Talk:cpp/numeric/valarray

One question about:

"Most C++ standard library implementations use expression templates to implement efficient operations on std::valarray, with rare exceptions ..."

Is this true (e.g. for MSVC)?

134.76.223.3 06:52, 10 August 2012 (PDT)

Good question, I wrote most, but I only checked GNU's lihbstdc++ and LLVM's libc++. I don't have access to MSVC, but looks like IBM and Sun's libraries (based on Dinkumware and STLport, repesectively) return temporaries, for shame. Changed to be explicit on who implemented what. --Cubbi 08:07, 10 August 2012 (PDT)

Some suggestions on valarray pages

Pages about valarray seemed to be confused, because

• every Non-member functions says "applies the function std::xxx to each element of valarray". However, in its own page, "Unqualified function (xxx) is used to perform the computation".
• most of these non-member functions have a "Possible implementation" in their pages. But they look silly enough to manufacture due to nothing differences but functions' name. maybe using apply(xxx) to explain would be better?
• shall we replace return types of these member and non-member functions with /*see below*/ to highlight notes below?

I like the /*see below*/ idea and won't object to "Applies the function cos to each element" as long as it still links to std::cos (which is the function 99.999% of uses will apply). Possible implementations are silly because they are not returning expression templates, but who is up to the task of writing up a realistic yet readable ET? --Cubbi (talk) 07:19, 11 November 2019 (PST)

So my suggestion is

• remove all possible implementations in those pages
• add "this function performs as if va.apply(xxx)" to every note which is generated by a template
• provide a silly possible implementation in apply's page with comments saying something like "ET can be used here"

By the way, I found it evident that more templates can be used in those pages.

Yaossg (talk) 21:22, 14 November 2019 (PST)

Excessive template transclusion is deemed not a good idea. See Talk:Main_Page/Archive_2#Abandoning_complex_templates --D41D8CD98F (talk) 22:24, 14 November 2019 (PST)

However, there are at least 15 pages owning exactly the same content with only their name different. any change to one of them may need to apply 14+ another times! This number is even larger than container's, whose pages are still using templates to build. Yaossg (talk) 00:09, 15 November 2019 (PST)

The Progress

Now I've applied some changes in apply and abs. What do you think about them? Shall I make further changes (including creating a new template metioned in abs)? Yaossg (talk) 21:33, 7 December 2019 (PST)