________________________________________________________________________________
It seems rude to call something a code smell the instant (or… two years before the instant?) it gets replaced. Though it is C++, where you're apparently never supposed to use several of the features.
Although, I mean, iostreams is pretty bad.
It’s easy to pick on C++, but you can’t have backward compatibility and foreword progress and also not have features that are considered out-of-date.
Well that's fine, but it would be nice to have more official guidance, like editions/profiles/versions that turn things off as well as on. There are some standards, but mostly you just get self-proclaimed C++ experts telling you you suck if you write a `new` statement and don't use whatever kind of smart pointer is the right one now.
There already is such a thing and its called the C++ Core Guidelines:
https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines
Using shift operators for reading and writing to streams never made sense to me
What operator would you have used?
None. Just use a simple function call like almost every other language. Why is
std::cout::write(myInt)
such a hard thing to write that we decided
std::cout << myInt
was a better idea?
Disclaimer: I dislike iostreams and in our environment <iostream> is completely banned.
However, in C++98, how else would you handle when you have an arbitrary number of things to print in a type-safe way? e.g.
cout << myInt << ": " << something << " - " << otherthing << " - " << etc << '\n';
printf is not type-safe, nor supports user defined types. I'd rather not write 8 lines of functions calls for the same thing either.
I think it made some sense C++98, though it is an awful API in 2021.
std::cout::write(myInt,": ",something," - ",otherthing," - ",etc,'\n');
Possible with preprocessor.
I'm not sure how that could be done before variadic templates were introduced in C++11.
I doubt that's possible with the C++98 preprocessor.
If I have to guess, it's probably to support operator overloading (printing custom data type) without requiring OOP e.g. must be a class that implements Writable
> What operator would you have used?
Qt does it right: fluent-style interface which does string interpolation.
Is that localizable, or does it compile the interpolated string into code so it can't be replaced?
Qt has a pretty extensive localization framework. You would probably want to wrap both strings in tr().
https://doc.qt.io/qt-5/i18n-source-translation.html
Again, C++ solves problems it itself created.
Someone needs to write a book "C++ The Good Parts", similar to Douglas Crockford's "Javascript the good parts".
C++ is insanely bloated and complex. If there is a book distilling the best and cleanest parts of the language, I might give it another look. Haven't considered it in earnest since College.
Finally, now they just need to add networking then C++ will have a “good enough” standard library.
Why do you feel it's a good idea to shove networking into the standard library? Doesn't your choice of network library depend on other tech decisions regarding things like sync/async IO?
And moreover, if you want a java-like standard library that's good enough then you already have a bunch of options like Qt, Boost, POCO, etc.
Because every other PL has networking now. Because the lack of a networking framework fragments the C++ library ecosystem and prevents the development of re-useable and interoperable higher level libraries. The lack of such interoperable libraries make C++ a third-class citizen in the choice of language for middleware development.
Wouldn't you say the same thing about file io, or, just about everything?
Networking is an extremely standard need, making the standard library a good place for it.