segfaults/inconsistent results with is.na

[andrewblim]

Tried this on dplyr 0.4.2, R 3.2.1, and OS X 10.10.3. Coworker got similar results on Ubuntu 14.04.2 LTS.

Setup:

library(dplyr)
df <- data_frame(x=c(rep(c(0,NA),100000)))

df %>% dplyr::mutate(z = is.na(x)) works fine.

df %>% dplyr::mutate(y = 1, z = is.na(x)) sometimes seems to work fine but sometimes either hangs R or crashes it with an error that sometimes is the usual segfault "address 0x10b63a000, cause 'memory not mapped'" and sometimes an error on malloc:

R(88224,0x7fff72ccd300) malloc: *** error for object 0x10f9fab88: pointer being freed was not allocated
*** set a breakpoint in malloc_error_break to debug
Abort trap: 6

It doesn't always happen - running this up to 10 times sequentially is usually enough to get it to happen on my machine.

df %>% dplyr::mutate(y = 1, z = ifelse(is.na(x), 0, 1)) suffers the same problems as above and also sometimes returns inconsistent/incorrect results, for example:

   x     y z
1  0 FALSE 1
2  0 FALSE 1
3  0 FALSE 1
4  0 FALSE 1
5  0 FALSE 1
6  0 FALSE 1
7  0 FALSE 1
8  0 FALSE 1
9  0 FALSE 1
10 0 FALSE 1

or

    x y z
1   0 1 1
2  NA 0 0
3   0 1 1
4  NA 0 0
5   0 1 1
6  NA 0 0
7   0 1 1
8  NA 0 0
9   0 1 1
10 NA 0 0

Similar issues with SE calls df %>% dplyr::mutate_(1, ~ is.na(x)) and df %>% dplyr::mutate_(1, ~ ifelse(is.na(x), 0, 1)).

[romainfrancois]

Did you try against the dev version ?

Envoyé de mon iPhone

Le 23 juil. 2015 à 17:53, Andrew Lim notifications@github.com a écrit :

Tried this on dplyr 0.4.2, R 3.2.1, and OS X 10.10.3. Coworker got similar results on Ubuntu 14.04.2 LTS.

Setup:

library(dplyr)
df <- data_frame(x=c(rep(c(0,NA),100000)))
df %>% dplyr::mutate(z = is.na(x)) works fine.

df %>% dplyr::mutate(y = 1, z = is.na(x)) sometimes seems to work fine but sometimes either hangs R or crashes it with an error that sometimes is the usual segfault "address 0x10b63a000, cause 'memory not mapped'" and sometimes an error on malloc:

R(88224,0x7fff72ccd300) malloc: *** error for object 0x10f9fab88: pointer being freed was not allocated
*** set a breakpoint in malloc_error_break to debug
Abort trap: 6
It doesn't always happen - running this up to 10 times sequentially is usually enough to get it to happen on my machine.

df %>% dplyr::mutate(y = 1, z = ifelse(is.na(x), 0, 1)) suffers the same problems as above and also sometimes returns inconsistent/incorrect results, for example:

x y z
1 0 FALSE 1
2 0 FALSE 1
3 0 FALSE 1
4 0 FALSE 1
5 0 FALSE 1
6 0 FALSE 1
7 0 FALSE 1
8 0 FALSE 1
9 0 FALSE 1
10 0 FALSE 1
or

x y z

1 0 1 1
2 NA 0 0
3 0 1 1
4 NA 0 0
5 0 1 1
6 NA 0 0
7 0 1 1
8 NA 0 0
9 0 1 1
10 NA 0 0
Similar issues with SE calls df %>% dplyr::mutate_(1, ~ is.na(x)) and df %>% dplyr::mutate_(1, ~ ifelse(is.na(x), 0, 1)).

—
Reply to this email directly or view it on GitHub.

[romainfrancois]

I strongly believe this has been fixed in the dev version. Probably along fixing #1231
Please reopen when you still see the problem with the dev version.