/ arrays-jvn /

Arrays by JVN

description:

Well, as I am sure everyone knows to the point of hurling (regurgitating)
by now, I chose a format that looks somewhat like Fortran. I could not
use the right parenthesis, ) , because it was taken as the closure for
a parenthesized comment. Could not use the [ ]'s because they also have
a definite meaning (turn compiler off and on by switching STATE).
So I was left with what I could do using curly braces { and }.

Now why did I want an array notation that looks like Fortran? There were
several reasons:

1. I wanted (at that time) to sell Forth to the Fortran community.
(Silly me! But I was young then.)

2. I wanted a notation that unequivocally said "I am an array!"
I had already written a number of linear equations and other
matrix ops programs and kept running into the problems of

a. different constructors for each data type;
b. reading the program when it was done.

3. I wanted something that would be completely transparent, would
permit address arithmetic, and would not require tons of comments
to be maintainable.

Eventually I hit on the Forthish solution of defining an array with a
header that contained its size and its data size. Then I wrote a de-
referencing operator that would dig into the header and calculate the
address of the datum being indexed. The notatin for this was

v{ I } ( -- adr of v[I] )

The left curly brace in an array's name was simply syntactic sugar.
But at some point I realized that by naming 2-dimensional arrays
something like M{{ (2 curly braces) I could have the name say
"I am a 2-dimensional array." Then one dereferences such by saying

M{{ I J }}

--that is, the 2-dim dereferencing operator expects a base address and
two indices and produces the correct address of the I,Jth element.

For those who want it, go to my home page (under construction) then to
the link "Computational Methods in the Physical Sciences", and thence
to "Forth system and example programs". There you can find the file
arrays.f that does all of this stuff. I have bounds checking in that
version because it was intended for student use. Experts can delete
that stuff.


implementation:


\ Toolset for one- and two-dimensional arrays in ANS Forth

\ ---------------------------------------------------
\ (c) Copyright 2001 Julian V. Noble. \
\ Permission is granted by the author to \
\ use this software for any application pro- \
\ vided this copyright notice is preserved. \
\ ---------------------------------------------------

\ Requires CORE and CORE EXT wordsets

: [undefined] BL WORD FIND NIP 0= ;

\ non-Standard word:
[undefined] CELL- [IF] : cell- [ -1 CELLS ] LITERAL + ; [THEN]


\ words for 1-dimensional arrays
: long ;
: 1array ( len #bytes/datum --) \ ( #b len data ...)
CREATE 2DUP , , * ALLOT ;
: _len ( base_addr -- len) \ determine length of an array
CELL+ @ ;

: } ( base_adr indx -- adr[indx] )
OVER _len OVER <= OVER 0< OR ABORT" Index out of range"
OVER @ * + CELL+ CELL+ ;

\ words for 2-dimensional arrays
: wide ;

: 2array ( hgt wid data_size --) \ ( wid #b len data ...)
CREATE >R TUCK , ( wid hgt)
R@ , * DUP , R> * ALLOT ;


: }} ( base_adr m n -- adr[m,n] ) \ data stored row-wise
2>R CELL+ DUP cell- @
R> * R> + ( base_adr+cell m+n*w)
} ;

FALSE [IF]
Usage examples:

20 long 2 FLOATS 1array a{ \ complex vector
20 long 20 wide 1 FLOATS 2array M{{ \ real matrix
20 long 1 CELLS 1array Irow{ \ single-length, integer-
\ valued vector
say

M{{ I J }} ( -- adr[m_ij] )

to dereference.
[THEN]




page composed from:


Message-ID: <3EF64F99.669AB670@virginia.edu> by Julian V. Noble
phys.virginia.edu/classes/551.jvn.fall01/arrays.f




generated Wed Jul 23 02:53:41 2003mlg