Yes a good talk. Amazing that there is no mention of Rust though. Which does exactly this. Rust "is" what C++ seems to want to be: Concepts, Algebraic types, Move semantics, Associated types... etc. already baked in.
No Rust is a new implementation of C++ best practices anno 2016, making those best practices native to the core language instead of optional new features with increasingly obscure syntax.
I would like to argue that your example of the "should not usually happen" case i.e. "out of memory" is actually a "might happen" case. I reason as follows: When your program wants more memory it sends a message to the operating system to request it. That is "new" in C++ which is actually good old "malloc" from C at the end of the day. The operating system may well refuse the request. I make no distinction between that and failure to get any other resource, like create_dir for example. Of course it's even tougher to deal with because Linux, for example, will return "OK" for such a memory request even if it does not have the memory. Your process will get killed with a page fault when it actually tries to use the non-existent memory. No amount of error/exception handling in your code will help you. More generally, your talk has convinced me that C++ has "jumped the shark" and is about to implode. No normal human can comprehend it's complexity. When it's easier to get the job done in a knowable, predictable, reliable fashion in assembler something is very wrong with the high level language. We could never have put man on the moon using C++.
Yes a good talk. Amazing that there is no mention of Rust though. Which does exactly this.
Rust "is" what C++ seems to want to be: Concepts, Algebraic types, Move semantics, Associated types... etc. already baked in.
No Rust is a new implementation of C++ best practices anno 2016, making those best practices native to the core language instead of optional new features with increasingly obscure syntax.
Great presentation.
I would like to argue that your example of the "should not usually happen" case i.e. "out of memory" is actually a "might happen" case.
I reason as follows:
When your program wants more memory it sends a message to the operating system to request it. That is "new" in C++ which is actually good old "malloc" from C at the end of the day.
The operating system may well refuse the request.
I make no distinction between that and failure to get any other resource, like create_dir for example.
Of course it's even tougher to deal with because Linux, for example, will return "OK" for such a memory request even if it does not have the memory. Your process will get killed with a page fault when it actually tries to use the non-existent memory. No amount of error/exception handling in your code will help you.
More generally, your talk has convinced me that C++ has "jumped the shark" and is about to implode. No normal human can comprehend it's complexity.
When it's easier to get the job done in a knowable, predictable, reliable fashion in assembler something is very wrong with the high level language. We could never have put man on the moon using C++.
God, is he collecting new terrible coding style fads as a hobby?
Otherwise awesome talk. sorry for nitpicking ;)
So many mouth sounds....