join() crash

[low-decarie]

Thank you for the tremendously good work on this essential package.

My current script that causes the crash is too bulky for upload. I am working on an example script that will cause the same crash.

join() crashes my R session with:

*** caught segfault ***
address 0x0, cause 'memory not mapped'

Traceback:
1: .Call("split_indices", index, group, as.integer(n))
2: split_indices(seq_along(keys$y), keys$y, keys$n)
3: join_ids(x, y, by, all = TRUE)
4: join_all(x, y, by, type)
5: join(counts.transplant, counts.clamy, by = "Water.plot")

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace

within RStudio, this causes the whole app to crash.

Thank you and have an excellent day,

Etienne

[mndrs]

I have this same exact issue. I've had join in plyr crash R 2.13.1 and 2.12.0 (as well as RStudio).

[hadley]

Here's a reproducible example from @imark:

m1<-data.frame(cl=c(1,2), file=c("hi", "low"))
m2<-data.frame(file=c("1776.txt", "About.txt"), actual=c(11.5, 4.5), stringsAsFactors=F)
join(m1, m2, "file")

[brendano]

I've been getting this too. There's something funny going on with factors vs character join columns, and the presence of NA's.

Factor vs. Character

Works:

d1 = data.frame(x=c('a','b'), y=1:2, stringsAsFactors=F)
d2 = data.frame(x=c('b','d'), z=1:2, stringsAsFactors=F)
join(d1,d2)

Works, even though the factors have different levels:

d1 = data.frame(x=c('a','b'), y=1:2)
d2 = data.frame(x=c('b','d'), z=1:2)
join(d1,d2)

Works, though gets the wrong answer:

d1 = data.frame(x=c('a','b'), y=1:2, stringsAsFactors=F)
d2 = data.frame(x=c('b','d'), z=1:2)
join(d1,d2)

Crashes:

d1 = data.frame(x=c('a','b'), y=1:2)
d2 = data.frame(x=c('b','d'), z=1:2, stringsAsFactors=F)
join(d1,d2)

Specifically, it's a segfault in split_indices.

NA's in factors

When the right join column (under a left join) is a factor and has an NA in it, it wants to crash.

Works:

d1 = data.frame(x=c('a','b'), y=1:2, stringsAsFactors=F)
d2 = data.frame(x=c('b',NA), z=1:2, stringsAsFactors=F)
join(d1,d2)

Works:

d1 = data.frame(x=c(NA,'b'), y=1:2)
d2 = data.frame(x=c('b','c'), z=1:2)
join(d1,d2)

Crashes:

d1 = data.frame(x=c('a','b'), y=1:2)
d2 = data.frame(x=c('b',NA), z=1:2)
join(d1,d2)

Again, the segfault is in split_indices.

NA's in numerics are fine

These both work. The problem seems constricted to factor vectors.

d1 = data.frame(x=c(10,11), y=1:2)
d2 = data.frame(x=c(11,12), z=1:2)
join(d1,d2)

d1 = data.frame(x=c(10,11), y=1:2)
d2 = data.frame(x=c(11,NA), z=1:2)
join(d1,d2)

d1 = data.frame(x=c(NA,11), y=1:2)
d2 = data.frame(x=c(11,12), z=1:2)
join(d1,d2)

Non-determinism

Sometimes, instead of a segfault I get a benign error message in split_indices. If I start R fresh and do a similar setup as the NA version above, just with larger data frames:

d1 = data.frame(x=letters[1:4], y=1:4)
d2 = data.frame(x=letters[2:5], z=1:4)
d2$x[2] = NA
join(d1,d2)

I only get the error

Error in split_indices(seq_along(keys$y), keys$y, keys$n) : 
  INTEGER() can only be applied to a 'integer', not a 'character'
Calls: join -> join_all -> join_ids -> split_indices -> .Call

(which, by the way, seems strange since the columns are factors, not characters.)

But if I do it a few more times, I get the segfault, same as all the above crashes:

 *** caught segfault ***
address 0x202, cause 'memory not mapped'

Traceback:
 1: .Call("split_indices", index, group, as.integer(n))
 2: split_indices(seq_along(keys$y), keys$y, keys$n)
 3: join_ids(x, y, by, all = TRUE)
 4: join_all(x, y, by, type)
 5: join(d1, d2)