pickleR

pickleR provides for the serializing and de-serializing of an R object structure, similar to that seen in the Python module pickle. pickleR traverses the object hierarchy, keeping track of any sub-objects it has already encountered so that they will not be stored again. Therefore, recursive, nested and shared objects are handled by pickleR.

The unpickling process reconstitutes the object with shared objects and recursions intact.

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("d-sharpe/pickleR")

Example

library(pickleR)
library(lobstr)

valueVector <- rnorm(1000)

## memory address of vector
obj_addr(valueVector)
#> [1] "0x56489ef30c70"

## object memory size
obj_size(valueVector)
#> 8,048 B

listOfValues <-
  list(valueVector,
       valueVector)

## Each element of the listOfValues points to the same memory address of the
## original valueVector

lobstr::sxp(x = listOfValues)
#> [1:0x5648a028a908] <VECSXP[2]> (named:4)
#>   [2:0x56489ef30c70] <REALSXP[1000]> (named:6)
#>   [2:0x56489ef30c70]

## listOfValues memory size
obj_size(listOfValues)
#> 8,112 B

RListOfValues <-
  unserialize(serialize(listOfValues, connection = NULL))

## Each element of the RListOfValues points to a new vector i.e. shared object
## reference were not maintained

lobstr::sxp(x = RListOfValues)
#> [1:0x56489f3cc3d8] <VECSXP[2]> (named:4)
#>   [2:0x5648a0ba9df0] <REALSXP[1000]> (named:1)
#>   [3:0x5648a22e75d0] <REALSXP[1000]> (named:1)

## RListOfValues memory size
obj_size(RListOfValues)
#> 16,160 B

# The serialized object consumes more memory than the original

pickleListOfValues <-
  unpickle(pickle(listOfValues))

## Each element of the pickleListOfValues points to the same memory 
## address i.e. shared object

lobstr::sxp(x = pickleListOfValues)
#> [1:0x5648a16cde98] <VECSXP[2]> (named:10)
#>   [2:0x5648a13840e0] <REALSXP[1000]> (named:9)
#>   [2:0x5648a13840e0]

## pickleListOfValues memory size
obj_size(pickleListOfValues)
#> 8,112 B

# The pickled object consumes the same memory as the original