Well, good on you for having learned another language first.

As for switching to C++... well, its going to take time. A lot of it. You will find the tools for C++, at least in relation to refactoring and such, to be a lot more primitive than the Java/C# versions. You will also find the language is a lot less friendly, and the standard library is quite a bit slimmer. This means you will spend a lot more time either investigating libraries to augment functionality, or implementing boilerplates.

As for resources: C++ Primer is a good start, especially if you get the latest edition which covers C++11 features. Do be warned though that many C++11 features are not yet supported by all compilers. The C++ Programming Language is also a good one to get. Then there are other books that are good reads, such as Effective C++.

Absolutely finish your game.

In principle, I agree, but since OP alludes to it taking a ton of time it *might* be worthwhile to port the game to C++ and complete it there, especially if OP feels that the lack of gaming support in Java is a hindrance.

I would say if you are less than 40% of the way through, or if you're more than 9 months from completion, then consider re-writing what you have in C++ and finish the game. Otherwise, stick it out with Java. Josh is absolutely right that a completed game in any language is way more valuable than numerous half-finished ones. A bird in hand is worth two in the bush, and all that.

However, be aware that you will certainly find C++ to be more slow-going. C++ has a great many points in its favor, but productivity is never counted among them -- even moreso for a beginner. You will not have good tools for rapid refactoring, error messages are often arcane, and your programs will leak memory gleefully if you fail to apply proper management patterns.

Modern, idiomatic use of C++11 better arms you against those sorts of problems than C++ ever has in the past, but its road is simply not as well-paved as languages like Java or C#. Since you have the benefit of starting now, in the age of C++11, I would recommend you against using dated resources for learning. Many of the core books -- Lippman's C++ Primer, Stroustrup's The C++ Programming Language, and Josutti's The C++ Standard Library -- are up-to-date with C++ 11 in their latest editions, try to learn from those. Stroustrup's A Tour of C++ is a great introductory text as well, short, but inexpensive. Meyer's is writing an updated Effective C++ as well, but its not ready yet, that one looks like it will cover both C++11 and C++14, which is being voted as a final standard soon (I wouldn't wait for texts to be updated for C++14, the changes between it an C++11 are less radical/fundamental than between C++11 and what came before it.)

Absolutely finish your game.

In principle, I agree, but since OP alludes to it taking a ton of time it *might* be worthwhile to port the game to C++ and complete it there, especially if OP feels that the lack of gaming support in Java is a hindrance.

I would say if you are less than 40% of the way through, or if you're more than 9 months from completion, then consider re-writing what you have in C++ and finish the game. Otherwise, stick it out with Java. Josh is absolutely right that a completed game in any language is way more valuable than numerous half-finished ones. A bird in hand is worth two in the bush, and all that.

However, be aware that you will certainly find C++ to be more slow-going. C++ has a great many points in its favor, but productivity is never counted among them -- even moreso for a beginner. You will not have good tools for rapid refactoring, error messages are often arcane, and your programs will leak memory gleefully if you fail to apply proper management patterns.

Modern, idiomatic use of C++11 better arms you against those sorts of problems than C++ ever has in the past, but its road is simply not as well-paved as languages like Java or C#. Since you have the benefit of starting now, in the age of C++11, I would recommend you against using dated resources for learning. Many of the core books -- Lippman's C++ Primer, Stroustrup's The C++ Programming Language, and Josutti's The C++ Standard Library -- are up-to-date with C++ 11 in their latest editions, try to learn from those. Stroustrup's A Tour of C++ is a great introductory text as well, short, but inexpensive. Meyer's is writing an updated Effective C++ as well, but its not ready yet, that one looks like it will cover both C++11 and C++14, which is being voted as a final standard soon (I wouldn't wait for texts to be updated for C++14, the changes between it an C++11 are less radical/fundamental than between C++11 and what came before it.)

Yes, getting the C++11 versions of the books (which I linked) is definitely important. It is important to note though that most compilers do not fully implement the C++11 standard yet, usually its either parts of the standard library missing (such as regex) or its language things like constexpr.

Effective C++ is a great book regardless of C++11 or not, but since he'll be spending his time learning, its not a necessary book to have in the near future... So... he can wait to pick that up until the 2014 edition is out.

As far as the project goes... honestly, its nice seeing projects in multiple languages. It shows a level of flexibility that many junior programmers do not have. So if this project is too far in the future, it would be good to do up another project that simpler in Java.

Yes, getting the C++11 versions of the books (which I linked) is definitely important. It is important to note though that most compilers do not fully implement the C++11 standard yet, usually its either parts of the standard library missing (such as regex) or its language things like constexpr.

Many of the compilers are more than good enough to basically ignore outmoded techniques and idioms though.

Notably, Microsoft's compiler is furthest behind -- they plan on delivering a sort of mishmash of C++11 and C++ 14 features going forward, until around the time that everyone ought to be delivering C++14, when they'll achieve rough parity. If you've read any of the blogs, they've stated several times that the simple fact is that their compiler internals had been quite dated, and they're having to refactor great swaths of it concurrently to implementing new language features -- I mean, the compiler didn't even have an intermediate representation before.

Still, even today on any mainstream compiler including Microsoft's, you get things like auto, trailing return types, lambdas, variadic templates, parameter packs, shared_ptr/weak_ptr/unique_ptr, make_shared, chrono and random, range-based for-loops, rvalue references and move semantics. initializer lists, decltype, defaulted and deleted member functions, default template arguments for template functions, delegating constructors, atomics. The most useful stuff is widely available (with the notable exception of constexpr), the stuff that really impacts the way you architect and write your code.