The graph goes up for me when I find my comfortable little subset of C++ but goes back down when I encounter other people’s comfortable little subset of C++ or when I find/remember another footgun I didn’t know/forgot about.
When I became a team leader at my last job, my first priority was making a list of parts of the language we must never use because of our high reliability requirement.
Sure, strtok is a terrible misfeature, a relic of ancient times, but it’s plainly the heritage of C, not C++ (just like e.g. strcpy). The C++ problems are things like braced initialization list having different meaning depending on the set of available constructors, or the significantly non-zero cost of various abstractions, caused by strange backward-compatible limitations of the standard/ABI definitions, or the distinctness of vector<bool> etc.
No you are right! Honestly it was several years ago and I struggled to remember exactly what I came up with before I left.
In our application we for example never use dynamic memory allocation. It has to be done very carefully so we don’t crash. Problem is there’s lots of sneaky ways one can accidentally do it from the standard library.
C is almost the perfect subset for me, but then I miss templates (almost exclusively for defining generic data structures) and automatic cleanup. That’s why I’m so interested in Zig with its comptime and defer features.
You may also like Odin if you haven’t already started zig. It’s less of a learning curve and feels more like what c should have always been. It has defer and simple generics, but doesn’t have the magic of comptime.
That’s one thing that always shocks me. You can have two people writing C++ and have them both not understand what the other is writing. C++ has soo many random and contradictory design patterns, that two people can literally use it as if it were 2 separate languages.
The graph goes up for me when I find my comfortable little subset of C++ but goes back down when I encounter other people’s comfortable little subset of C++ or when I find/remember another footgun I didn’t know/forgot about.
When I became a team leader at my last job, my first priority was making a list of parts of the language we must never use because of our high reliability requirement.
Care to share any favourites?
strtok is a worst offender that comes to mind. Global state. Pretty much just waiting to bite you in the ass and it did, multiple times.
Faust bless those who added strtok_s to C11.
Sure, strtok is a terrible misfeature, a relic of ancient times, but it’s plainly the heritage of C, not C++ (just like e.g. strcpy). The C++ problems are things like braced initialization list having different meaning depending on the set of available constructors, or the significantly non-zero cost of various abstractions, caused by strange backward-compatible limitations of the standard/ABI definitions, or the distinctness of
vector<bool>etc.No you are right! Honestly it was several years ago and I struggled to remember exactly what I came up with before I left.
In our application we for example never use dynamic memory allocation. It has to be done very carefully so we don’t crash. Problem is there’s lots of sneaky ways one can accidentally do it from the standard library.
I also have one. I call it “C”
C is almost the perfect subset for me, but then I miss templates (almost exclusively for defining generic data structures) and automatic cleanup. That’s why I’m so interested in Zig with its comptime and defer features.
You may also like Odin if you haven’t already started zig. It’s less of a learning curve and feels more like what c should have always been. It has defer and simple generics, but doesn’t have the magic of
comptime.This comment smells like unix
Damm, C23 has a lot of changes. Some of them are really good, some of them I strongly dislike(keyword auto, addition of nullptr).
That’s one thing that always shocks me. You can have two people writing C++ and have them both not understand what the other is writing. C++ has soo many random and contradictory design patterns, that two people can literally use it as if it were 2 separate languages.