inconsistent output of add column on join

[jangorecki]

See the example and comment below.

library(data.table)
dt1 <- data.table(a = c(1,2), key="a")
dt2 <- data.table(b = 2, v = 1, key="b")
dt1[dt2, v := i.v]
#    a  v
#1: 1 NA
#2: 2  1

That looks OK.

library(data.table)
dt1 <- data.table(a = 1, key="a")
dt2 <- data.table(b = 2, v = 1, key="b")
dt1[dt2, v := i.v]
#    a
#1: 1

That looks not OK.
I would expect to get:

#    a  v
#1: 1 NA

Of course it should also retain data types, so NA_real_ in that case

R version 3.2.0 (2015-04-16)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 14.04.2 LTS

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_DK.UTF-8        LC_COLLATE=en_US.UTF-8     LC_MONETARY=en_US.UTF-8    LC_MESSAGES=C              LC_PAPER=en_US.UTF-8       LC_NAME=C                  LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] anchormodeling_0.3.8 testthat_0.10.0      devtools_1.8.0       R6_2.0.1             data.table_1.9.5    

loaded via a namespace (and not attached):
 [1] Rcpp_0.11.6          magrittr_1.5         roxygen2_4.1.1       MASS_7.3-40          munsell_0.4.2        colorspace_1.2-6     stringr_1.0.0        plyr_1.8.2           tools_3.2.0          grid_3.2.0           gtable_0.1.2         git2r_0.10.1         rversions_1.0.0     
[14] digest_0.6.8         crayon_1.2.1         reshape2_1.4.1       ggplot2_1.0.1        microbenchmark_1.4-2 bitops_1.0-6         RCurl_1.95-4.6       memoise_0.2.1        stringi_0.4-1        scales_0.2.4         XML_3.98-1.1         chron_2.3-45         proto_0.3-10  

[franknarf1]

Maybe a relevant excerpt from the news/changelog for 1.9.2:

"X[Y, col:=value] when no match exists in the join is now caught early and X is simply returned. Also a message when datatable.verbose is TRUE is provided. In addition, if col is an existing column, since no update actually takes place, the key is now retained." in response to #354

[arunsrinivasan]

Possible duplicate of #820. A PR would be great for this. IIRC we've to not return when i evaluates to all FALSE and j has := (it's all R AFAICT).

[jangorecki]

Another case, more problematic as cannot be easily fixed by if(!"x" %in% names(d1)) d1[, x := NA]. The column which we want to lookup from d2 already exists in d1 and we want to override it.

library(data.table)
d1 <- data.table(a="a",b=FALSE,x=1,key=c("a","b"))
d2 <- data.table(a="a",b=TRUE,x=5,key=c("a","b"))
d1[d2, x :=  i.x]
#    a     b x
# 1: a FALSE 1
## while I would expect to get
#    a     b  x
# 1: a FALSE NA

[jangorecki]

Duplicate of #759, closing