Re: [proposal] a new keyword : final

["J.Jørgen von Bargen" <jjvb.primus@...>]

Am 23.03.2010 19:27, schrieb David Given:
> Unless it becomes utterly trivial --- that is, the const-ness applies to
> the variable only, and not the contents --- you very quickly start
> straying down the path towards a full type system...
>    
I'd like to say, the const-ness at least applies to to type-nees of the 
variable :-)
And also to the contents, if the value itself is immutable. So:

final x=true  will always be of type boolean, and will always be true 
(true is immutable)
final y="YES" will always be a string and will always be the same string 
(strings are immutable [PiL 2.4] )
final t={} will always be a table (and will always be the same table (!) 
), but the table itself is mutable, so t.x=1 is fine, even t[#t+1]="x" 
is perfect legal.
final fd=io.open("text.txt","w") will always be a userdata (and always 
the same userdata), but the underlying c-data (in this case the 
file-state) may be altered with
fd:write(...) or fd:close()

Conclusions:
final x=false if x then something() end -- eligible for optimization
final t="YES" if t=="no" then something() end -- eligible for optimization
final y={} if y then something() end -- too! (y will never be nil or false)
final z={} if #z>0 then something() end -- not (because the table is 
mutable so #z can vary)

I dont know, how complicated this would be to implement (as a pure 
lua-application-writer I did not look into luac yet), but this looks 
promising for me.

Regards Jørgen