struggling with integers (maximum integer size)

Question

I was using a function that requires input as integers.

So I have been trying to read up on making things integers:

y <- 3.14
as.integer(y)
[1] 3              # all cool

All good, but if I have

 x <- 1639195531833
 as.integer(x)
 [1] NA
 Warning message:
 NAs introduced by coercion

I had options(digits = 15) on and it confused my why it wasn't working but in a clean session it must be to do with the scientific notation.

I also tried to trick R but it was not happy:

  as.integer(as.character(x))
[1] 2147483647
Warning message:
inaccurate integer conversion in coercion

How do I defeat scientific notation and get my integers?

See `?as.integer`, which tells you integers are restricted to approximately +/-2*10^9. — Joshua Ulrich, Jan 29 '13 at 18:08
forget it. Integers in R are still 32bit, so you can't have integers larger than about 1.2E9. This will probably change in the next major R version, if rumours are right. — Joris Meys, Jan 29 '13 at 18:08
ouch. This requires a lot of renaming. thanks guys, shall I delete this Q or would you like an answer here? — user1320502, Jan 29 '13 at 18:10
it's a valid question with now a valid answer, so leave it be I'd say — Joris Meys, Jan 29 '13 at 18:13

Dason · Accepted Answer · 2014-02-25T17:57:24.557

The largest integer R can hold is

.Machine$integer.max
# [1] 2147483647

This has nothing to do with scientific notation and everything to do with how the computer actually stores the numbers. The current version of R stores integers still as 32bit, regardless of the architecture. This might change in the future though.

see also ?as.integer

Currently you can get access to 64 bit integers through the int64 package

> as.integer(.Machine$integer.max)
[1] 2147483647
> # We get problems with this
> as.integer(.Machine$integer.max + 1)
[1] NA
Warning message:
NAs introduced by coercion 
> # But if we use int64
> library(int64)
> as.int64(.Machine$integer.max) + 1L
[1] 2147483648

score 15 · Answer 2 · answered May 22 '17 at 16:34

Update to Dason's answer (don't have enough reputation to make a comment):

The int64 package has now been deprecated, however there is now a package called bit64. I was able to achieve the same result using it. The syntax changed only slightly from "as.int64" to "as.integer64".

I.e.

library(bit64)
as.integer64(.Machine$integer.max) + 1L

score 4 · Answer 3 · edited May 22 '17 at 20:28

4

seems int64 is already deprecated. I use package bit64, method as.integer64() and solve the problem. details are referred to here.

edited May 22 '17 at 20:28

demongolem

9,474
36
90
105

answered Apr 06 '17 at 08:24

HappyCoding

5,029
7
31
51

score 2 · Answer 4 · answered Jan 29 '13 at 18:14

2

There are classes that can handle large integers. I use int64, its vignette is here:

http://cran.r-project.org/web/packages/int64/vignettes/int64.pdf

To use this you just place some normal number into that class then you can add or multiply it across the threshold of what the normal 32bit integer max. Good luck.

answered Jan 29 '13 at 18:14

Seth

4,745
23
27

+1 this looks interesting thank you, although getting that class into the function im running might be tough but good to know. – user1320502 Jan 29 '13 at 18:16

struggling with integers (maximum integer size)

4 Answers4

Linked

Related