Re: struct.el -- a package to encode/decode binary data

[Kim F. Storm]

Eli Zaretskii <address@hidden> writes:

> [I think we should conduct this discussion on emacs-devel: it might be 
> useful for others.]
I've CC:ed the list.

> 
> On 19 Mar 2002, Kim F. Storm wrote:
> 
> > I used string-make-unibyte because I was having a great many
> > difficulties with "struct-pack" generating what seems to be a bogus
> > string, i.e. if I used `length' on the result, it didn't even have the
> > expected length.  So it didn't DTRT for me in this case.
> 
> You would have to show fragments of Lisp for me to see what could be the 
> problem.
> 
> > I did manage to control that via the current buffer's coding system,
> > but that seemed to make pretty random results, depending on what
> > buffer I happened to eval the function in (e.g. my source buffer
> > vs. *scratch* vs. a process buffer).
> 
> This seems to indicate that you forced Emacs to convert from unibyte to 
> multibyte and back.  When you do that, it uses the current defaults to 
> guess what you mean as TRT.  That's why I said we shouldn't stress this 
> ad-hoc'ery too much--it's too fragile.

The problem probably was that I took some binary data from one buffer,
put it into a (multibyte) string, and then tried to unpack that string
in another buffer.  It didn't work - probably because those buffers
had different coding systems.

Since I really wanted this to work on binary, byte-oriented data,
I thought unibyte would be the correct mantra for me to use (and
it does work for this purpose).

> 
> > So I probably just have to warn users that this package will only work
> > on binary data if the current buffer's coding system is raw-text-unix.
> 
> I don't understand this--when you read binary data, what other decoding 
> can you possibly use in the buffer into which the data is read?  

True.  It just puzzles me why I always seem to run into problems when
emacs DTRT with this coding stuff.  It *was* easier in the old days,
and my mind probably just can't cope with this "new" coding stuff :-)


>                                                                    Why 
> should users bother about what kind of buffer are you using behind the 
> scenes to convert binary garbage into human-readable description?  What 
> am I missing?

In this scope, the `users' I was talking about are really `programmers'.
So `real users' definitely shouldn't bother -- and they will probably be
unaware of the existence of the struct.el package too.

> 
> > To me it would make sense to have a coding system named `binary' which
> > would work across all platforms.
> 
> `binary' _does_ work this way.  My problem was not with the coding system 
> you used, it was with the fact that you used unibyte strings.  That's a 
> different issue.
> 
> > Besides setting the buffer's coding system, is there some other way to
> > ensure that I'll always *use* a coding system which ensures that
> > `(substring ... 4 5)' really does only take one *byte* from the
> > string?
> 
> `substring' is a string operation.  Your data is not text, so you 
> should avoid string operations, I think.  Again, it's hard to give 
> specific advice without a specific example.  In general, functions like 
> char-after are much better.

My code is typically used in a network process filter, and in that
scope, the received (binary) data is delivered in a string, so yes, my
data isn't text, but it stored in a string...

In any case, I'll follow your advice and rely on using the proper
coding systems for the buffers and processes involved (and keep my
fingers crossed :-).

-- 
Kim F. Storm <address@hidden> http://www.cua.dk