[fpc-devel] RFC: Support for new type "tuple" v0.1

[Mark Morgan Lloyd]

Sven Barth wrote:

>>> b : tuple of (integer, integer);
>>> a : integer;
>>>
>>> begin
>>>   b:=(1,2);
>>>   a:=b[0]; // a = 1 after this.
>>> end;
>>
>> Iff a tuple contains elements of the same type and if tuples with this
>> restriction are compatible with arrays, wouldn't this behaviour be
>> implicit?
>>
> 
> The [] operator would then also be allowed for tuples with different types.
> Also I'd make tuples only assignment compatible to arrays (thus you 
> could not use the array's [] operator on a tuple). But it seems anyway 
> that the extensions aren't favored that much...

Cautionary tale follows, slightly OT. I don't know whether you've ever 
looked at APL but given an array X you can either select an element 
X[1,2] or you can specify the significant axis [2]X  Somebody tried to 
write a parser using YACC and found this was ambiguous, they ended up 
having to recognise one of the cases in the lexer and (in effect) change 
the brackets into something else. Fun language: you can parse it 
left-to-right but have to remember that it executes right-to-left.

-- 
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]