To: J3                                                     J3/23-151r1
From: Steve Lionel & Van Snyder
Subject: F202Y Allow I/O of enumeration type names
Date: 2023-June-13

In paper 23-151, Van Snyder notes that I/O of enumeration types
(introduced in F2023 and distinct from interoperable enum types)
are usable in I/O operations only for their ordinal values.
Van properly notes that it is desirable to offer a means
of reading and writing enumeration values by name, and suggests
extending character output, for example A and AT formats, to
support enumeration types. JOR prefers a different option.

Proposal
--------

Extend the G format edit descriptor to support enumeration types.
For output, the field is that of the enumerator name(*) in upper
case, padded with blanks to the right if necessary to satisfy a
nonzero w field. If w is nonzero but is shorter than the width
of the enumerator name, the field is filled with asterisks. For
G0 or G0.d, the output field is the exact length of the enumeration
name.

For input with a G edit descriptor, the character string consists
of a valid enumerator name for the type of the input list item;
irrespective of case, with trailing blanks ignored. If the input
field does not contain such a name, an error condition occurs.

(*) An enumerator name consists of the characters in the name in the
    Fortran character set; the default character set is required to
    include those characters.

    This is an issue which applies to other I/O; for example,
    Japanese character sets contain full-width and half-width versions
    of digits, letters, special characters, etc. It would be
    unfortunate if someone thought that full-width characters should
    be used for any formatted non-character I/O.

    (The half-width ones are the ones that correspond to ASCII digits etc.)

For list-directed output, enumeration type values are produced
with the effect of a G0 descriptor, preceded and/or followed by
a value separator, the same as for other non-character data.

For list-directed input, when the next effective item is of an
enumeration type, the input form consists of characters containing
a valid enumerator name in the item's enumeration type, insensitive
to case.

For namelist formatting, the appropriate list-directed rules apply.