| Re: little and big endian -- once more [message #15329] |
Thu, 13 May 1999 00:00 |
Dick Jackson
Messages: 347
Registered: August 1998
|
Senior Member |
|
|
Hi,
Martin Schultz wrote:
>
> Hi all,
>
> took me a while to realize that it is the machine architecture not
> the OS that determines the byte swapping -- in fact I needed to have IDL
> on linux (after using it on an SGI and with Windows) to figure that out
> ... Anyway, I now devised the following test for byte swapping which is
> applied in my open_file routine and handles everything transparently.
I'm not sure if I'm missing some subtlety here, but wouldn't the (newish)
keywords /SWAP_IF_[BIG|LITTLE]_ENDIAN to IDL's OPEN and BYTEORDER routines
work in all cases? I've used them to good effect for cross-platform work.
From the help file:
[...] it only takes effect if the current system has [big|little] endian
byte ordering. This keyword does not refer to the byte ordering of the input
data, but to the computer hardware.
Cheers,
--
-Dick
Dick Jackson Fanning Software Consulting, Canadian Office
djackson@dfanning.com Calgary, Alberta Voice/Fax: (403) 242-7398
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: little and big endian -- once more [message #15330 is a reply to message #15329] |
Thu, 13 May 1999 00:00  |
Liam Gumley
Messages: 473
Registered: November 1994
|
Senior Member |
|
|
Martin Schultz wrote:
> So, I guess, I should give Liam's suggestion a try.
> Unfortunately, it seems that you must know what's in the file in order
> to test it, but that's probably a good idea anyway when you are dealing
> with binary files ;-)
Yeah, it's pretty hard to read a binary file if you don't know the
format.
> Just for clarification: is it true that the
> endian matters for integers as well as floats?
Yes, it's true.
Here at SSEC we've messed with porting binary datafiles for years.
However in the last year or so, netCDF has become very widely used in
our applications. Some of the reasons are:
(1) No more byte-swapping worries, ever.
(2) The same IDL code can read the same datafile and give the same
results on any IDL platform (Unix, Linux, Windows, Mac).
(3) If you give someone the file and tell them it's netCDF, they can
then easily figure out what it contains. You don't need to tell them the
format details.
(4) The netCDF API is available for Fortran77/90, C, C++, IDL, Matlab,
perl, Python, and Java.
(5) The API is very simple and concise: the only things you can store in
a netCDF file are arrays, attributes, and dimensions. However, you can
create structures using these building blocks which are as simple or as
complex as you like.
(6) netCDF supports the following IDL datatypes: BYTE, STRING, INT,
LONG, FLOAT, DOUBLE.
More information is available at
http://www.unidata.ucar.edu/packages/netcdf/faq.html
---
Liam E. Gumley
Space Science and Engineering Center, UW-Madison
http://cimss.ssec.wisc.edu/~gumley
|
|
|
|
| Re: little and big endian -- once more [message #15338 is a reply to message #15329] |
Wed, 12 May 1999 00:00 |
Martin Schultz
Messages: 515
Registered: August 1997
|
Senior Member |
|
|
David Fanning wrote:
>
> Martin Schultz (mgs@io.harvard.edu) writes:
>
> [...]
>
> Here is a little function, Martin, that will make sure
> your code works without having to track every machine
> architecture down. Found it on my web page. :-)
>
> Cheers,
>
> David
well, that's not all you need. Of course I replaced my string('x86')
thing with the inline code of little_endian() and at first it seemed to
work fine. But I encountered one situation when the SWAP_IF keyword
would not detect things properly which is reading data from a mounted
network drive with different endian (i.e. a generic unix drive mounted
to a linux box). So, I guess, I should give Liam's suggestion a try.
Unfortunately, it seems that you must know what's in the file in order
to test it, but that's probably a good idea anyway when you are dealing
with binary files ;-) Just for clarification: is it true that the endian
matters for integers as well as floats?
Thanks again,
Martin.
--
|||||||||||||||\\\\\\\\\\\\\-------------------///////////// //|||||||||||||||
Martin Schultz, DEAS, Harvard University, 29 Oxford St., Pierce 109,
Cambridge, MA 02138 phone (617) 496 8318 fax (617) 495 4551
e-mail mgs@io.harvard.edu web http://www-as/people/staff/mgs/
|
|
|
|
| Re: little and big endian -- once more [message #15339 is a reply to message #15338] |
Wed, 12 May 1999 00:00 |
Liam Gumley
Messages: 473
Registered: November 1994
|
Senior Member |
|
|
Martin Schultz wrote:
> took me a while to realize that it is the machine architecture not
> the OS that determines the byte swapping -- in fact I needed to have IDL
> on linux (after using it on an SGI and with Windows) to figure that out
> ... Anyway, I now devised the following test for byte swapping which is
> applied in my open_file routine and handles everything transparently.
> Please tell me if there are any other machine architectures that are big
> endian and what their !version.arch tag would be.
The way I usually handle this is to let the *data* tell me whether
byte-swapping is required. For example, my colleagues generate datafiles
in host order using the same code on SGI and Linux boxes. Rather than me
keeping track of what platform was used to create the data, when I open
the datafile in IDL my code interrogates the data itself to see if
appears to be swapped. There is usually some data item (e.g. a date
and/or time) that can be checked to see if it's in the correct range. If
it's in the correct range, then proceed. If not, check swap_endian(date)
to see if it's in the right range. If it looks ok, swap everything. If
it doesn't look ok, then the data is not in the expected format.
I usually only apply this method to files that others supply to me. If I
want to create platform-independent datafiles, I use netCDF or HDF.
---
Liam E. Gumley
Space Science and Engineering Center, UW-Madison
http://cimss.ssec.wisc.edu/~gumley
|
|
|
|
| Re: little and big endian -- once more [message #15340 is a reply to message #15338] |
Wed, 12 May 1999 00:00 |
davidf
Messages: 2866
Registered: September 1996
|
Senior Member |
|
|
Martin Schultz (mgs@io.harvard.edu) writes:
> took me a while to realize that it is the machine architecture not
> the OS that determines the byte swapping -- in fact I needed to have IDL
> on linux (after using it on an SGI and with Windows) to figure that out
> ... Anyway, I now devised the following test for byte swapping which is
> applied in my open_file routine and handles everything transparently.
> Please tell me if there are any other machine architectures that are big
> endian and what their !version.arch tag would be.
Here is a little function, Martin, that will make sure
your code works without having to track every machine
architecture down. Found it on my web page. :-)
Cheers,
David
*****************************************************
FUNCTION Little_Endian
; Returns 1 if Little-Endian, returns 0 if Big-Endian.
little_endian = (BYTE(1, 0, 1))[0]
RETURN, little_endian
END
*****************************************************
David Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438 E-Mail: davidf@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
