Indent indices for token "EQ_SUB" are different.

[lorenzwalthert]

The problem with the token "EQ_SUB" mentioned here arises because of the fact that expression "a = 3" and "b = 44" are not nested one level deeper, i.e. they are all direct sub-expressions of the root token:

styler:::create_tree("data_frame(a = 3, b = 44)")
#>                                              levelName
#> 1  ROOT (token: short_text [lag_newlines/spaces] {id})
#> 2   °--expr:  [0/0] {18}                              
#> 3       ¦--expr:  [0/0] {3}                           
#> 4       ¦   °--SYMBOL_FUNCTION_CALL: data_ [0/0] {1}  
#> 5       ¦--'(': ( [0/0] {2}                           
#> 6       ¦--SYMBOL_SUB: a [0/1] {4}                    
#> 7       ¦--EQ_SUB: = [0/1] {5}                        
#> 8       ¦--expr:  [0/0] {7}                           
#> 9       ¦   °--NUM_CONST: 3 [0/0] {6}                 
#> 10      ¦--',': , [0/1] {8}                           
#> 11      ¦--SYMBOL_SUB: b [0/1] {11}                   
#> 12      ¦--EQ_SUB: = [0/1] {12}                       
#> 13      ¦--expr:  [0/0] {14}                          
#> 14      ¦   °--NUM_CONST: 44 [0/0] {13}               
#> 15      °--')': ) [0/0] {15}

We used compute_indent_indices() so far to compute the indices that need indention. However, with all other operators such as braces or + %>%, the indention is always from the token on until the last or the penultimate token in that parse table. Here, it is only the next token, so we could go about it as follows:

indent_eq_sub <- function(pd,
                          indent_by,
                          token = "EQ_SUB") {
  eq_sub <- which(pd$token == "EQ_SUB")
  if (length(eq_sub) == 0) return(pd)
  has_line_break <- which(pd$lag_newlines > 0)
  indent_indices <- intersect(eq_sub + 1, has_line_break)
  pd$indent[indent_indices] <- pd$indent[indent_indices] + indent_by
  pd
}

Alternatively, we had to modify compute_indent_indices(), which has already an argument indent_last. We could make that argument non-boolean and change it to indent_up_to and allow values "next", "penulatimate", "last".

[krlmlr]

Do we treat EQ_SUB as operator by any chance? In that case, one option would be simply not to treat it as operator but apply a different rule.

[lorenzwalthert]

No, it's not treated as an operator in the sense that it is not indented with the other operators such as + with indent_op(). We have the function indent_eq_sub() but it does not account for the structure described above properly, so I suggested to modify it.

[krlmlr]

Modifying indent_eq_sub() sounds good.