0

I am a quite new with R programming and now developing a system that is suppose to interact with R. My question is :

How to get quotes from hard coded from the scripts rather than from various sources like “yahoo” “google” etc?

Why do I need quotes hard coded in the script?

I am using Rserve as my downstream system, the main system does fetching of data and will perform other portfolio checks then it makes a call to R-TTR-quantmod packages for calculation of financial numbers. So I don’t want R to refetch those quotes so I want the quotes to be hardcoded and sent from my system to Rserve where it gets executed and results is returned from there. This way my code would rely on standard calculation on R and user can focus on other business logic.

Why am I not using csv file approach ?

I am in a realtime system and file io would take huge time and would slow down my system.

FOR EXAMPLE:

library(quantmod)

library(TTR)

Pull S&P500 index data from Yahoo! Finance

getSymbols("^RIL", from="2000-01-01", to="2008-12-07")

Calculate the RSI indicator

rsi <- RSI(Cl(RIL),2)

So this is what I need:

  • Rather than calling getSymbol I would like to pass the data as a variable in the script.
  • I assume the data may be very large at time or very small at times.
  • So what should I do in this scenario?
Pop
  • 12,135
  • 5
  • 55
  • 68
Prahalad Das
  • 27
  • 6
  • If I understand correctly, you might find `dput` to be of use. It returns the code of an R data structure that you can put into scripts to replicate a data item. – James Aug 07 '12 at 10:30
  • If the data is already there, as an `xts` object, you can use it directly. For instance, if the data is in a variable `x`, you can just use `rsi <- RSI(x,2)` or `rsi <- RSI(Cl(x),2)`. – Vincent Zoonekynd Aug 07 '12 at 12:23

1 Answers1

0

I've just been working on some code to use load/save to cache xts objects. The code is trivial:

getSymbols("^RIL", from="2000-01-01", to="2008-12-07")
save(RIL,file="cache/RIL.rdata")

Then in your other script that does the analysis:

load("cache/RIL.rdata")
rsi <- RSI(Cl(RIL),2)

As an added bonus the .rdata files are gzipped.

Some stats: Exhibit 1

My previous approach had 1 minute bars, as one csv file per day. 303 files, summing to 12MB, loaded into a 86,590 row XTS object took 5.64s (elapsed; the user time was 5.34s).

The .rdata file is 2.8MB and takes 0.056s.

Exhibit 2:

First approach: Running RSI freshly on the data (as a pre-cursor to another calculation). Second approach: Have another offline process run RSI on the data, and cache it to an .rdata file.

Loading the RSI file worked out 3 times quicker than calculating the RSI data freshly.

Warning: in my timing tests I used a loop where it is created/saved on pass 1, and then loaded on passes 2 to 20. It was the latter loading I was timing. So it would've been fresh in any OS file caches.

Darren Cook
  • 27,837
  • 13
  • 117
  • 217