Convenience constructor for seconds + nanoseconds about nanotime

[eddelbuettel]

Do you mean at the R or at the C++ level? In R, seconds since epoch is the usual conversion from POSIXct:

R> library(nanotime)
R> now <- Sys.time()
R> format(nanotime(now))
[1] "2017-01-11T17:59:03.210716+00:00"
R> format(nanotime(as.numeric(now)*1e9))
[1] "2017-01-11T17:59:03.210716672+00:00"
R> 

The two are the same (relative to rounding error) as nanotime has nanoseconds since epoch \approx (fractional) seconds since epoch \times 1e9

Nanoseconds since last second sounds weird. Just instantiate that last second and add to it:

R> format(nanotime(now)  + 42)    # 42 nsec later
[1] "2017-01-11T17:59:03.210716042+00:00"
R> 

[dcdillon]

Like I said, just convenience for times coming from other sources. struct timespec has exactly that: seconds since epoch and nanoseconds since last second.

[eddelbuettel]

Right. But isn't what RcppCCTZ does with parseDatetime (returning two values) which we then use here in a somewhat ill-named helper.

[eddelbuettel]

Here is a minimal answer. Start with sourceCpp() on this snippet to get struct timeres to R:

#include <Rcpp.h>

// [[Rcpp::plugins(cpp11)]]

// [[Rcpp::export]]
Rcpp::NumericMatrix hiresNow() {
  struct timespec ts;
  timespec_get(&ts, TIME_UTC);
  Rcpp::NumericMatrix M(1,2);
  M(0,0) = ts.tv_sec;
  M(0,1) = ts.tv_nsec;
  return M;
}

This basically just wraps access to struct timeres into a function returning the Matrix representation I already needed anyway.

Then:

R> library(Rcpp)
R> sourceCpp("/tmp/danCtor.cpp")
R> M <- hiresNow(); now <- Sys.time()
R> M
           [,1]      [,2]
[1,] 1484183403 196793033
R> library(nanotime)
R> nanotime:::nanotime.matrix(M)
integer64
[1] 1484183403196793033
R> format(nanotime:::nanotime.matrix(M))    # at UTC
[1] "2017-01-12T01:10:03.196793033+00:00"
R> now                                                           # local POSIXct
[1] "2017-01-11 19:10:03.207415 CST"         
R> format(now, tz="UTC")                             # POSIXct at UTC
[1] "2017-01-12 01:10:03.207415"
R> 

and the difference is the difference in the time in M <- hiresNow(); now <- Sys.time() which appears to be 10 msec or so. Sounds about right.

So you just need to write yourself a convenience wrapper for nanotime:::nanotime.matrix().