mombasa@kronos.arc.nasa.gov (Tarang Kumar Patel) writes:
> thompson@orpheus.nascom.nasa.gov (William Thompson) writes:
>> daffer@primenet.com writes:
>>> In <3fa92o$4ve@mojo.eng.umd.edu>, surinder@eng.umd.edu (Surinder P. Singh) writes:
>>>>
>>>>
>>>>
>>>> I get data out put from
>>>> FORTRAN 77, a function
>>>> F(i,j,k) .
>>>>
>>>> What is the best way to write this
>>>> data and read by idl ?
>>>> Can i increse speed of reading
>>>> and writing by unformatted data ?
>>>>
>>> Definately use unformated data. To write a four byte float in full precision as
>>> formatted data (i.e. text) takes 8 or [9 bytes] (7 digits + decimal
>>> [+ minus sign]), four of five bytes more than in binary representation.
>>> Just remember that IDL and Fortran think of arrays different ways. I forget
>>> exactly the way you say this, but I think the expression is that Fortran is row
>>> major and IDL is column major. If you don't impose some output
>>> format by means of a do loop around the output statment or an implied
>>> do loop in the format statement, then if the array that is written out has
>>> dimensions 3 X 4 X 5 than it should be read in as 5 X 4 X 3.
>> That's wrong, if one is speaking about Interactive Data Language from Research
>> Systems Incorporated. Perhaps the author was thinking about Interface
>> Definition Language--I don't know anything about that. However, both RSI's IDL
>> and FORTRAN address arrays in the same way. I don't remember the terminology
>> either, but in both IDL and Fortran the first dimension in the array is the one
>> that changes first--i.e. the element (2,1) follows immediately after the
>> element (1,1). Both behave in exactly the opposite way from C.
> Yes, the storage is same for IDL and FORTRAN i.e 1st DIMENSION varying
> fastest. However IDL arrays differ in that the 1st dimension is the COLUMN
> index and not the ROW index.
> for example
> a =indgen(2,3) ; 2 columns, 3 rows
> print,a
> 0 1
> 2 3
> 4 5
> I
> where a(0,0) = 0, a(0,1) = 2, a(0,2) = 4 and so on
> FORTRAN and C adhere to MATRIX notation. Thus in FORTRAN this would appear
> as a(1,1) = 0 a(2,1) = 2, and in C a[0][0]=1, a[1][0] = 2
> The fact that FORTRAN stores an array in ,e,ory as row major has nothing
> to do with the way a user address's the array, thats really code efficiency
> issue. Though as far as storage is concerned IDL and FORTRAN are the same i.e
> 1st DIMENSION varying the fastest, however the 1st dimension has different
> meanings
> So storage wise the elements of the array would appear in a consecutive order
> as follows
>
> Thus for the above example 0,1,2,3,4,5 are stored in that sequence
> In IDL notation a(0,0), a(1,0), a(0,1) note the 1st dimension is the COLUMN
> In FORTRAN notation a(1,1), a(2,1), a(3,1) " 1st " " " ROW
> and thus this would be 0,2,4,1,3,5
>
> In C notation a[0][0],a[0][1],a[1][0],a[1][1] " 1st " " " ROW
> and thus this would be in memory as 0,1,2,3,4,5
> So you see this really confuses matters. The situation is made worse by RSI's
> manual claiming that IDL and FORTRAN ordering are a like i.e row major.
> Well IDL is not row major in the classic notation of MATRICES.
You must be using a different version of Fortran than I'm using. I put
together the following short test program in both Fortran and IDL
Fortran:
program test
dimension a(5,3)
open(unit=1,file='test.dat',form='formatted',readonly,status ='old')
read (1,*) a
read (*,*) i,j
write (*,*) a(i,j)
end
IDL:
pro test
a = fltarr(5,3)
openr,1,'test.dat'
readf,1,a
read,i,j
print,a(i-1,j-1)
end
test.dat
1 20 300 4000 50000
11 12 13 14 15
111 222 333 444 555
Running this on a VAX/VMS computer, I get the same results from both the VMS
and Unix versions no matter what indices I type in. Both IDL and any version
of Fortran I've ever run into work exactly the same way. They're both the
opposite of the way C treats arrays.
Bill Thompson
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
