joaquintides / boost_epoch Goto Github PK

The geometry library is really a C++14 library, not C++03. Attempting to test it with C++03 there are many errors because C++11 and C++14 constructs, and there are warnings which say that it will soon be a C++14 library. I know it claims C++03 compliance right now but I am not going to bother to report to them all the errors because of C++11 and C++14 constructs, since it will soon be C++14.

Hi @grafikrobot

I wonder if you'd be willing to add some detail on the tooling section so as to provide a more concrete idea of the changes needed there, with a focus on the binary name qualifiers for BOOST_ASSUME_STD (renamed from BOOST_MIN_EPOCH, you may want to re-read the proposal).

Thank you for your help!

(ongoing commentary)

There is an unacceptably high level of internal dependencies among libraries ("pulling one library drags most of Boost in").

This one could be true, or it might not be. But at least the "unacceptably" aspect is subjective. What one user finds acceptable depends on their needs. Someone who wants to use Boost.Beast is likely willing to accept a large set of dependencies. Yet someone who wants Boost.Any is likely unwilling to accept any (yes, pun intended) dependency. Measuring what the "weight" of a library is would be helpful in various ways. It would help inform, and likely dispel, the view of what is acceptable for libraries. It would help library authors understand and, perhaps, work towards improving the footprint of libraries. There are various ways to model this measurement also. It could be in terms of the super-set of library dependencies, and/or the amount of total code of the super-set, and/or the compiler performance complexity of using the code of a library (this one is not so easily achieve). I think this is a valuable enough measure that I'm willing to work towards generating it (grafikrobot/boost_lib_stats#1).

The callable_traits library is a C++11 library. Therefore it is not c03.

The convert library is a C++11 library. See boostorg/convert#48.

(ongoing commentary)

Much stuff in Boost is now useless as it was adopted into the C++ standard itself.

This is a supposition not likely based on fact. I doubt that "much" is either in the standard itself nor useless. To determine this we would need to count what percentage of Boost was adopted into the standard in comparison to the rest of Boost. Of course any such comparison would need to be multidimensional as measuring utility is subjective. You could attempt to use objective measure to compensate. For example, number of code lines, number of downloads (hard to determine), number of stars, number of followers, number of reported issues, perhaps even search results of referenced headers in open source software.

The qvm library is a C++11 library. The documentation specifically says so.