R data.table bug cannot rename column names which are NA

Question

I have a data.table where some columns names are NA. trying to change them to a character name fails and they stay NA.

I manage to replace them by switching to a data.frame, though, but is there a way with data.table?

dt <- data.table(a = 1:2, b = 2:3)
setDF(dt)
names(dt) <- c(NA,"c")
setDT(dt)
names(dt) <- c("a","b")
names(dt)
# [1] NA  "b"`

Using data.frame it works:

setDF(dt)
names(dt) <- c("a", "b")
names(dt)
# [1] "a" "b"`

EDIT: @akrun suggests to use NA_character_ but that doesn't work for several NA in the names (which is my case but the example above was simplified)

dt <- data.table(a = 1:2, b = 2:3, c = 2:3)
setDF(dt)
names(dt) <- c(NA,NA,"c")
setDT(dt)    
setnames(dt, NA_character_, c('a','b'))

Error in setnames(dt, NA_character_, c("a", "b")) : Some items of old are duplicated (ambiguous) in column names: NA

setnames(dt, c(NA_character_,NA_character_), c('a','b'))

Error in setnames(dt, c(NA_character_, NA_character_), c("a", "b")) : Some duplicates exist in old: NA

PS:

sessionInfo()

R version 3.4.2 (2017-09-28)
Platform: x86_64-suse-linux-gnu (64-bit)
Running under: SUSE Linux Enterprise Desktop 12 SP2

Matrix products: default
BLAS/LAPACK: /usr/lib64/libopenblas_serial.so.0

locale:
 [1] LC_CTYPE=en_GB.UTF-8       LC_NUMERIC=C               LC_TIME=en_GB.UTF-8       
 [4] LC_COLLATE=en_GB.UTF-8     LC_MONETARY=en_GB.UTF-8    LC_MESSAGES=en_GB.UTF-8   
 [7] LC_PAPER=en_GB.UTF-8       LC_NAME=C                  LC_ADDRESS=C              
[10] LC_TELEPHONE=C             LC_MEASUREMENT=en_GB.UTF-8 LC_IDENTIFICATION=C       

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

other attached packages:
 [1] rvest_0.3.2         xml2_1.1.9000       bindrcpp_0.2        shiny_1.0.5        
 [5] dplyr_0.7.3         RUnit_0.4.31        gjpoisson_0.4      
[17] gplots_3.0.1        moments_0.14        foreach_1.4.3       ggplot2_2.2.1      
[21] RODBC_1.3-15        data.table_1.10.4-3 mgcv_1.8-21         nlme_3.1-131       
[25] pacman_0.4.6        devtools_1.13.3    

loaded via a namespace (and not attached):
 [1] bitops_1.0-6        xts_0.10-0          lubridate_1.6.0     bit64_0.9-7        
 [5] httr_1.3.1          quantDb_0.4.0       RColorBrewer_1.1-2  tools_3.4.2        
 [9] backports_1.1.1     rredis_1.7.0        R6_2.2.2           
[13] KernSmooth_2.23-15  rpart_4.1-11        Hmisc_4.0-3         DBI_0.7            
[17] lazyeval_0.2.0      colorspace_1.3-2    nnet_7.3-12         withr_2.0.0        
[21] gridExtra_2.3       curl_2.8.1          bit_1.1-12          compiler_3.4.2     
[25] htmlTable_1.9       caTools_1.17.1      scales_0.5.0        dygraphs_1.1.1.4   
[29] checkmate_1.8.3     odbc_1.1.1          speedglm_0.3-2      stringr_1.2.0      
[33] digest_0.6.12       foreign_0.8-69      datashop_0.13.2     base64enc_0.1-3    
[37] pkgconfig_2.0.1     htmltools_0.3.6     htmlwidgets_0.9     rlang_0.1.2        
[41] ggthemes_3.4.0      bindr_0.1           zoo_1.8-0           gtools_3.5.0       
[45] acepack_1.4.1       inline_0.3.14       marketUtils_0.3.8   magrittr_1.5       
[49] Formula_1.2-2       Matrix_1.2-11       Rcpp_0.12.12        munsell_0.4.3      
[53] stringi_1.1.5       yaml_2.1.14         MASS_7.3-47         RJSONIO_1.3-0      
[57] plyr_1.8.4          grid_3.4.2          blob_1.1.0          gdata_2.18.0       
[61] ggrepel_0.6.5       lattice_0.20-35     splines_3.4.2       fasttime_1.0-2     
[65] hms_0.3             knitr_1.17          reshape2_1.4.2      codetools_0.2-15   
[69] fctsUtils_0.4.7     XML_3.98-1.9        glue_1.1.1          latticeExtra_0.6-28
[73] selectr_0.3-1       httpuv_1.3.5        gtable_0.2.0        purrr_0.2.4        
[77] tidyr_0.7.1         assertthat_0.2.0    mime_0.5            xtable_1.8-2       
[81] survival_2.41-3     quantum_0.13.1      tibble_1.3.4        iterators_1.0.8    
[85] memoise_1.1.0       cluster_2.0.6
> 

Answer 1

This was a bug, thanks for identifying the issue and providing a reproducible example! You should be able to install the current development version of data.table (1.10.5) with:

install.packages('data.table', type = 'source',
                 repos = 'http://Rdatatable.github.io/data.table')

If that doesn't work directly, please consult the Installation Wiki.

If you're unable to install this version (administrative rights or can only install from CRAN), here's a workaround: the bug emerges when only the old argument of setnames is present (in which case it is somewhat paradoxically --though I think intuitively in usage -- interpreted as new).

So to get around this, we need only be sure to use both old and new arguments to setnames:

setnames(dt, seq_along(dt), c('a', 'b', 'c'))
dt
#    a b c
# 1: 1 2 2
# 2: 2 3 3

We can't use names(dt) in the old argument, because there are duplicates in names(dt), and when old is character, we need to be able to match 1-1 the old and new names (i.e., does a belong to the first NA or the second? The same problem would arise if names(dt) was c('a', 'a', 'b') to start (i.e., that's a separate issue). To get around this, we specify the positions instead of the names.