Re: [R] efficiency in merging two data frames

From: Petr Pikal <>
Date: Tue 02 May 2006 - 17:55:10 EST


and you would better to give up thinking in "for cycles" as a good tool for data manipulation and calculation.

R has usually functions for direct manipulation and calculation with vectors or matrices, vhich are far more effective than calculation in cycle.

I also recommend you to go through Paul Johnsons Rtips which is set of many good ideas how to do some things.


On 1 May 2006 at 7:54, Steve Miller wrote:

From:           	"Steve Miller" <>
To:             	"'Guojun Zhu'" <>,
Date sent:      	Mon, 1 May 2006 07:54:54 -0500
Subject:        	Re: [R] efficiency in merging two data frames

> I'm sure you'll get ingenious responses to help you optimize your R
> code. I deal with similar investment data in even larger numbers (e.g.
> 10 years of daily return data for each stock in the Russell 3000), and
> prefer reading and consolidating the data in Python using dictionaries
> and lists, then either piping the data to R in a read statement
> (read.table("pipe python...")) or using Rpy to write R data frames
> directly from Python. Python is more facile with these basic data
> manipulations for hundreds of thousands or even millions of records,
> and performance is generally considerably better.
> Steve Miller
> -----Original Message-----
> From:
> [] On Behalf Of Guojun Zhu
> Sent: Monday, May 01, 2006 2:35 AM To:
> Subject: [R] efficiency in merging two data frames
> I have two data sets about lots of companies' stock
> and fiscal data. One is monthly data with about
> 144,000 lines, and the other is quaterly with about
> 56,000. Each data set takes different company code.
> I need to merge these two together. I read both ask
> cvs. And the other file with corresponding firm code.
> Now I have three data sets. return$PERMNO,
> account$GVKEY. id is the data frames of the
> corresponding relation and has both id$PERMNO and
> id$GVKEY. Also, I need to convert the return's month
> into quarter and finally merge two data frames(return
> and account). I end up write a short program for
> this, but it runs very slow. 15+ minutes. Is there
> quick way to do it. Here is my original codes.
> id$fy=rep(0,length(id$PERMNO))
> for (i in 1:length(id$PERMNO))
> id$fy[[i]]<-account$FYR[id$GVKEY[[i]]==account$GVKEY][[1]]
> return$GVKEY=rep(0,length(return$PERMNO))
> return$fyy=rep(0,length(return$PERMNO))
> return$fyq=rep(0,length(return$PERMNO))
> for (i in i:length(return$PERMNO)) {
> temp<-id$PERMNO==return$PERMNO[[i]];
> tempmon<-id$fy[temp][[1]];
> if (return$month[[i]]<-tempmon) {
> return$fyy[[i]]<-return$year[[i]];
> return$fyq[[i]]<-4-(tempmon-return$month[[i]])%/%3;
> }
> else{
> return$fyy[[i]]<-return$year[[i]]+1;
> return$fyq[[i]]<-(return$month[[i]]-tempmon-1)%/%3;
> }
> return$GVKEY[[i]]<-id$GVKEY[temp][[1]];
> }
> returnnew=merge(return,account,by.x<-c("GVKEY","fyy","fyq"),by.y<-c("G
> VKEY", "fyy","fyq"))
> ______________________________________________
> mailing list
> PLEASE do read the posting guide!
> ______________________________________________
> mailing list
> PLEASE do read the posting guide!

Petr Pikal mailing list PLEASE do read the posting guide! Received on Tue May 02 20:58:41 2006

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.1.8, at Tue 02 May 2006 - 22:09:57 EST.

Mailing list information is available at Please read the posting guide before posting to the list.