What is the most memory efficient and easiest (yes, I know those are sometimes mutually exclusive) way to create an R data frame then save it to an .Rdata file using Java?
Go easy on me though, I'm not a Java developer.
4
-
1stackoverflow.com/questions/4034936/using-r-programming-in-java asked a week ago will probably help.Spacedman– Spacedman2010-10-30 07:04:25 +00:00Commented Oct 30, 2010 at 7:04
-
Does it need to be in .RData form? A csv file would import/save just as well.Gavin Simpson– Gavin Simpson2010-10-31 09:27:44 +00:00Commented Oct 31, 2010 at 9:27
-
Csv is what I've been using. Works fine 99% of the time. Sometimes it gets the column data types wrong.JD Long– JD Long2010-10-31 11:04:14 +00:00Commented Oct 31, 2010 at 11:04
-
Would just using colClasses help?Aaron Statham– Aaron Statham2010-10-31 12:54:48 +00:00Commented Oct 31, 2010 at 12:54
Add a comment
|
3 Answers
How about building a text datafile with structure() and retrieving it with dget()?
data.frame(x= 1:5, y= as.factor(1:5), z= as.character(1:5))
gives the same result as:
structure(list(x = 1:5, y = structure(1:5, .Label = c("1", "2",
"3", "4", "5"), class = "factor"), z = structure(1:5, .Label = c("1",
"2", "3", "4", "5"), class = "factor")), .Names = c("x", "y",
"z"), row.names = c(NA, -5L), class = "data.frame")
It is not memory efficient per se, but you have more control over the data types. From R, you can show a data frame in the above long format by using dput() and retrieve it from a text file with dget(), and it shouldn't take too much parsing to write it from Java.
1 Comment
JD Long
Hey that's a neat idea I had not thought of. I have not used
structure() before so it hadnt crossed my mind. I will play with that. Thanks!It might be a bit of an overkill but rJava/JRI (http://rosuda.org/rJava/) give you a Java API to R. Essentially you get an R process that you can control programmatically from your Java code and obviously you can share data and create a .RData file through R calls.
Comments
My first inclination is to throw stuff in MySQL, but the overhead of creating tables, etc. probably doesn't make sense if these files are temporary in nature.
I agree with the others that if you want to run R from Java, rJava is the way to go, but this solution seems a little clumsy.
Along the lines of the simplicity of CSV files, but how about using a portable data format like NetCDF http://en.wikipedia.org/wiki/NetCDF instead? They should preserve data formats better and can be accessed from Java ( http://www.unidata.ucar.edu/software/netcdf-java/ ), R ( http://cran.r-project.org/web/packages/RNetCDF/ ), and even GDAL.
(My astro background forces me to mention FITS as an option too.)
