On Sep 12, 2007, at 8:48 AM, Matthias WçÄhter wrote:
> On 12.09.2007 13:45, F. Senault wrote:
>> You should look in the 3rd part, chapter 22, under "The Basic Types".
>> Table 22.2, page 306 (in my PDF edition).
>>
>> It says :
>>
>> |     Table 22.2. Substitutions in double-quoted strings
>> | --------------------------------------------------------------
>> | \a Bell / alert (0x07)                \nnn     Octal nnn
>> | \b Backspace (0x08)                   \xnn     Hex nn
>> | \e Escape (0x1b)                      \cx      Control-x
>> | \f Formfeed (0x0c)                    \C-x     Control-x
>> | \n Newline (0x0a)                     \M-x     Meta-x
>> | \r Return (0x0d)                      \M-\C-x  Meta-control-x
>> | \s Space (0x20)                       \x       x
>> | \t Tab (0x09)                         #{code}  Value of code
>> | \v Vertical tab (0x0b)
>> | --------------------------------------------------------------
>>
>> So, your example uses the octal representation.
>
> Does anyone know a good reason for outputting 8 bit byte characters
> as octals in String#inspect?
>
> see string.c (1.8.6):
>
> [...]
> 	else {
> 	    sprintf(s, "\\%03o", c & 0377);
> 	    rb_str_buf_cat2(result, s);
> 	}
> [...]
>
> why not make it:
>
>             sprintf(s, "\\x%02x", c & 0377);
>
> I can't understand why it is desirable to introduce yet another base
> that is rarely used outside of chmod and od (without options). We
> learned decimals, we got used to binary and even hex, but why do we
> need octals just for these control characters when there is no
> benefit compared to the hex representation (note: both take 4 chars
> to display)?
>
> - Matthias

This is pure history.  The use of octal notation predates hex by a  
long time.  In my "The C Programming Language" [1978], 0ddd and 0xddd  
were both valid integer literals, but only octal \ddd was valid for  
character constants.

In fact, there's a table of internal sizes for data types on various  
machines which includes the Honeywell 6000 having a 9-bit char type  
and 36-bit types for short, int, long, and float and 72-bit doubles.   
For that machine, octal representation of bit patterns makes perfect  
sense.

-Rob

Rob Biedenharn		http://agileconsultingllc.com
Rob / AgileConsultingLLC.com