Trace as many attributes as possible when combining

[bart1]

Currently when using a combine transformation all attributes except for the target attributes get omitted. I think if it it not to hard to implement it would be great to retain those attributes that remain valid. There is currently a work around by adding those attributes to the .by arguments to combine call (second example below) but i think it could be desirable to trace these atributes by default. Thanks in advance for the great work.

require(drake)
#> Loading required package: drake
drake_plan(
  i=target(p,transform=map(p=!!(1:2))),
  a=target(x*i, transform=cross(i,x=!!(1:2))),
  b=target(a*y, transform=cross(a,y=!!(1:2),.id=c(p,x))),
  d=target(c(b), transform=combine(b, .by=c(a))),
  trace = T
)
#> # A tibble: 18 x 9
#>    target    command        p     i     x     a       y     b       d      
#>    <chr>     <expr>         <chr> <chr> <chr> <chr>   <chr> <chr>   <chr>  
#>  1 i_1L      1L           … 1L    i_1L  <NA>  <NA>    <NA>  <NA>    <NA>   
#>  2 i_2L      2L           … 2L    i_2L  <NA>  <NA>    <NA>  <NA>    <NA>   
#>  3 a_1L_i_1L 1L * i_1L    … 1L    i_1L  1L    a_1L_i… <NA>  <NA>    <NA>   
#>  4 a_2L_i_1L 2L * i_1L    … 1L    i_1L  2L    a_2L_i… <NA>  <NA>    <NA>   
#>  5 a_1L_i_2L 1L * i_2L    … 2L    i_2L  1L    a_1L_i… <NA>  <NA>    <NA>   
#>  6 a_2L_i_2L 2L * i_2L    … 2L    i_2L  2L    a_2L_i… <NA>  <NA>    <NA>   
#>  7 b_1L_1L   a_1L_i_1L * 1… 1L    i_1L  1L    a_1L_i… 1L    b_1L_1L <NA>   
#>  8 b_1L_1L_2 a_1L_i_1L * 2… 1L    i_1L  1L    a_1L_i… 2L    b_1L_1… <NA>   
#>  9 b_1L_2L   a_2L_i_1L * 1… 1L    i_1L  2L    a_2L_i… 1L    b_1L_2L <NA>   
#> 10 b_1L_2L_2 a_2L_i_1L * 2… 1L    i_1L  2L    a_2L_i… 2L    b_1L_2… <NA>   
#> 11 b_2L_1L   a_1L_i_2L * 1… 2L    i_2L  1L    a_1L_i… 1L    b_2L_1L <NA>   
#> 12 b_2L_1L_2 a_1L_i_2L * 2… 2L    i_2L  1L    a_1L_i… 2L    b_2L_1… <NA>   
#> 13 b_2L_2L   a_2L_i_2L * 1… 2L    i_2L  2L    a_2L_i… 1L    b_2L_2L <NA>   
#> 14 b_2L_2L_2 a_2L_i_2L * 2… 2L    i_2L  2L    a_2L_i… 2L    b_2L_2… <NA>   
#> 15 d_a_1L_i… c(b_1L_1L, b_… <NA>  <NA>  <NA>  a_1L_i… <NA>  <NA>    d_a_1L…
#> 16 d_a_1L_i… c(b_2L_1L, b_… <NA>  <NA>  <NA>  a_1L_i… <NA>  <NA>    d_a_1L…
#> 17 d_a_2L_i… c(b_1L_2L, b_… <NA>  <NA>  <NA>  a_2L_i… <NA>  <NA>    d_a_2L…
#> 18 d_a_2L_i… c(b_2L_2L, b_… <NA>  <NA>  <NA>  a_2L_i… <NA>  <NA>    d_a_2L…
drake_plan(
  i=target(p,transform=map(p=!!(1:2))),
  a=target(x*i, transform=cross(i,x=!!(1:2))),
  b=target(a*y, transform=cross(a,y=!!(1:2),.id=c(p,x))),
  d=target(c(b), transform=combine(b, .by=c(a,x,i,p))),
  trace = T
)
#> # A tibble: 18 x 9
#>    target      command      p     i     x     a      y     b      d        
#>    <chr>       <expr>       <chr> <chr> <chr> <chr>  <chr> <chr>  <chr>    
#>  1 i_1L        1L         … 1L    i_1L  <NA>  <NA>   <NA>  <NA>   <NA>     
#>  2 i_2L        2L         … 2L    i_2L  <NA>  <NA>   <NA>  <NA>   <NA>     
#>  3 a_1L_i_1L   1L * i_1L  … 1L    i_1L  1L    a_1L_… <NA>  <NA>   <NA>     
#>  4 a_2L_i_1L   2L * i_1L  … 1L    i_1L  2L    a_2L_… <NA>  <NA>   <NA>     
#>  5 a_1L_i_2L   1L * i_2L  … 2L    i_2L  1L    a_1L_… <NA>  <NA>   <NA>     
#>  6 a_2L_i_2L   2L * i_2L  … 2L    i_2L  2L    a_2L_… <NA>  <NA>   <NA>     
#>  7 b_1L_1L     a_1L_i_1L *… 1L    i_1L  1L    a_1L_… 1L    b_1L_… <NA>     
#>  8 b_1L_1L_2   a_1L_i_1L *… 1L    i_1L  1L    a_1L_… 2L    b_1L_… <NA>     
#>  9 b_1L_2L     a_2L_i_1L *… 1L    i_1L  2L    a_2L_… 1L    b_1L_… <NA>     
#> 10 b_1L_2L_2   a_2L_i_1L *… 1L    i_1L  2L    a_2L_… 2L    b_1L_… <NA>     
#> 11 b_2L_1L     a_1L_i_2L *… 2L    i_2L  1L    a_1L_… 1L    b_2L_… <NA>     
#> 12 b_2L_1L_2   a_1L_i_2L *… 2L    i_2L  1L    a_1L_… 2L    b_2L_… <NA>     
#> 13 b_2L_2L     a_2L_i_2L *… 2L    i_2L  2L    a_2L_… 1L    b_2L_… <NA>     
#> 14 b_2L_2L_2   a_2L_i_2L *… 2L    i_2L  2L    a_2L_… 2L    b_2L_… <NA>     
#> 15 d_a_1L_i_1… c(b_1L_1L, … 1L    i_1L  1L    a_1L_… <NA>  <NA>   d_a_1L_i…
#> 16 d_a_2L_i_1… c(b_1L_2L, … 1L    i_1L  2L    a_2L_… <NA>  <NA>   d_a_2L_i…
#> 17 d_a_1L_i_2… c(b_2L_1L, … 2L    i_2L  1L    a_1L_… <NA>  <NA>   d_a_1L_i…
#> 18 d_a_2L_i_2… c(b_2L_2L, … 2L    i_2L  2L    a_2L_… <NA>  <NA>   d_a_2L_i…

^{Created on 2019-09-12 by the reprex package (v0.3.0)}

[wlandau]

Seems like an easier problem than provenance in map() and cross(). If there is a quick fix, it will probably be here.

[wlandau]

It took a little work, but drake now keeps non-missing singleton grouping variables within each combine() split.