position = 2
    for (i in 2:length(cusippresent)) {
        if(cusippresent[i] %in% cusiplist) {
            workinglist[position]=as.character(cusippresent[i]);
            position=position+1
        }
    }

    rm(position)

    Data=data.frame()

    #On remplit la première ligne du dataframe
    #avec des éléments du bon type sinon il y a des problèmes

    Data[1,1]=as.character('1987-11-12')

    Data[1,2]=as.numeric(1)
    for (i in 2:length(workinglist)){Data[1,i]=as.numeric(1)}
    for (i in 1:length(workinglist)){colnames(Data)[i]=workinglist[i]}

    return(Data)    
}

To fill the data frame:

function(DATA11C,TCusipC){
    nloop = 1
    positionorigine = 1
    positioncible = 1

    #copy of dates
    datelist = DATA11C[,"Date"]
    datelist = unique(datelist)
    for (i in 1:length(datelist)) {
        TCusipC[i,"Date"]=as.character(datelist[i])
    }

    #creation of needed columns
    TCusipC[,ncol(TCusipC)+1] = as.Date(TCusipC[,"Date"])
    colnames(TCusipC)[ncol(TCusipC)] = 'TRI'

    #ordering of tables
    DATA11C=DATA11C[with(DATA11C,order(TRI)),]
    TCusipC=TCusipC[with(TCusipC,order(TRI)),]

    longueur = nrow(TCusipC)

    #filling of the table 
    while(nloop<longueur) {
        while(DATA11C[positionorigine,"TRI"]==TCusipC[positioncible,"TRI"]){
            nom = as.character(DATA11C[positionorigine,"CUSIP"]);
            TCusipC[positioncible,nom]=as.numeric(DATA11C[positionorigine,"Pclose"]);
            positionorigine=positionorigine+1
        };
        nloop=nloop+1;
        positioncible=positioncible+1
    }

    return(TCusipC)
}

Any suggestion on what function to dig up? potential improvement?

Many thanks,

Vincent

Answer 1

That is exactly what the reshape packages does

library(reshape)
cast(Date ~ CUSIP, data = DATA11C, value = "Pclose")