Well at the risk of stating the obvious, there's a lot of armchair programmer style "read this", "watch this" instruction here. It's much better to do something concrete.
Go and do a trivial project in it, cppreference.com and cplusplus.com have c++11 tags on the new features. I'm sure you've had ideas recently and I'm sure the project is less trivial than you thought!
I think looking at language features first is "a solution looking for a problem" attitude.
People who learn C++ this way tend to miss stuff out, I think - there are a lot of thing that you can just muddle your way though without really understanding.
I'm sure there are some people that think this will be OK; after all, lots of people say you don't really have to understand templates unless you're writing a library (and so on). But I don't think it will be OK. There was very little in C++98 (that I knew well, at the time) that it was OK not to know; unless you knew all of it, pretty much, you couldn't really use any of it safely. (This is how Scott Meyers makes his living.) I can't imagine C++11 will be any different; I might not know it well, but I do note that it has many of the same people behind it.
i also disagree with that statement, while amusing, it's untrue. you can use C++ without the things I have mentioned quite safely. in fact the less features you use, the less they interact and the safer the whole experience is.
obviously we disagree! i think it's ok to miss out on language features.
templates, as you say yourself you're ok not to know. you're also not ok not to know exceptions.
there's other stuff too, you can get away without the C preprocessor. C++ is too big to know all of and it's a waste of time to invest time trying to, unless you are on a standards committee or are writing a compiler...(and even then you can get away with being a specialist!).
it's much better to be a program writer than a language lawyer and dare I say it: more satisfying too.
i draw a huge distinction between people that understand a language and people that can use a language, and i think people underestimate the time and effort in understanding a language. i think that time and effort could be better spent writing programs.
unless you are on a standards committee or are writing a compiler...
Or you work on an existing codebase. Most C++ code that I worked on had templates, exceptions, and used the C preprocessor.
I used to have your attitude. This works for many languages, e.g. if you know some of the syntax of Python or Go and the usual stuff from the standard library, you can get pretty far.
If you want to write modern C++, this is not really true. You need to know about const correctness, lifetimes, some templates, etc. Or you cannot deal with most projects or libraries, like Boost, Eigen, etc.
It's like saying you can program Haskell in 2014 without knowing typeclasses, lenses, GADTs, etc. You can write some educational form of Haskell, sure. But many libraries use these.
Modern C++ is a moving target. By the time you get around to doing a project with one set of features they've been made redundant. Exceptions are a good example of this with noexcept etc coming in and replacing what went before.
I know nothing of haskell. My point is that you should learn the features of the codebase you work on, and it's a waste of time to learn all features, ultimately for a codebase you are not working on or may never work on.
incidentally, ditto, i used to have your attitude.
obviously we disagree! i think it's ok to miss out on language features.
Sure, but there are a ridiculous number of edge cases which result in undefined behavior -- this is something that cannot be avoided when learning C++. You really need to understand what undefined behavior means and what precautions you need to take to avoid it. (Don't get me wrong, recent standards let you write code that's a lot less likely to run into UB, but you need to learn that from somewhere. You should definitely not be approaching modern C++ as an improved C!)
the edge cases are multiplicative, the more features you use, the more they interact. therefore by reducing language features that you use you reduce the number of edge cases you run in to. most shops agree on a subset of C++ before embarking on a project.
i didn't say to approach C++ as an improved C. you are more likely to learn undefined behaviour from compiler warnings than reading a merry array of 500 page books before sitting down to not finish a program.
The thing is that most of the things that your parent mentions are not just some theory. Exceptions, templates, and the C preprocessor are used in most real-life C++ code.
I think starting from somewhere with modern C++ means not starting at C strings, arrays, and bare pointers but starting with STL containers, STL algorithms, and smart pointers etc. to build good habits. But eventually any good introduction has to cover C arrays etc. as well.
Back when I started C++, Accelerated C++ was that sort of introduction. But AFAIK it's not updated for C++ 11 or 14.
I agree that picking a subset is the way to go. Google avoid two out of the three i mentioned that you take issue with avoiding.
Places I have worked the code is divided into libraries and coding standards per library are different to the whole. The point is to lazily learn the part that the codebase you are working on uses and get cracking rather than waste time getting book smart on parts you may never use.
I think the correct approach is a mix. I firmly believe that one of the things that makes good programmers good is developing the correct habits. Once something has been ingrained as a habit you don't even think about it any more and just do it.
I'd take a book like Scott Meyer's new Effective Modern C++ book (which I recommend) and focus on applying one item every week. At the start of the week read through the item. Every time you write code for that week make an effort to see if that item applies to what you're doing. Maybe read the item again midweek or towards the end of the week to see if the practical experience changes your view on what he wrote. The book contains 42 items so if you go through it like this you'd be done in less than a year.
I'm not saying that people shouldn't read the whole book at once. However, if you just read it through once and put it on the shelf I don't think you'll get the same benefit that you would get from spending time focusing on individual areas.
I think you make a very valid point and I would agree to an extent. However I wouldn't discount the value in seeking to understand the personality and capabilities of a language as it evolves over time.
I think of any language as another tool in my toolbox. The better I know a tool, the better I know when and how to apply it to a specific problem.
I'd wager that just one trivial project won't fully demonstrate the personality and capabilities of a language.
absolutely, but I think acquiring knowledge by reading a 500 page book dutifully is a time sink. much better to get a touristic overview and solve actual problems rather than trying to get book smart.
Comments
Well at the risk of stating the obvious, there's a lot of armchair programmer style "read this", "watch this" instruction here. It's much better to do something concrete.
