I've noticed he doesn't know everything, but he knows a lot.
There are exceptions to statements like this.
I'd change "never" to "almost never".
I've noticed he doesn't know everything, but he knows a lot.
There are exceptions to statements like this.
I'd change "never" to "almost never".
In a personal conversation, that person told me that he estimated that he knew about half of the language, and that average career length wasn't enough to learn all of it.
This was way before C++11.
I've been working primarily in C++ since the days when there weren't even C++ compilers (you instead ran the C++ code through a program that turned it into C that you compiled with a C compiler). By any reasonable measure, I am proficient in the language -- but I think I'm actually proficient in about half of the language, same as your coworker.
This is also why I've moved away from C++ in the last few years. It's too unwieldy for me to use as a default language anymore. Now, my default language has become C++-as-a-better-C rather than full-blown C++.
For one thing, when a language feature doesn't really work out, the user community will notice and it'll disappear from most codebases. If a Java developer with 20 years of experience has never worked with java webstart - is their mastery incomplete?
And even for features that aren't deprecated, almost no jobs will demand every feature a language offers. A veteran Java developer with 20+ years working on high-performance backend server code might not know much about JavaCard, or packaging their code for use on Windows, or the state of the art in making GUIs, or the latest kitchen-sink enterprise framework - is their knowledge incomplete?
Of course, some would say if your definition of "knowing a language" is so demanding that nobody meets it, some might say that's an unrealistic bar....
But this is true of almost any popular language! By the metric people are applying to C++ here, there is no one on earth who "knows Java", which is patently ridiculous.
A better metric would be "Can write code in that language for some domain".
Can you though? C++ chooses to sidestep Rice's Theorem with IFNDR (Ill-formed No Diagnostic Required). IFNDR is sprinkled on lots of the ISO Document and what it basically does is, everywhere the language needs semantic constraints for soundness Rice's Theorem would otherwise make it theoretically impossible to write a C++ compiler so they just say well, if you break this semantic constraint you haven't written a C++ program. The compiler won't know, and neither will you, but it's not our fault if your program doesn't work, it was actually never a C++ program anyway.
This is a great rhetorical device - if the goal was to win a debate championship this sort of sidestep is very clever. But obviously it's going to produce a vast amount of nonsense, it's terrible engineering practice, and yet for decades people acted like this was OK.
Also, this graph is from 2010, so it's missing some lambda magic, auto vomit, and static object initialization got easier since.