Milliseconds rounding issues about anytime

[eddelbuettel]

Formating is done by format.POSIXct and is an R feature. Little we can do here, besides formatting ourselves which I'd rather not get into.

[eddelbuettel]

Demonstration not involving any code of this package:

R> txt <- c("2016-02-25 17:34:00.375",
+          "2016-02-25 17:34:00.376",
+          "2016-02-25 17:34:00.377")
R> pt <- as.POSIXct(txt)
R> pt
[1] "2016-02-25 17:34:00.375 CST"
[2] "2016-02-25 17:34:00.375 CST"
[3] "2016-02-25 17:34:00.377 CST"
R> 

My default is already options(digits.secs=6) which I didn't change. I'll close this as it really is an R issue. I don't like it either but that is the way it it, And even when using Boost to format, we get the same issue:

R> anytime:::testFormat("%Y-%m-%d %H:%M:%S%f", txt[1])
[1] "2016-02-25 17:34:00.375 CST"
R> anytime:::testFormat("%Y-%m-%d %H:%M:%S%f", txt[2])
[1] "2016-02-25 17:34:00.375 CST"
R> anytime:::testFormat("%Y-%m-%d %H:%M:%S%f", txt[3])
[1] "2016-02-25 17:34:00.377 CST"
R> 

[eddelbuettel]

Minor thinko on my part in the last bit. That isn't Boost formating the output. I may have a different helper for that or try one.

But the issue in this report is still with format.POSIXct() and hence an R issue.

[eddelbuettel]

But there is hope. I just added a format() function to Rcpp in the transition to the 'new' Datetime classes. Using that, we can do

options(width=50)
print(anytime:::format(pt))

which yields

edd@max:~/git/anytime(master)$ Rscript /tmp/formating.R 
[1] "2016-02-25 17:34:00.375000"
[2] "2016-02-25 17:34:00.376000"
[3] "2016-02-25 17:34:00.377000"
edd@max:~/git/anytime(master)$ 

I'll roll that out in anytime at some point -- I can't quite depend on that yet as not all Rcpp installations have it. And by the same token, I can probably take a shortcut and put a format() routine based on Boost in.

[cactus74]

That this could have been a pure R issue did not even come to my mind. Great to hear this can be overcome in the near future using your format() routine. Thanks a lot.

[eddelbuettel]

It's messy. I just looked at it on the train -- where code comments by myself reminded me that I had looked into this. But in order to use Boost code for string formatting, we need the Boost timezone file which this package does not have :-/ RcppBDT so maybe we can suggest it.

Longer term the Rcpp-based solution will do.