Hacker News new | ask | show | jobs
Show HN: Nameof Operator for Modern C++ (github.com)
65 points by neargye 2650 days ago
5 comments
koyote 2650 days ago
This is a good use of the more modern C++ features such as string_view, constexpr and template metafunctions such as decay and enable_if.

I do wonder how much this would impact compile-times if used moderately in a large code-base.

link
neargye 2649 days ago
In our project, NAMEOF_TYPE and NAMEOF almost do not add time relative to the entire project. But NAMEOF_ENUM can add a couple of seconds to large enums.
link
CalChris 2650 days ago
This is similar to the C preprocessor trick STRINGIZE:

  #define STRINGIZE_DETAIL(x) #x
  #define STRINGIZE(x)        STRINGIZE_DETAIL(x)
link
gpderetta 2650 days ago
It probably uses stringification under the hood, but then relies on constexpr to "canonicalize" the name (removing everything before the dot, template parameters etc).
link
Someone 2650 days ago
From cursory reading, it seems it doesn’t, at all. It uses compiler-specific predefined values __PRETTY_FUNCTION__ (gcc, clang) or __FUNCSIG__ (MS Visual C++), and doesn’t work on other compilers.

It also has to do work to extract the function name from the char arrays/C strings that those macros expand to.

Reading answers to https://stackoverflow.com/questions/4384765/whats-the-differ..., it seems it would have been easier to use the __FUNCTION__ in gcc, but I’m not familiar with that (or the other predefined values) at all, so who knows?

link
anon4242 2650 days ago
Actually, __FUNCTION__ won't work as it only expands to the name of the function, not any templates or arguments.

It's actually a pretty clever way to automagically deduce not only names of variables but also of enums :). Too bad it's most likely not maintainable in the long run as compilers probably change what their __PRETTY_FUNCTION__ (or equivalent) macros expand to in new versions.

link
neargye 2649 days ago
I usually check the newest versions of compilers, and if the behavior changes, then I update the code
link
asveikau 2650 days ago
Last I knew __FUNCTION__ in GCC doesn't even do name de-mangling. You'd get the raw C symbol name with no C++ types.
link
neargye 2649 days ago
https://stackoverflow.com/a/31163837/9514814
link
asveikau 2649 days ago
Gcc 7. I could be wrong but I seem to recall a long time ago, when I was learning this stuff in the 90s, __FUNCTION__ wouldn't even demangle.
link
neargye 2649 days ago
You are absolutely right :)
link
asveikau 2650 days ago
That was what I was thinking, then I saw that this will stringify enum values, eg. nameof(c) returns "RED", which the preprocessor #x will not do.
link
neargye 2649 days ago
To be sure: nameof(c) returns "c" (nameof return name of variable) nameof_enum(c) returns "RED" (nameof_enum return enum name of enum variable)
link
CalChris 2650 days ago
I only use STRINGIZE for error messages. I learned about it from reading the source for assert macros.
link
gumby 2650 days ago
Nice work, and the header is nowhere near as complex as I had feared it would need to be.
link
neargye 2649 days ago
With features C++17 this header looks not so terribly.
link
shhehebehdh 2649 days ago
I wonder what including and using this does to compile times. Nothing good, I suspect.
link
neargye 2649 days ago
In fact, not so much about the whole big project. May add a few seconds to the compilation.
link
jhasse 2649 days ago
nameof_enum doesn't work with GCC :(
link
neargye 2649 days ago
Unfortunately, yes, I'm still looking for a solution for gcc
link
neargye 2648 days ago
New version supported GCC >= 9
link
jhasse 2648 days ago
Awesome :)
link