styling dplyr::case_when()

[lorenzwalthert]

Reference: #714, i.e. should be one condition per line. Also note overlap with #258 for ~.

Unfortunately, the structure of the parse table in c(x~y) is not equivalent to c(x = y) so we can't rely on the alignment detection algorithm as is 🤔

styler::cache_deactivate()
#> Deactivated cache.
styler:::create_tree(
  "c(x = y) ;c(x~y)"
)
#>                                                  levelName
#> 1  ROOT (token: short_text [lag_newlines/spaces] {pos_id})
#> 2   ¦--expr: c(x = [0/1] {1}                              
#> 3   ¦   ¦--expr: c [0/0] {3}                              
#> 4   ¦   ¦   °--SYMBOL_FUNCTION_CALL: c [0/0] {2}          
#> 5   ¦   ¦--'(': ( [0/0] {4}                               
#> 6   ¦   ¦--SYMBOL_SUB: x [0/1] {5}                        
#> 7   ¦   ¦--EQ_SUB: = [0/1] {6}                            
#> 8   ¦   ¦--expr: y [0/0] {8}                              
#> 9   ¦   ¦   °--SYMBOL: y [0/0] {7}                        
#> 10  ¦   °--')': ) [0/0] {9}                               
#> 11  ¦--';': ; [0/0] {10}                                  
#> 12  °--expr: c(x~y [0/0] {11}                             
#> 13      ¦--expr: c [0/0] {13}                             
#> 14      ¦   °--SYMBOL_FUNCTION_CALL: c [0/0] {12}         
#> 15      ¦--'(': ( [0/0] {14}                              
#> 16      ¦--expr: x~y [0/0] {15}                           
#> 17      ¦   ¦--expr: x [0/0] {17}                         
#> 18      ¦   ¦   °--SYMBOL: x [0/0] {16}                   
#> 19      ¦   ¦--'~': ~ [0/0] {18}                          
#> 20      ¦   °--expr: y [0/0] {20}                         
#> 21      ¦       °--SYMBOL: y [0/0] {19}                   
#> 22      °--')': ) [0/0] {21}

^{Created on 2021-02-15 by the reprex package (v1.0.0)}

i.e. when at the nest expr / ~ / exp, is_alined is false because that's one level up. Without these two being on the same level, a comparison is very hard. We potentially have to move these expressions one level up with flatten_operators_one(), but potentially only within function calls. This probably implies is_tilde_expr() has to be re-thought too as it is not guaranteed that ~ has position 2 or 3.