Go and do a trivial project in it, cppreference.com and cplusplus.com have c++11 tags on the new features. I'm sure you've had ideas recently and I'm sure the project is less trivial than you thought!
I think looking at language features first is "a solution looking for a problem" attitude.
People who learn C++ this way tend to miss stuff out, I think - there are a lot of thing that you can just muddle your way though without really understanding.
I'm sure there are some people that think this will be OK; after all, lots of people say you don't really have to understand templates unless you're writing a library (and so on). But I don't think it will be OK. There was very little in C++98 (that I knew well, at the time) that it was OK not to know; unless you knew all of it, pretty much, you couldn't really use any of it safely. (This is how Scott Meyers makes his living.) I can't imagine C++11 will be any different; I might not know it well, but I do note that it has many of the same people behind it.
unless you knew all of it, pretty much, you couldn't really use any of it safely.
Thank you for perhaps the best summation of everything wrong with using C++ in a project, and everything wrong with C++ as a language!
i also disagree with that statement, while amusing, it's untrue. you can use C++ without the things I have mentioned quite safely. in fact the less features you use, the less they interact and the safer the whole experience is.
You know what we call (C++)--?
C.
:)
you should actually preincrement there to avoid the copy constructor, the correct formulations is (++C)--
How do you know that my preincrement and postincrement are the same? ;)
Well played, at any rate.
obviously we disagree! i think it's ok to miss out on language features.
templates, as you say yourself you're ok not to know. you're also not ok not to know exceptions.
there's other stuff too, you can get away without the C preprocessor. C++ is too big to know all of and it's a waste of time to invest time trying to, unless you are on a standards committee or are writing a compiler...(and even then you can get away with being a specialist!).
it's much better to be a program writer than a language lawyer and dare I say it: more satisfying too.
i draw a huge distinction between people that understand a language and people that can use a language, and i think people underestimate the time and effort in understanding a language. i think that time and effort could be better spent writing programs.
unless you are on a standards committee or are writing a compiler...
Or you work on an existing codebase. Most C++ code that I worked on had templates, exceptions, and used the C preprocessor.
I used to have your attitude. This works for many languages, e.g. if you know some of the syntax of Python or Go and the usual stuff from the standard library, you can get pretty far.
If you want to write modern C++, this is not really true. You need to know about const correctness, lifetimes, some templates, etc. Or you cannot deal with most projects or libraries, like Boost, Eigen, etc.
It's like saying you can program Haskell in 2014 without knowing typeclasses, lenses, GADTs, etc. You can write some educational form of Haskell, sure. But many libraries use these.
Modern C++ is a moving target. By the time you get around to doing a project with one set of features they've been made redundant. Exceptions are a good example of this with noexcept etc coming in and replacing what went before.
I know nothing of haskell. My point is that you should learn the features of the codebase you work on, and it's a waste of time to learn all features, ultimately for a codebase you are not working on or may never work on.
incidentally, ditto, i used to have your attitude.
Sure, but there are a ridiculous number of edge cases which result in undefined behavior -- this is something that cannot be avoided when learning C++. You really need to understand what undefined behavior means and what precautions you need to take to avoid it. (Don't get me wrong, recent standards let you write code that's a lot less likely to run into UB, but you need to learn that from somewhere. You should definitely not be approaching modern C++ as an improved C!)
the edge cases are multiplicative, the more features you use, the more they interact. therefore by reducing language features that you use you reduce the number of edge cases you run in to. most shops agree on a subset of C++ before embarking on a project.
i didn't say to approach C++ as an improved C. you are more likely to learn undefined behaviour from compiler warnings than reading a merry array of 500 page books before sitting down to not finish a program.
I agree; sometimes you have to start somewhere. Then come back and improve knowledge. Trying to be a theoretical expert early on will bite you
The thing is that most of the things that your parent mentions are not just some theory. Exceptions, templates, and the C preprocessor are used in most real-life C++ code.
I think starting from somewhere with modern C++ means not starting at C strings, arrays, and bare pointers but starting with STL containers, STL algorithms, and smart pointers etc. to build good habits. But eventually any good introduction has to cover C arrays etc. as well.
Back when I started C++, Accelerated C++ was that sort of introduction. But AFAIK it's not updated for C++ 11 or 14.
I agree that picking a subset is the way to go. Google avoid two out of the three i mentioned that you take issue with avoiding.
Places I have worked the code is divided into libraries and coding standards per library are different to the whole. The point is to lazily learn the part that the codebase you are working on uses and get cracking rather than waste time getting book smart on parts you may never use.
I think the correct approach is a mix. I firmly believe that one of the things that makes good programmers good is developing the correct habits. Once something has been ingrained as a habit you don't even think about it any more and just do it.
I'd take a book like Scott Meyer's new Effective Modern C++ book (which I recommend) and focus on applying one item every week. At the start of the week read through the item. Every time you write code for that week make an effort to see if that item applies to what you're doing. Maybe read the item again midweek or towards the end of the week to see if the practical experience changes your view on what he wrote. The book contains 42 items so if you go through it like this you'd be done in less than a year.
I'm not saying that people shouldn't read the whole book at once. However, if you just read it through once and put it on the shelf I don't think you'll get the same benefit that you would get from spending time focusing on individual areas.
I think you make a very valid point and I would agree to an extent. However I wouldn't discount the value in seeking to understand the personality and capabilities of a language as it evolves over time.
I think of any language as another tool in my toolbox. The better I know a tool, the better I know when and how to apply it to a specific problem.
I'd wager that just one trivial project won't fully demonstrate the personality and capabilities of a language.
absolutely, but I think acquiring knowledge by reading a 500 page book dutifully is a time sink. much better to get a touristic overview and solve actual problems rather than trying to get book smart.