Re: [Feature Request?] __key

["Soni L." <fakedme@...>]

On 18/07/16 04:21 AM, Tim Hill wrote:
> > On Jul 17, 2016, at 8:32 PM, Philipp Janda <siffiejoe@gmx.net 
> > <mailto:siffiejoe@gmx.net>> wrote:
> >
> > > The point is really this: __key modifies the contract for a table
> > > dramatically.
> >
> > What exactly is the contract for a table? Taking metamethods into 
> > account there is not much we can say for sure about the behavior of a 
> > table ...
>
> To my mind there is a huge difference between explicitly modifying 
> table behavior via a metatable, and having behavior changed as a side 
> effect of storing a userdata item as a key.

You don't seem to have any issues with integral floats. How are they any 
different?
> As I and others have noted, the functionality needed by the OP can be 
> done by interning, or by the appropriate use of metatables.

With returning plain integers (for interoperability with plain Lua keys) 
I'd have to use bigint.add(bigint(1), bigint(2)) instead of bigint(1) + 
bigint(2), which makes it literally impossible to write code that works 
with both bigints and plain numbers.

With __index and __newindex I can't use code I don't control.

> And as others have noted, this thread is really going nowhere and is 
> now drifting into ad hominem (not by you), so I’m dropping out.

This isn't ad hominem.
> —Tim

--
Disclaimer: these emails may be made public at any given time, with or without reason. If you don't agree with this, DO NOT REPLY.