Extends bigmemory's big.matrix to support big vectors of character strings.
Current package build status via use of
Travis CI:
This package extends bigmemory's big.matrix objects to support big vectors of character strings. The only constraint is the user-specified maximum character length which must be satisfied for every string in the vector. Shorter strings are padded with NAs, consuming (and wasting) space, but this is unavoidable given the use of memory-mapped files.
This package is very much under development. The basic functionality
seems fairly solid. The implementation may not be as efficient as
it could be. Not all usage errors may be trapped or handled
in a friendly or appropriate manner.
For example, if a big.char vector, x has length 5, then
x[6] generates an error (this would not be the case with a regular
vector of strings). Please feel free to email me with problems,
questions, or requests.
Known issues to be resolved:
- index recycling (bug in bigmemory, fix there)
- indices out of bounds (needs fixing in big.char for consistency with R's default behavior/behaviour)
The next to last command triggers a warning because the string is truncated to three (3) characters.
> require(devtools)
> install_github('big.char', 'jayemerson')
> library(big.char)
>
> x <- big.char(5, 3, init="ABC")
> x[]
[1] "ABC" "ABC" "ABC" "ABC" "ABC"
> x[1] <- ""
> x[-1] <- c(NA, "*", "--", "ABCD")
Warning message:
In `[<-`(`*tmp*`, -1, value = c(NA, "*", "--", "DEFG")) :
Long string(s) truncated to maxchar characters
> x[]
[1] "" NA "*" "--" "DEF"
The big.char package is licensed under LGPL-3.
After creating the public repository on GitHib, essentially empty
with only README.md, I cloned to my laptop:
git clone https://github.com/jayemerson/big.char.git
I then moved the package contents into the local big.char directory,
at which point the following workflow appears sufficient for my own
purposes:
git status
git commit -am "Short commit message here"
git push origin