[ruby-talk:4853] Re: FEATURE REQUEST: Fixnum bitfields

Yukihiro Matsumoto wrote:

> Hi,
>
> In message "[ruby-talk:4849] FEATURE REQUEST: Fixnum bitfields"
>     on 00/09/12, Wayne Scott <[e-mail]> writes:
>
> |why not:
> |
> |fixnum[bit,len]    access a bitfield in number
> |bignum[bit,len]    access a bitfield in number
>
> Do you mean 0xbabe[16,8] returns 0xba?
> Hmm, is this useful?  Or just for completeness?
>
> |fixnum[Range]      access a bitfield in number
> |bignum[Range]      access a bitfield in number
>
> Same as above.  I agree with your friend.  Persuade me.
>
> |While I am on the subject of bitfields where is another question.
> |Supose I read a series or raw 256 bit data chunks from a file into
> |Strings.  How can I transfer them to Bignum's so i can do bitwise ORs
> |and ANDs on them.  (I want to convert my CRC checksum generator for
> |CPU caches from perl to Ruby.)
>
> def bits2num(str)
>   n = 0
>   str.each_byte do |c|
>     n <<= 8
>     n |= c
>   end
>   n
> end
>
>                                                         matz.

Some implementation of bit fields would certainly be nice.  Personally, I
lean toward packed arrays of boolean, but something that could extract a
set of flags from a word in one pass is a bit more flexible.  The
proposal of syntax makes sense, and increasing the symmetries is good.

OTOH, this isn't a frequently needed feature, so perhaps library routines
are a better approach.  Still, there's lots of things that are best
implemented as sets, and bit strings are the classic way to implement
small sets.  Sometimes making a feature available opens up whole new
styles of programming.  Look at what making grep style patterns available
within the language did to the way they are used.