C++ Logo

STD-DISCUSSION

Advanced search

Subject: Re: Two issues about the description of the instantiation of fold-expression in [temp.variadic]
From: Andrew Schepler (aschepler_at_[hidden])
Date: 2021-03-14 00:07:35


On Mon, Mar 8, 2021 at 3:06 AM jim x via Std-Discussion <
std-discussion_at_[hidden]> wrote:
>
> In the current draft, there are two issues about the instantiation for a
fold-expression.
> The first is the standard seems not clear for stating which are the
elements of the pack expansion parameters that appear in the pattern of
fold-expression. It could be assumed they're stated in paragraph 8
[temp.variadic#8]
>
> However, such a layout of the rule seems that the description for what
these elements of pack expansion parameters are is only applying to the
case that is: `the instantiation of a pack expansion that is neither a
sizeof... expression nor a fold-expression`.
>
> Is it necessary to take the rule out from paragraph 8 to be an individual
rule? That would cover the case that what these elements of the pack
expansion parameters are which appears in the fold-expression.

I agree, it would make sense if things were rearranged so that
subparagraphs (8.1)-(8.3) also applied to fold-expressions. Perhaps
something like:

----
[Paragraph 8]:
<del>The instantiation of a pack expansion that is neither a sizeof...
expression nor a *fold-expression* produces</del>
<ins>The *expanded pack list* for an instantiation of a pack expansion that
is not a sizeof... expression is</ins> a list of elements...
[(8.3)] ... the init capture-pack.
[Add paragraph break and renumber.]
<del>All of the E_i become</del>
<ins>The instantiation of a pack expansion that is neither a sizeof...
expression nor a *fold-expression* produces the elements of the expanded
pack list as</ins>
items in the enclosing list.
[Here, keep the following note, normative sentences, and example from the
original paragraph 8.]
[Paragraph 10]:
In each case, *op* is the *fold-operator*, *N* is the number of elements in
the pack expansion parameters, and
<del>each E_i is generated by instantiating the pattern and replacing each
pack expansion parameter with its i-th element.</del>
<ins>the expressions E_i are the elements of the expanded pack list.</ins>
For a binary ...
----
>
> The second issue is about the **binary fold-expression** when the number
of the elements is zero. The whole paragraph 10 only talks about how the
result is for the unary fold-expression when the number of the elements is
zero. It lacks the case of a binary fold-expression. (Presumably, the
result of this case is that of the expression `E` that didn't contain an
unexpanded pack).
I don't think any changes are needed here. The descriptions of the
instantiations use the informal ellipsis as often seen in mathematical
formulas. The meaning of that type of formula, although seldom defined in
very strict detail, is very widely understood as a pattern giving a formula
involving K values/operands and K-1 operators, as long as K>=1. The K=0
case is in general not valid, but there can be other conventions, stated or
unstated, about what it means then, just as we see C++ defining explicitly
for unary fold-expressions with N=0. (In mathematical writing, it's often
assumed if the operation is called "addition" the result is the set's
appropriate "zero", and if the operation is called "multiplication" the
result is the set's appropriate "one".) So a binary fold-expression where
N=0 is a case of the informal ellipsis with K=1, and yes, the result is E.
This is no more or less ambiguous than a unary fold-expression where N=1,
since for interpretation of the informal ellipsis pattern, the fact that E
has different properties than the E_i expressions doesn't matter.
-- Andrew Schepler
> --
> Std-Discussion mailing list
> Std-Discussion_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-discussion


STD-DISCUSSION list run by std-discussion-owner@lists.isocpp.org

Older Archives on Google Groups