J3/03-150

Date:      13 March 2003
To:        J3
From:      Nick Maclaren
Subject:   Recursive I/O

______________________________________
Summary
______________________________________

Some implementations of Fortran have supported recursive external I/O
since the 1960s, and all experience is that it is both trivial to
provide and immensely useful.  Vastly the most useful aspect is the
ability to call tracing and diagnostic I/O routines in a subprogram used
within an I/O statement, and it seems extraordinary that Fortran still
does not permit that.  The only requirement is that the recursive I/O
operations should be independent of one another, and most aspects of
that are already required by Fortran.

This caused considerable confusion with Fortran IV, where function calls
were allowed in subscript expressions, unlike in the standard Fortran 66
language.  Some compilers (e.g. IBM Fortran G and WATFIV) did not handle
recursive I/O (or, indeed, recursive anything) and failed horribly,
whereas others did.  For that reason, Fortran 77 explicitly forbade
recursive I/O (sections 6.6 and 12.11), but that decision has never been
revisited.  Fortran 90 reflected the modern acceptance of recursion, but
this area seemed to have been missed.

____________________________________
Edits
______________________________________

_____________
216:25-26
Replace:
"an external unit"
by:
"an external unit that is identified by another I/O statement being
executed"