Path: utzoo!utgpu!water!watmath!clyde!bellcore!decvax!ucbvax!WYOCDC1.BITNET!ZSYJKAA
From: ZSYJKAA@WYOCDC1.BITNET (Jim Kirkpatrick 307 766-5303)
Newsgroups: comp.os.vms
Subject: FORT vs. zero-length strings
Message-ID: <880207233405.000025A8.GBJR.83@WYOCDC1>
Date: 8 Feb 88 06:34:05 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 31

Regarding the following, earlier posting --

     I tripped over the following this week (VMS 4.6, FORTRAN 4.71-271)
     ------------------------------------------------------------------------
     $TYPE TEST.FOR
         CHARACTER*8 RED
         RED = ''
         END
     $ FORTRAN TEST
     %FORT-E-ZERLENSTR, Zero-length string
[and so on]

What the REAL problem is, is that FORTRAN character variables always have
fixed length; if you say RED = 'HI' it really gets padded to RED = 'HI      '
(since you declared RED as CHARACTER*8).  Thus, you cannot have zero-length
character variables.  Somebody at DEC (or would it be ANSI) mis-interpreted
this as meaning you cannot have zero-length character strings.  It seems to
me, if I say RED=character-string, and if character-string is less than 8
characters long, it gets padded with (8-size) blanks;  why this does not apply
if character-string is zero characters long, in which case 8 padding bytes
should simply be added, baffles me.  Every computer system I've been on has
some mentally-retarded feature that doesn't handle some sort of "zero"
condition properly.  I remember a big argument with SDS (long ago!) that
their SORT utility should not abort if the input file contains zero records,
it should simply sort all zero records and build an output file containing
all zero records in proper order!  They insisted sorting zero records was an
error condition.

If anybody has the ANSI standard handy, could you look up the formal definition
of a character string and let us (or me) know if zero-length strings are OK?
I'd hate to flame DEC if ANSI deserves it instead.