Re: [TCLCORE] Fwd: TIP 672 Implementation Complete - Ready for Sponsorship

[Colin M. <col...@ya...>]

One point I forgot - a serious C implementation of this should probably 
disallow any further $ or [] substitution after substituting the bare 
variable names.  This could also have a secondary use-case as a safe way 
to run user-specified calculations.

On 01/11/2025 16:49, Colin Macleod via Tcl-Core wrote:
> On 30/10/2025 01:18, Andreas Leitgeb wrote:
>>    - [= ...] with a modified expr-language using barewords as variable
>>      names would probably solve many (if not most) of the cases, where
>>      [expr {...}] really "hurts" for its verbosity.
>
> Here is a pure Tcl implementation of that:
>
>     proc = args {
>         set ex [join $args]
>         set exp [regsub -all
>     {(::)?[[:alpha:]]([[:alnum:]_]|::)*([^[:alnum:]_(]|$)} $ex {$&}]
>         uplevel expr $exp
>     }
>
> This just finds all the possible variable names in the expression, 
> prefixes them with $ and then runs expr on the result.  The regex is a 
> little complicated because:
>
>   * The variable name could include :: at the beginning or in the
>     middle for namespacing.
>   * We disallow single colon : within the name because that creates
>     ambiguity with the ternary operator ?: .
>   * We also disallow array references because we can't distinguish
>     them from mathfunc calls.
>
> The fact that this can be implemented in pure Tcl shows that it 
> doesn't require any change to the dodekalogue rules, and so there is 
> no consequence for subst either.
>
> Of course a real implementation should be done in C.  It could then 
> also be possible to implement the extension Kevin Kenny suggested of 
> adding `=` as an assignment operator as in C within an expression, and 
> perhaps even chaining multiple assignments with `,` .
>
> I saw the `expr` question is on the agenda for the call on Monday.  I 
> don't usually join these calls but I would join it for this issue 
> except that I have an urgent dentist's appointment at that time.  
> Perhaps I should create a new TIP for this proposal?
>
> Colin.
>
>
>
> _______________________________________________
> Tcl-Core mailing list
> Tcl...@li...
> https://lists.sourceforge.net/lists/listinfo/tcl-core