F2007: another possible improvement

INMHO, array constructors should allow us to declare the types of the
entries in a declaration, so that something like the following would
be possible:

CHARACTER(LEN=16), PARAMETER::  WEEKDAYS( 7 ) = &
CHARACTER(LEN=9)(/
'Sunday',    'Monday',   'Tuesday', &
'Wednesday', 'Thursday', 'Friday' &
'Saturday',  'Sunday'
/)

so that all of the names of the days would be coerced into the same
length and therefore could be used as the PARAMETER-value for the
array WEEKDAYS, for example.

Likewise, one should be able to do an array constructor like
REAL(KIND=foo)(/ 1, 2.0, 3.0D0 /)
where all of the entries are coerced into REAL(KIND=foo) before
being used as array-entries in the array-constructor.


fwiw--

Carlie J. Coats, Jr.                        carlie.coats@baronams.com
Environmental Modeling Center               carlie_coats@ncsu.edu
c/o North Carolina State University                   MEAS Department
1125 Jordan Hall                                      Campus Box 8208
Raleigh, NC 27695           phone: (919)515-7076 / fax: (919)515-7802
"My opinions are my own, and I've got *lots* of them!"

Carlie J. Coats wrote:
>
> INMHO, array constructors should allow us to declare the types of the
> entries in a declaration, so that something like the following would
> be possible:
>
>    CHARACTER(LEN=16), PARAMETER::  WEEKDAYS( 7 ) = &
>       CHARACTER(LEN=9)(/
>            'Sunday',    'Monday',   'Tuesday', &
>            'Wednesday', 'Thursday', 'Friday' &
>            'Saturday',  'Sunday'
>             /)
>
> so that all of the names of the days would be coerced into the same
> length and therefore could be used as the PARAMETER-value for the
> array WEEKDAYS, for example.

I'm . What length do you want? Len=16 or Len=9? What's wrong with so
mething like:

CHARACTER(*), PARAMETER::  WEEKDAYS( 7 ) = &
(/'Sunday',    'Monday',   'Tuesday', &
'Wednesday', 'Thursday', 'Friday' &
'Saturday',  'Sunday'/)

and with the result that LEN(WEEKDAYS(1)) = the length of the longest compon
ent.


> Likewise, one should be able to do an array constructor like
>     REAL(KIND=foo)(/ 1, 2.0, 3.0D0 /)
> where all of the entries are coerced into REAL(KIND=foo) before
> being used as array-entries in the array-constructor.

This one I like even less - for personal reasons, since I think it encourage
s sloppiness
(goodness, I sound like a school marm!) I think having to specify the kind t
ype in the
declaration ala

REAL(foo) :: x(3)

x = (/ 1.0_foo, 2.0_foo, 3.0_foo /)

makes one think more about the precision required (e.g. do I even need KIND=
foo? Or will
KIND=bar be enough? Is "x" declared as KIND=foo or KIND=bar?) I've "fixed" p
lenty of code
that didn't work right because people assumed, e.g., the variable assignment
 x=0.1 meant x
really was 0.100000000000....etc, and not x=0.09999999485763.....whatever. (
Aside: the
thing that gets me about this is that a few times the result is said people 
thinking
Fortran is a stupid language. Oof. I'm amazed at the number of people who st
ill don't grok
how numbers are represented in computers.)

> Carlie J. Coats, Jr.                        carlie.coats@baronams.com
> Environmental Modeling Center               carlie_coats@ncsu.edu
> c/o North Carolina State University                   MEAS Department

Hey! I thought *I* was at EMC! :o)

cheers,

paulv

--
Paul van Delst
CIMSS @ NOAA/NCEP/EMC

I think F2003 does what you want, you can
put a type-spec as the leading thing in an array
constructor.  The syntax rule is

R465     array-constructor is (/ ac-spec /)
...
R466     ac-spec  is type-spec ::
or [type-spec ::] ac-value-list

And "type-spec" is
R401 type-spec    is intrinsic-type-spec
or derived-type-spec

which covers anything you should want ;).  It coerces all
of the following constructor values into the specified type.

An example is:
(/ CHARACTER(LEN=7) :: 'Takata', 'Tanaka', 'Hayashi' /)

Before anyone asks, the first form
(/ type-spec:: /)
allows you to create a zero-size array of a specific type.
You couldn't do that before F2003.  (There was an ambiguity
in what (/ (I, I=1,0) /) meant.)


Dick Hendrickson

Carlie J. Coats wrote:
>
> INMHO, array constructors should allow us to declare the types of the
> entries in a declaration, so that something like the following would
> be possible:
>
>    CHARACTER(LEN=16), PARAMETER::  WEEKDAYS( 7 ) = &
>       CHARACTER(LEN=9)(/
>            'Sunday',    'Monday',   'Tuesday', &
>            'Wednesday', 'Thursday', 'Friday' &
>            'Saturday',  'Sunday'
>             /)
>
> so that all of the names of the days would be coerced into the same
> length and therefore could be used as the PARAMETER-value for the
> array WEEKDAYS, for example.
>
> Likewise, one should be able to do an array constructor like
>     REAL(KIND=foo)(/ 1, 2.0, 3.0D0 /)
> where all of the entries are coerced into REAL(KIND=foo) before
> being used as array-entries in the array-constructor.
>
>
> fwiw--
>
> Carlie J. Coats, Jr.                        carlie.coats@baronams.com
> Environmental Modeling Center               carlie_coats@ncsu.edu
> c/o North Carolina State University                   MEAS Department
> 1125 Jordan Hall                                      Campus Box 8208
> Raleigh, NC 27695           phone: (919)515-7076 / fax: (919)515-7802
> "My opinions are my own, and I've got *lots* of them!"