Rule feedback: reorder definitions to the top of routine

I have an issue with this check existing at all. It violates the good practice of using the most restrictive visibility.

Example:

method test.
DATA: 
  some_flag type abap_bool,
  some_string type string.

some_flag = abap_false.

"
"imagine 15 lines of code here 
"

some_string = 'value'.

" some more code here
endmethod.

By placing the declaration at the top, I’m making the string visible to the part of the method I’m not using it in. I’d rather do

method test.
DATA:  some_flag type abap_bool.

some_flag = abap_false.

"
"imagine 15 lines of code here 
"
DATA:  some_string type string.
some_string = 'value'.

" some more code here
endmethod.

If I don’t intend to use the string in the first half of the method, I shouldn’t need to declare it at the beginning - then I’m protected against accidental usage and I’m making my reasoning about the usage explicit to future maintainers.

This is precisely why inline data declarations are good (they have their own scoping issues, but that is another matter), but these might not be available on your release.

[FreHu]

as close to the first use as possible - not entirely.

LOOP AT table INTO DATA(x).
  variable = x-blah.
ENDLOOP.
variable = 123.

should declare like this

DATA(variable) = 0.
LOOP AT table INTO DATA(x).
  variable = x-blah.
ENDLOOP.
variable = 123.

not

LOOP AT table INTO DATA(x).
  DATA(variable) = 0.
  variable = x-blah.
ENDLOOP.
variable = 123.

so it should be as close as possible without breaking the code or violating the sane block scoping rules in other languages

[larshp]

ABAP is a dangerous place, anyhow, I’m having a go at it https://github.com/abaplint/abaplint/pull/933