C++17 feature proposal: Allow if constexpr()

[Jayson Adams]

Hello,

I was looking through the codebase today and happened to run across the following comment in checked_math_impl.h:

// TODO(jschuh) Make this "constexpr if" once we're C++17.

This seems like a reasonable thing to do here but I actually don't have strong feelings about it. From a codebase hygiene standpoint, now that C++17 is available I'd like to move forward with this TODO or remove it. Making a decision to allow or ban if constexpr() is the first step.

Best,

__jayson

[Avi Drissman]

+1 for this; this would be a strict improvement.

--
You received this message because you are subscribed to the Google Groups "cxx" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cxx+uns...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/cxx/CAC2KseO8z7JUbjpM_avE0je01x7s_jG%3D--r_Xq1sVBBoAtupxA%40mail.gmail.com.

[Jeremy Roman]

Strictly speaking we would also need to wrap the remainder of the function in an else branch (contrary to common style recommendations to the contrary), since the rules don't discard statements due to early-return AFAIK. Right?

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/cxx/CACWgwAZkY9ab2UBRH4dvsPW3BcGkW-c49H3tUAUHv7ybEqry4w%40mail.gmail.com.

[Peter Kasting]

I was about to write this proposal since I have a local cl using constexpr if here and in a few other places :)

+1

PK

[Peter Kasting]

On Thu, Jan 6, 2022 at 1:27 PM Jeremy Roman <jbr...@chromium.org> wrote:

Strictly speaking we would also need to wrap the remainder of the function in an else branch (contrary to common style recommendations to the contrary), since the rules don't discard statements due to early-return AFAIK. Right?

At least in this case, wrapping the remaining function in "else" has no effect on the generated code size.

PK 

[Jeremy Roman]

Personally I'm looking forward to cleaning up stuff like this.

[Jayson Adams]

3370526: [c++17] Add "if constexpr" to the allowed C++17 feature list | https://chromium-review.googlesource.com/c/chromium/src/+/3370526

[Roland Bock]

+1 

"if constexpr" can make a massive difference in readability of C++ template code.

--

You received this message because you are subscribed to the Google Groups "cxx" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cxx+uns...@chromium.org.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/cxx/CAC2KsePGH_wgnbn1J%3DMeOpTzs8qkXPkBUuLm0s%2BffU8%2B_Phc3w%40mail.gmail.com.

[Jayson Adams]

The change has landed.

[Victor Vianna]

Curious: what's the guidance on using this new feature vs ifdefs? It seems like it could fundamentally replace ifdefs. If so, do we intend to migrate things like the very common "#if defined(OS_FOO)"?

[Jeremy Roman]

I wouldn't expect it to significantly displace #if, because it only works in certain contexts. I think it's more often going to be a replacement for using additional layers of templated functions/structs to avoid compiler errors.