Re: preliminary patch for zsh/random module

[Clinton Bunch <cdb_zsh@zentaur.org>]

On 10/23/2022 11:47 PM, Bart Schaefer wrote:
> On Sun, Oct 23, 2022 at 7:57 PM Clinton Bunch <cdb_zsh@zentaur.org> wrote:
>> On 10/23/2022 9:01 PM, Bart Schaefer wrote:
>>> * Typo in the introductory comment.
>> I'm pretty sure I've since found and changed all the instances of
>> Zoltan's name.
> I was referring to the spelling of "randome".
>
>>> Actual question:  What's the use case for returning or printing a
>>> block of random bytes?  Why does this need to be a builtin?
>> Mostly because I see constructs like read -k6 -u3 3</dev/urandom
> ... because?

I did it to seed rand48.  Roman did it to generate a 32-bit random 
integer.  Both of which are obsoleted by other parts of this module.

I'm sure there are other reasons someone might want to read more or less 
than 4 bytes of random data.  One that comes to mind is generating a 
password in a platform-independent way without assuming perl or python.

>
>> it would be nice to initialize an array with random numbers without
>> having to use a loop to access SRANDOM n times.
> But you're not filling an array with random numbers, you're filling a
> string (scalar) with random bytes.
Actually, it does both.
>
>>> Suggestion:  Treat SRANDOM like SECONDS, in that you can change the
>>> type from integer to floating-point.  Then maybe the zrandom() math
>>> function isn't needed?
>> That would seem confusing to me, and too easy to forget which state you
>> left it in.
> You make it local so you're not leaving it.
>
> () {
>   print $SECONDS;
>   () {
>    local -F SECONDS
>    print $SECONDS
>   }
>   print $SECONDS
> }
> 56
> 0.0000050000
> 56
Still seems like something you do because you can rather than because 
it's intuitive to the user.  Also, your example assumes you'd only want 
to do this in a function (or wrap it in an anonymous function just for 
this purpose)
>
>> zrandom was meant to be a replacement for rand48
> OK.
>