Re: Nil and false in Lua design

[Sean Conner <sean@...>]

It was thus said that the Great steve donovan once stated:
> On Mon, Oct 14, 2013 at 1:25 PM, Carsten Fuchs <carsten.fuchs@cafu.de> wrote:
> > I'm certainly not feverish about this, but for consistency's sake, I think
> > it would be nice to have either -- or none.   ;-)
> 
> Up to some point, consistency is overvalued.  

  "A foolish consistency is the hogoblin of little minds."
		-- Ralph Waldo Emmerson

> No point in breaking all
> those programs because implicit tostring was offending my delicate
> tastes - because tastes differ.  I'm with Dirk when he suggests an
> explicit tobool, then everyone can see what's happening and (NB) not
> paying for something they don't need.  This is a famous C++ principle,
> but then it hides so much implicit complexity (in theory, 'fire =
> true' could launch a rocket *) that one has to be extra awake.

  I deal with some pretty complex protocols [1] at work.  These protocols
tend to have lots (and lots) of optional parts.  I don't like it, but I
can't change that.  So, my code tends to have:

	if info.use_extension_a then
	  ...
	end

	if info.use_option_2 then
	  ...
	end

  If nil becomes non-boolean, then I would either have to change a lot of
code to:

	info.use_extension_a = false
	info.use_option_2 = false
	-- and a billion other options I don't care about

or

	if info.use_extention_a ~= nil and info.use_extension_a then
	  ...
	end

	if info.use_option_2 ~= nil and info.use_option_2 then
	  ...
	end

  In other words---it becomes a royal pain in the butt.  And for what?  A
bit of purity? [2]  

> (*) Silly example, but when dealing with others' code you learn to
> hate cleverness.  It remains true, of course, that 'rocket.fire =
> true' could do the same in Lua.

  True enough.  I have code where you can do:

	proc.limits.hard.cpu = 300

to limit the CPU usage of a program to 5 minutes [4]; underneath, it's a
call to setrlimit().

  -spc 

[1]	SS7, which is used in telephony.  

[2]	Even Haskell had to ditch 100% mathematical purity just so we could
	get useful programs [3]

[3]	Monads solely exist to make Haskell useful.
	
[4]	That's 300 seconds of CPU usage, not 300 seconds of wall clock time. 
	There is a difference.