| Re: Assoc and byteorder keywords [message #36236] |
Thu, 28 August 2003 08:44  |
Richard French
Messages: 173
Registered: December 2000
|
Senior Member |
|
|
On 8/28/03 7:45 AM, in article BB7363AB.C9D%rfrench@wellesley.edu, "Richard
G. French" <rfrench@wellesley.edu> wrote:
> I have several large binary data sets (image cubes) that I access from
> multiple platforms, using IDL. Typically, I do something like:
>
> Openr,lun,/GET_LUN,'images.bin'
> Images=assoc(lun,fltarr(nx,ny))
>
> I might then want to do something like this:
>
> Norm=(images[5]-images[4])/(images[3]>threshold)
>
> Unfortunately, from some platforms, the images are not in the correct byte
> order. Because byteorder is a procedure and not a function, I am forced into
> locutions such as:
>
> im5=images[5]
> Byteorder,im5, /SWAP_IF_BIG_ENDIAN
>
> im4=images[4]
> Byteorder,im4, /SWAP_IF_BIG_ENDIAN
>
> im3=images[3]
> Byteorder,im3, /SWAP_IF_BIG_ENDIAN
>
> Nomr=(im5-im4)/(im3>threshold)
>
> This seems to be negating the syntactical efficiency of the ASSOC function.
> I could wrap all of this into a function, of course, but it seems to me that
> an easier way would be for ASSOC to have the capability of doing the
> byte-swapping on the fly, by having the same keywords as BYTEORDER. So, for
> example, I would like to see:
>
>
> Images=ASSOC(lun,fltarr(nx,ny),/SWAP_IF_BIG_ENDIAN)
>
> If RSI could just then call byteorder internally to the ASSOC function
> before delivering the data to the user, it would result in much cleaner code
> at my end. Does anyone see a way to do something like this already, or see
> any problems with this suggestion?
>
> Dick French
>
The fine folks at RSI wrote to me directly to let me know that there is such
a keyword to the OPENR command, which does exactly what I am after. Not sure
how I missed this one - probably was introduced a few versions ago and I did
not pay attention to the change. Thought others might want to know.
Dick French
|
|
|
|
| Re: Assoc and byteorder keywords [message #36238 is a reply to message #36236] |
Thu, 28 August 2003 07:43  |
Liam E. Gumley
Messages: 378
Registered: January 2000
|
Senior Member |
|
|
"Richard G. French" <rfrench@wellesley.edu> wrote in message
news:BB7363AB.C9D%rfrench@wellesley.edu...
> I have several large binary data sets (image cubes) that I access from
> multiple platforms, using IDL. Typically, I do something like:
>
> Openr,lun,/GET_LUN,'images.bin'
> Images=assoc(lun,fltarr(nx,ny))
>
> I might then want to do something like this:
>
> Norm=(images[5]-images[4])/(images[3]>threshold)
>
> Unfortunately, from some platforms, the images are not in the correct byte
> order. Because byteorder is a procedure and not a function, I am forced
into
> locutions such as:
>
> im5=images[5]
> Byteorder,im5, /SWAP_IF_BIG_ENDIAN
>
> im4=images[4]
> Byteorder,im4, /SWAP_IF_BIG_ENDIAN
>
> im3=images[3]
> Byteorder,im3, /SWAP_IF_BIG_ENDIAN
>
> Nomr=(im5-im4)/(im3>threshold)
>
> This seems to be negating the syntactical efficiency of the ASSOC
function.
> I could wrap all of this into a function, of course, but it seems to me
that
> an easier way would be for ASSOC to have the capability of doing the
> byte-swapping on the fly, by having the same keywords as BYTEORDER. So,
for
> example, I would like to see:
>
>
> Images=ASSOC(lun,fltarr(nx,ny),/SWAP_IF_BIG_ENDIAN)
>
> If RSI could just then call byteorder internally to the ASSOC function
> before delivering the data to the user, it would result in much cleaner
code
> at my end. Does anyone see a way to do something like this already, or see
> any problems with this suggestion?
Why not try this?
openr, lun, /get_lun, 'images.bin', /swap_if_big_endian
Cheers,
Liam.
Practical IDL Programming
http://www.gumley.com/
|
|
|
|
| Re: Assoc and byteorder keywords [message #36311 is a reply to message #36238] |
Thu, 28 August 2003 16:25 |
Richard French
Messages: 173
Registered: December 2000
|
Senior Member |
|
|
>
> Why not try this?
>
> openr, lun, /get_lun, 'images.bin', /swap_if_big_endian
>
> Cheers,
> Liam.
> Practical IDL Programming
> http://www.gumley.com/
>
>
Well, I had hoped that that would work, but for my floating point images, I
seem to need to use the /LSWAP option of BYTEORDER in order to get things to
work - just doing /SWAP_IF_BIG_ENDIAN fails to work. To be specific, I want
the data to be in 'native' format on my Dec Alpha (oops - Compaq -ooopps,
HP), but to be readable/writable from a Sun workstation. If, on the Sun, I
do:
Read,1,array
Byteorder,/LSWAP,/SWAP_IF_BIG_ENDIAN,array
Then everything works
Dick
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
