add xdr argument

[traversc]

Hi Dirk, what do you think about adding an argument for choosing between XDR and binary formats? From my testing serialization using binary fromat is about 3.5x faster.

Comparison

library(Rcpp)
sourceCpp(code="#include <Rcpp.h>
#include "RApiSerializeAPI.h"
// [[Rcpp::depends(RApiSerialize)]]
// [[Rcpp::export(rng=false)]]
SEXP rapi_serializeToRaw(SEXP x, SEXP use_xdrSexp) {
  return serializeToRaw(x, R_NilValue, use_xdrSexp);
}
// [[Rcpp::export(rng=false)]]
SEXP rapi_unserializeFromRaw(SEXP x) {
  return unserializeFromRaw(x);
}")

x <- 1:1e8/2
system.time(out1 <- rapi_serializeToRaw(x, TRUE)) # use XDR
# user  system elapsed
# 0.608   0.069   0.676
system.time(out2 <- rapi_serializeToRaw(x, FALSE)) # use Binary
# user  system elapsed
# 0.104   0.081   0.186

system.time(y1 <- rapi_unserializeFromRaw(out1)) # use XDR
# user  system elapsed
# 0.508   0.051   0.558
system.time(y2 <- rapi_unserializeFromRaw(out2)) # use Binary
# user  system elapsed
# 0.507   0.050   0.557

identical(y1, y2) # TRUE

[eddelbuettel]

Love it. Saw the discussion on the list too.

[eddelbuettel]

This looks good. Left one minor comment.

Thinking it could do with a test. Should we add one? Maybe just extending tests/simpleTests.R (and its reference output) ?

[eddelbuettel]

Let's make this 0.1.2.1, please. I reserve 'a.b.c' versions for actual uploads I trigger.

[eddelbuettel]

Actually the rest of the file uses four spaces indentation.

[eddelbuettel]

I will just make the changes. Also needs a new argument at the R caller.

[eddelbuettel]

Ok, there were one or two little niggles missing I took care of (along with the one nag at CRAN about R_NO_REMAP).

This should be ready for a new CRAN version now. Let me know your thoughts.

[traversc]

Looks good! :)

[eddelbuettel]

I shipped it to CRAN a few hours ago, but I suspect they may have an issue with their test runners as we should long have heard back from the few reverse depends. I guess we will hear by tomorrow.

[eddelbuettel]

Booo. Was out for a nice bike ride and as I get back I am greeted by Uwe with a 'change to worse' for.... wait for it..... qs.

[eddelbuettel]

Fixed the blunder and re-uploaded. Because the existing package is tagged 'NoRemap' it requires a manual check which may happen early tomorrow during European morning hours.

[eddelbuettel]

And as expected the package is now at CRAN. Yay.