Normalizing Fortran code

I am reminded of a discussion with Michael Metcalf at CERN circa 1990, where I said that the Fortran language was what users wrote, not what was defined by the (then F77) standard. Michael politely disagreed. But that is the world in which I must work. We analyse and re-engineer large Fortran programs written by other people. They use what works: MAP and UNION, DEC Dollar constructs, Open MP … To us that is the language.

I would like to know which features outside the standard are considered essential by other users.

I think many f77 compilers used that character as an extension and also they used that character in data files with namelist input and output (which itself was also an extension).

Even in f90, the square bracket characters [ and ] were not in the language character set. That is why arrays were delimited with (/ and /) two-character tokens in f90. The square brackets were added later, f2003 I think, but the two-character tokens are still part of the standard.

and then just when you thought you did not have to check out which characters terminal keyboards had anymore before buying them along came APL. Try writing in that language with a Lear-Zeigler AMD-3 terminal.