I would love to write modern C++--the language looks remarkably usable; however, the build system and tooling are still overwhelmingly bad (at least for cross platform projects). When something like `cargo` or the `go` tool appear (in the mainstream), give me a call. The build system is the deal breaker for me, not the language.
For an IDE I use the eclipse CDT which has improved I leaps and bounds. The one thing I do is make sure I give eclipse enough memory (a couple gig usually so that its snappy)
I've used CMake + C++ for several years professionally. I know it well enough to know that I don't want to use it if I don't have to. The other aspects of C++ just aren't compelling enough to suffer CMake. In other words, even though I already know C++ fairly well, I would rather abandon it entirely and start fresh with Rust than "invest" more time into CMake. My hobby time is too limited to spend it on frustrating tools.
That is good to know. I have enjoyed writing C++ in the past - I like the clarity and forced obviousness of specifying whether you're talking about a variable value or a pointer reference to it. When switching languages I always have to remember which types and assignment operations duplicate values on assignment vs just copy references.
Maybe I'm just bad at it, but one of the big things that keeps me from writing more C and C++ is getting a dev environment setup, though I did find it a lot easier once I left the Windows Visual Studio ecosystem (ambiguous missing msvc103kd.dll error messages and the like) and drove gcc straight from the command line in a Unix environment.
The main issue comes up with compiler standards compatibility.
In the past with very large C++ projects, we've wanted to move towards newer C++ programming idioms but cross platform compiler compatibility has held us back. More specifically not all of these are on the same page when it comes to standards compatibility to use these features:
- GCC
- MS VCC
- xlC
- Clang
Standards compatibility is one thing, good performance with standards compatibility is another.
So unless you're only targeting a single platform, it'll be a while till everyone has full support for modern C++.
In cases like these, Python, Java, Go or any other VM based language just makes sense unless performance is absolutely paramount.
With data analysis, performance can really help even when it isn't paramount. If I am doing some exploratory analysis and it takes 20 minutes to run instead of a couple seconds, I can iterate much, much faster. This is huge.
I've been wanting to learn C++ (as a data scientist), but never know where to start. There are so many books that were about modern c++ when they were written, it's impossible for an outsider to know what's up to date. Is the book the article mentioned a good one for learning modern C++?
[blog author here] I learned C++ a while ago so I never got to read an intro to modern C++ cover-to-cover. However, a friend asked me to suggest a book and this one seemed to both cover everything you need to understand the language and provide several useful examples for scientific computing (e.g. there's a nice example of gradient descent in the preface).
It's definitely not the only good option. "C++ Primer" is probably the most popular introduction and the language's creator wrote an intro to programming with C++ ("Programming: Principles and Practice Using C++"). They all cover modern C++.
10 comments
[ 4.5 ms ] story [ 43.3 ms ] threadhttps://github.com/Zomojo/Cake
For an IDE I use the eclipse CDT which has improved I leaps and bounds. The one thing I do is make sure I give eclipse enough memory (a couple gig usually so that its snappy)
https://github.com/Akagi201/learning-cmake
But of course I understand your frustration.
Maybe I'm just bad at it, but one of the big things that keeps me from writing more C and C++ is getting a dev environment setup, though I did find it a lot easier once I left the Windows Visual Studio ecosystem (ambiguous missing msvc103kd.dll error messages and the like) and drove gcc straight from the command line in a Unix environment.
The main issue comes up with compiler standards compatibility.
In the past with very large C++ projects, we've wanted to move towards newer C++ programming idioms but cross platform compiler compatibility has held us back. More specifically not all of these are on the same page when it comes to standards compatibility to use these features:
- GCC
- MS VCC
- xlC
- Clang
Standards compatibility is one thing, good performance with standards compatibility is another.
So unless you're only targeting a single platform, it'll be a while till everyone has full support for modern C++.
In cases like these, Python, Java, Go or any other VM based language just makes sense unless performance is absolutely paramount.
It's definitely not the only good option. "C++ Primer" is probably the most popular introduction and the language's creator wrote an intro to programming with C++ ("Programming: Principles and Practice Using C++"). They all cover modern C++.