Question about String.unpack


#1

Hi,

Say i’ve a C struct like that:

typedef struct {
char name[30];
double size;
int age;
char stuff;
} enreg_t;

This record is written in a file that i want to re-read with a Ruby
script (just a toy example… to understand).

I’ve not found a “clean” way to unpack such a record…

My C compiler gives me a total sizeof of 48, so i use a
fd.sysread(48).unpack(A32dIC) to re-read. But that seems rather
‘tricky’ (i suppose we don’t have always such a
information). Furthermore, the first element is completed with
garbage if the name field have less than 30 “useful” chars (“Doe”,
for example).

I’m sure i’m missing something but i’ve not managed to find the right
way to do that.

Any clue?


#2

On Feb 21, 2006, at 9:28 AM, Eric J. wrote:

for example).

I’m sure i’m missing something but i’ve not managed to find the right
way to do that.

Any clue?

Eric J., ne il y a 1443970729 secondes

there isn’t really a “clean” way to do this, since you are doing
something inherently un-clean.
However, to fix the string issue, a possible solution is:

str.gsub!(/\000.*/, ‘’)

assuming of course that name is a c-string.


#3

I’m curious about binary protocol parsing in ruby and have scratched
around for some kind of BinaryReader-ish interface for strings. Seems
like you could solve this if you know that your struct is in a string
you could say:

str = fd.sysread(n)
name = str.read_str(30)
size = str.read_double
age = str.read_int
stuff = str.read_c

I was going to write a native extension to do this after I finish my
current Ruby/AIO extension project. Unless there is something that
does this already (maybe in a cleaner/more ruby-ish way?)


#4

On Feb 21, 2006, at 2:19 PM, Jacob Repp wrote:

I was going to write a native extension to do this after I finish my
current Ruby/AIO extension project. Unless there is something that
does this already (maybe in a cleaner/more ruby-ish way?)

This is pretty much exactly what unpack does anyway., except its one
method that uses a format string. If you wanted a more stream
oriented wrapper around this, you certainly could write one fairly
easily.

ri String#unpack


#5

On Feb 21, 2006, at 6:28 AM, Eric J. wrote:

script (just a toy example… to understand).

I’ve not found a “clean” way to unpack such a record…

My C compiler gives me a total sizeof of 48, so i use a
fd.sysread(48).unpack(A32dIC) to re-read. But that seems rather

Your struct says char name[30] not 32…

‘tricky’ (i suppose we don’t have always such a
information). Furthermore, the first element is completed with
garbage if the name field have less than 30 “useful” chars (“Doe”,
for example).

Did you initialize the C struct correctly? You should zero out all
of name.

I’m sure i’m missing something but i’ve not managed to find the right
way to do that.


Eric H. - removed_email_address@domain.invalid - http://segment7.net
This implementation is HODEL-HASH-9600 compliant

http://trackmap.robotcoop.com


#6

On 2/22/06, ts removed_email_address@domain.invalid wrote:

“E” == Eric H. removed_email_address@domain.invalid writes:

My C compiler gives me a total sizeof of 48, so i use a
fd.sysread(48).unpack(A32dIC) to re-read. But that seems rather

E> Your struct says char name[30] not 32…

Yes, but there is alignement

He just want Z32 rather than A32

I would suggest that Z30xx would be a better choice.

pth


#7

“E” == Eric H. removed_email_address@domain.invalid writes:

My C compiler gives me a total sizeof of 48, so i use a
fd.sysread(48).unpack(A32dIC) to re-read. But that seems rather

E> Your struct says char name[30] not 32…

Yes, but there is alignement

He just want Z32 rather than A32

biocluster1% cat a.c
#include <stdio.h>
#include <string.h>

typedef struct {
char name[30];
double size;
int age;
char stuff;
} enreg_t;

main()
{
FILE *fd;
enreg_t enreg;
strcpy(enreg.name, “abc”);
enreg.size = 12.0;
enreg.age = 12;
enreg.stuff = ‘a’;
fd = fopen(“aa”, “w”);
fwrite(&enreg, sizeof(enreg), 1, fd);
fclose(fd);
}
biocluster1%

biocluster1% ruby -e ‘p a = IO::read(“aa”); p a.unpack(“Z32dIC”)’
“abc\000\030\373\377\277U\203\004\010tp\001@\360z\001@8\373\377\277K\205\004\010\200\270\024@\000\000\000\000\000\000(@\f\000\000\000ae\001@”
[“abc”, 12.0, 12, 97]
biocluster1%

Guy Decoux


#8

ts removed_email_address@domain.invalid writes:

E> Your struct says char name[30] not 32…

Yes, but there is alignement

He just want Z32 rather than A32

Thanks…

BTW: Eric, your message was not propagated to my Usenet Server…


#9

On Wed, Feb 22, 2006 at 04:19:07AM +0900, Jacob Repp wrote:

I was going to write a native extension to do this after I finish my
current Ruby/AIO extension project. Unless there is something that
does this already (maybe in a cleaner/more ruby-ish way?)

Yes, I’ve needed one of these too (read + write). My application was
handling a binary UDP protocol. String#unpack doesn’t support some of
the bizarro data formating that this protocol seems to have, so I cooked
up ReadBuffer/WriteBuffer[1] classes with a few methods a bit like those
above.

…just my 2-pence-use-case.

dave

[1] http://svn.badgers-in-foil.co.uk/alchemic/trunk/na_protocol.rb