as_data_frame.matrix can't handle subclassed matrices

[tjmahr]

I was writing a tidier function with as_data_frame to convert a matrix generated by a model comparison function. The conversion appeared to work until tried to mutate the tibble. Then I got the error Error: matrix as column is not supported. I was able to trace the problem to how the columns in tibble retained their "matrix" class.

Here is an example using stats::poly, which also produces a subclassed matrix.

I decided to file an issue because this behavior is unexpected and because as.data.frame does work as expected.

library("dplyr", warn.conflicts = FALSE)
#> Warning: package 'dplyr' was built under R version 3.3.1
library("tibble")

poly(1:6, 3) %>% str
#>  poly [1:6, 1:3] -0.598 -0.359 -0.12 0.12 0.359 ...
#>  - attr(*, "dimnames")=List of 2
#>   ..$ : NULL
#>   ..$ : chr [1:3] "1" "2" "3"
#>  - attr(*, "coefs")=List of 2
#>   ..$ alpha: num [1:3] 3.5 3.5 3.5
#>   ..$ norm2: num [1:5] 1 6 17.5 37.3 64.8
#>  - attr(*, "degree")= int [1:3] 1 2 3
#>  - attr(*, "class")= chr [1:2] "poly" "matrix"

poly(1:6, 3) %>% as_data_frame %>% mutate(Condition = "Test")
#> Error in eval(expr, envir, enclos): matrix as column is not supported

poly(1:6, 3) %>% as_data_frame %>% str
#> Classes 'tbl_df', 'tbl' and 'data.frame':    6 obs. of  3 variables:
#>  $ 1:Classes 'poly', 'matrix'  atomic [1:6] -0.598 -0.359 -0.12 0.12 0.359 ...
#>   .. ..- attr(*, "coefs")=List of 2
#>   .. .. ..$ alpha: num [1:3] 3.5 3.5 3.5
#>   .. .. ..$ norm2: num [1:5] 1 6 17.5 37.3 64.8
#>   .. ..- attr(*, "degree")= int [1:3] 1 2 3
#>  $ 2:Classes 'poly', 'matrix'  atomic [1:6] 0.546 -0.109 -0.436 -0.436 -0.109 ...
#>   .. ..- attr(*, "coefs")=List of 2
#>   .. .. ..$ alpha: num [1:3] 3.5 3.5 3.5
#>   .. .. ..$ norm2: num [1:5] 1 6 17.5 37.3 64.8
#>   .. ..- attr(*, "degree")= int [1:3] 1 2 3
#>  $ 3:Classes 'poly', 'matrix'  atomic [1:6] -0.373 0.522 0.298 -0.298 -0.522 ...
#>   .. ..- attr(*, "coefs")=List of 2
#>   .. .. ..$ alpha: num [1:3] 3.5 3.5 3.5
#>   .. .. ..$ norm2: num [1:5] 1 6 17.5 37.3 64.8
#>   .. ..- attr(*, "degree")= int [1:3] 1 2 3

# workaround
poly(1:6, 3) %>% as.data.frame %>% as_data_frame %>% mutate(Condition = "Test")
#> # A tibble: 6 x 4
#>            1          2          3 Condition
#>        <dbl>      <dbl>      <dbl>     <chr>
#> 1 -0.5976143  0.5455447 -0.3726780      Test
#> 2 -0.3585686 -0.1091089  0.5217492      Test
#> 3 -0.1195229 -0.4364358  0.2981424      Test
#> 4  0.1195229 -0.4364358 -0.2981424      Test
#> 5  0.3585686 -0.1091089 -0.5217492      Test
#> 6  0.5976143  0.5455447  0.3726780      Test

devtools::session_info()
#> Session info --------------------------------------------------------------
#>  setting  value                       
#>  version  R version 3.3.0 (2016-05-03)
#>  system   x86_64, mingw32             
#>  ui       RTerm                       
#>  language (EN)                        
#>  collate  English_United States.1252  
#>  tz       America/Chicago             
#>  date     2016-06-28
#> Packages ------------------------------------------------------------------
#>  package    * version date       source                        
#>  assertthat   0.1     2013-12-06 CRAN (R 3.0.2)                
#>  DBI          0.4-1   2016-05-08 CRAN (R 3.2.5)                
#>  devtools     1.12.0  2016-06-24 CRAN (R 3.3.1)                
#>  digest       0.6.9   2016-01-08 CRAN (R 3.2.2)                
#>  dplyr      * 0.5.0   2016-06-24 CRAN (R 3.3.1)                
#>  evaluate     0.9     2016-04-29 CRAN (R 3.2.5)                
#>  formatR      1.4     2016-05-09 CRAN (R 3.2.3)                
#>  htmltools    0.3.5   2016-03-21 CRAN (R 3.2.4)                
#>  knitr        1.13    2016-05-09 CRAN (R 3.2.3)                
#>  lazyeval     0.2.0   2016-06-12 CRAN (R 3.3.0)                
#>  magrittr     1.5     2014-11-22 CRAN (R 3.1.2)                
#>  memoise      1.0.0   2016-01-29 CRAN (R 3.2.3)                
#>  R6           2.1.2   2016-01-26 CRAN (R 3.2.3)                
#>  Rcpp         0.12.5  2016-05-14 CRAN (R 3.2.5)                
#>  rmarkdown    0.9.6   2016-05-01 CRAN (R 3.2.3)                
#>  stringi      1.1.1   2016-05-27 CRAN (R 3.2.5)                
#>  stringr      1.0.0   2015-04-30 CRAN (R 3.2.0)                
#>  tibble     * 1.0-12  2016-06-28 Github (hadley/tibble@1e5b140)
#>  withr        1.0.2   2016-06-20 CRAN (R 3.3.1)                
#>  yaml         2.1.13  2014-06-12 CRAN (R 3.1.0)

[krlmlr]

Each column gets a copy of the input object's attributes, including class: https://github.com/hadley/tibble/blob/6340652ee468bffcffa8180234d00f282a0ec55e/src/matrixToDataFrame.cpp#L21. This is necessary e.g. to support conversion of Date matrices, but unhelpful in the case shown here.

[krlmlr]

A factor (or otherwise classed) matrix can be created via dim<-:

a <- factor(letters)
dim(a) <- c(13, 2)
class(a)
is.matrix(a)

Turns out the poly class is a very special case. I'll simply implement as_data_frame.poly().

[krlmlr]

@hadley: Can you think of other common matrix subclasses that may require special treatment here?

[hadley]

The most important would be table, but I think you already handle that. Maybe ftable? (But that's v. low priority)

[krlmlr]

Thanks. These are handled via the as.data.frame() forward. Not adding a test for now.