'memory not mapped' in setdiff when NA present in a factor column (dev version)'

[shntnu]

> sessionInfo()
R version 3.2.1 (2015-06-18)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: OS X 10.10.5 (Yosemite)

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     
> packageVersion("dplyr")
[1] ‘0.4.3.9000’
> 
> dplyr::setdiff(data.frame(var = c(NA, "a"), stringsAsFactors = F), data.frame(var = c("a"), stringsAsFactors = F))
   var
1 <NA>
> 
> dplyr::setdiff(data.frame(var = c(NA, "a"), stringsAsFactors = T), data.frame(var = c("a"), stringsAsFactors = T))

 *** caught segfault ***
address 0x7f9f9ef31028, cause 'memory not mapped'

Traceback:
 1: .Call("dplyr_setdiff_data_frame", PACKAGE = "dplyr", x, y)
 2: setdiff_data_frame(x, y)
 3: setdiff.data.frame(data.frame(var = c(NA, "a"), stringsAsFactors = T),     data.frame(var = c("a"), stringsAsFactors = T))
 4: dplyr::setdiff(data.frame(var = c(NA, "a"), stringsAsFactors = T),     data.frame(var = c("a"), stringsAsFactors = T))

[hadley]

Can you please provide a reproducible example? i.e. something I can copy and paste into R directly.

[shntnu]

dplyr::setdiff(data.frame(var = c(NA, "a"), stringsAsFactors = F), data.frame(var = c("a"), stringsAsFactors = F))
# var
# 1 <NA>

dplyr::setdiff(data.frame(var = c(NA, "a"), stringsAsFactors = T), data.frame(var = c("a"), stringsAsFactors = T))
# *** caught segfault ***
#   address 0x7fbd8b3ff058, cause 'memory not mapped'
# 
# Traceback:
# 1: .Call("dplyr_setdiff_data_frame", PACKAGE = "dplyr", x, y)
# 2: setdiff_data_frame(x, y)
# 3: setdiff.data.frame(data.frame(var = c(NA, "a"), stringsAsFactors = T),     data.frame(var = c("a"), stringsAsFactors = T))
# 4: dplyr::setdiff(data.frame(var = c(NA, "a"), stringsAsFactors = T),     data.frame(var = c("a"), stringsAsFactors = T))

[hadley]

Minimal reprex:

library(dplyr)

df1 <- data_frame(x = factor(c(NA, "a")))
df2 <- data_frame(x = factor("a"))

setdiff(df1, df2)

But it works for me in the dev version.

[hadley]

That sha was spurious, but the problem does seem to have come back. @romainfrancois can you please take a look?

[romainfrancois]

Went a little beyond that as setdiff was coercing factors to characters even when the levels were identical. Not anymore:

> df1 <- data_frame(x = factor(c(NA, "a")))
> df2 <- data_frame(x = factor("a"))
>
> setdiff(df1, df2)
Source: local data frame [1 x 1]

       x
  <fctr>
1     NA