endian question

[James Leddy]

I run an intel i686 processor.

I just wanted to know weather it mattered because of endian how I put data
in a file. If I open a FILE *, say conf, and then I fwrite() to it, will I
mave to make any conversions if I want to fread() from it?

example by c code:
//assume conf is opened for writing and is a new file
int j = 12793;
fwrite(&j, sizeof(int), 1, conf);

//conf is same file, opened for reading and at the start
int j
fread(&j, sizeof(int), 1, conf);
printf("%d", j); //should output 12793

It is my understanding that I should be able to do this without
a problem, but because I have little endian, the actual numbers will be
stored when read in a hex editor as somthing else. (ie 12793 = 0x0000n31F9
will be stored as F9310000) Will this affect the program if I stick to
fread()ing and fwrite()ing?

Thanks
--
-J. Leddy (a.k.a. iustitia)

[Marc Rochkind]

On Wed, 16 Jul 2003 02:32:49 GMT, James Leddy <iustitia@optonline.net>
wrote:

> I run an intel i686 processor.
>
> I just wanted to know weather it mattered because of endian how I put
> data
> in a file.

[snip]

It matters for the same reasons and under the same circumstances as it
matters when you're passing data over a network. Namely:

1. Never for characters, but always for binary data that's more than a
byte.
2. Never when you're internal to a machine, but possibly (which means you
have to account for it) when going between machines.

So... feel free to write binary data however you like if the binary file
stays within a machine, or another machine with the same hardware. But,
take endian into account if you're writing a binary file that's supposed to
be portable.

For this reason, binary interchange formats like, say, TIFF, record the
endian state in the file, and the reading program must take it into
account.

Or... use a 100% character based format, like, say, XML, and the problem
goes away.

--Marc

[Valentin Nechayev]

>>> Marc Rochkind wrote:

MR> It matters for the same reasons and under the same circumstances as it
MR> matters when you're passing data over a network. Namely:
[...]
MR> 2. Never when you're internal to a machine, but possibly (which means you
MR> have to account for it) when going between machines.

There are little exceptions for some unusual;) architectures as ia64 and
power[pc]: endianness can be switched on them, and different OSes can use
different endianness.

MR> 1. Never for characters, but always for binary data that's more than a
MR> byte.

UTF-16 have two different endiannesses.

[...]
MR> Or... use a 100% character based format, like, say, XML, and the problem
MR> goes away.

XML with UTF-8, of course.


-netch-

[Erik Max Francis]

Valentin Nechayev wrote:

> UTF-16 have two different endiannesses.

True enough, but any sane UTF-16 stream is going to have the endianness
marker in it to eliminate any question.

--
Erik Max Francis && [email]max@alcyone.com[/email] && [url]http://www.alcyone.com/max/[/url]
__ San Jose, CA, USA && 37 20 N 121 53 W && &tSftDotIotE
/ \ You are inspiration to my life / You are the reason why I smile
\__/ India Arie