search

ORNL / cpp-proposals-pub

Collaborating on papers for the ISO C++ committee - public repo
26 stars 26 forks source link

P1673: LWG review 2023/10/18 #416

Open mhoemmen opened 10 months ago

mhoemmen commented 10 months ago

P1673: LWG review 2023/10/18

vector_sum_of_squares

CT: We repeat Remarks a lot. Should we float this up? No, just leave it.

vector_two_norm

Para 2 mixes math with prose. Parens in the text are confusing and not needed. Rewrite as follows.

The square root of the sum of square of init and the squares of the absolute values of the elements of v.

Note creeps on normative. "Recommended practice" is the text we use, but should be fine as a Note. Change "suggested" to "possible."

Get rid of separate headers for "specified result type" vs. "default result type," generally throughout the norms etc.

For default result type, delete the arrow and "see-below." It's not observable if the implementation wants to spell out the return type.

What if abs-if-needed returns a proxy reference? That's OK, then the function fails to compile; that's what we want here.

vector_abs_sum

Regarding _GENERALIZEDSUM, why don't we do that with vector 2 norm?

We probably don't need it here, just math. Also need a math way to say "real-if-needed" and "imag-if-needed." At least, just drop _GENERALIZEDSUM and use words and/or math.

I think we could just say "the sum of init and the absolute values of elements in the domain of v", or just use a math formula.

We will come back next time with a formula version of this. If it looks better, we can use it.

2.2 is redundant but harmless. 2.3 suffices.

Remove section headers "specified result type" and "default result type."

Remove arrow and see-below for return type.

vector_idx_abs_max

Cpp17LessThanComparable semantics can't be checked in Mandates. Let U be decltype(...). Mandates declval<U>() < declval<U>() is a valid expression.

"Mandates: declval<T&>() < declval<T&>() is a valid expression" (but references?)

Or just say less-than comparable.

Also more parens in English, replace with math.

How much of using Cpp17LessThanComparable is about semantics rather than syntax? We require that it is a strict weak order, i.e., it can’t be actually mandated. So, “let T be decltype(...), Mandates: declval<T>() < declval<T>() is a valid expression.

...

Remove the parenthesis in (2.2) -- don't have a precondition -- keep the Mandates. Remove "(in the domain v)."

[linalg.algs.blas1.matfrobnorm]

Same changes as before: “see-below” and removing the subsections. Stopping before [linalg.algs.blas1.matonenorm].

Stopped after matrix Frobenius norm, before matrix one norm.

mhoemmen commented 10 months ago

"Frobenius Method is FOR DEFINITE a progressive rock band name."

mhoemmen commented 10 months ago

PR https://github.com/ORNL/cpp-proposals-pub/pull/417 fixes the issues above, except for exploring use of math formulas instead of _GENERALIZEDSUM.

mhoemmen commented 10 months ago

I've filed https://github.com/ORNL/cpp-proposals-pub/issues/418 as a follow-on issue to explore use of math formulas. PR 417 fixes everything else. Thanks! : - )