Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!indri!uflorida!haven!umbc3!mbph!hybl
From: hybl@mbph.UUCP (Albert Hybl  Dept of Biophysics  SM)
Newsgroups: comp.lang.fortran
Subject: Re: Character aliases are Satanic extensions
Summary: A nutritious way to read/write character data to and from a real array
Keywords: Using SAP inclusions
Message-ID: <600@mbph.UUCP>
Date: 31 May 89 17:50:43 GMT
References: <598@mbph.UUCP> <599@mbph.UUCP>
Organization: University of Maryland, School of Medicine, Baltimore, MD 21201
Lines: 77


In article <1515@hudson.acc.virginia.edu> from
gl8f@bessel.acc.Virginia.EDU (Greg Lindahl), "A challenge"
>If you can come up with an ANSI-Standard Approved Nutritious Scheme for
>writing character data into binary files with real numbers, or putting
>character data into commons with other data, let me know. ...

]khb%chiba@Sun.COM (Keith Bierman - SPD Languages Marketing -- MTS)
]in message <105811@sun.Eng.Sun.COM> responds:
]?? this is perfectly allowed; viz.
]	write(10) character*10,real*8,character*10
]is legal in x3.9-1978.
]
], or putting
]>character data into commons with other data, let me know
]
]this (and equivalence) is what was disallowed.
]

From the porting trenches, let me ask:  can we read/write character
information to and from a real array?  First, let's answer: WHY
DO IT?  Scientist use names for the molecules with which they're
working; they attach names to the residues from which the molecules
are built; they assign names to the the atoms from which the residues
are built; they identify the atom type with a atomic symbol.  Because
of the unpredictable size of a problem, the authors of the programs
I am porting implemented a dynamic storage scheme that uses two large
arrays--one REAL*4 and the other INTEGER*2.  Since many of the names
are four characters long, they were stashed into the REAL*4 array along
with the atomic coordinates and all the other real REAL parameters.
Ok, now lets see how this can be done without rewriting the dynamic
storage scheme and without choking a standards conforming compiler.

Let's try:
      CHARACTER*4 NAME
      REAL X(10)
      NAME = 'Na+ '
      OPEN (UNIT=10,FORM='UNFORMATTED',FILE='F10',STATUS='NEW')
      WRITE (UNIT=10) NAME
      REWIND (UNIT=10)
      READ (UNIT=10) X(4)
      WRITE (*,FMT='(1X,A4)') X(4)  ! Probably not standard and
* is used here only for testing.
      CLOSE (UNIT=10,STATUS='DELETE')
      END

This is, of course, an over simplified example.  The binary write
could be in a different program unit or a different program from the
unit that does the binary read.  The binary file is not portable
between differing hardware types but is the source code? My SVS-fortran
compiler think it's ok.  The binary WRITE has expunged the character
identification from the variable NAME; it becomes just a bit string.
If the length of the bit string is compatible with the length of a
REAL*4 variable, the READ binary stashes the bits into the REAL
array element.  -- We did it -- Right ??? .... 

Maybe not--from a private communication, I learned that there
exists hardware specifically designed for floating point operations.
Certain floating point registers will immediately normalize
the values loaded into them because that is what they were
designed to do.  NORMALIZATION OF A REAL VARIABLE CONTAINING
STASHED CHARACTER INFORMATION CAN DESTROY THE CHARACTER DATA!
The FORTRAN compiler can be written to suppress the use of
these registers when no floating point calculations are
specified thus allowing a programmer to move character data
around by using REAL or DOUBLE PRECISION variables.  But will
the next compiler you use be so thoughtful?

The moral is:  DON'T USE CHARACTER ALIASES!

----------------------------------------------------------------------
Albert Hybl, PhD.              Office UUCP: uunet!mimsy!mbph!hybl
Department of Biophysics       Home   UUCP: uunet!mimsy!mbph!hybl!ah
University of Maryland                CoSy: ahybl
School of Medicine
Baltimore, MD  21201                 Phone: (301) 328-7940 (Office)
----------------------------------------------------------------------