I work on a large multi-platform desktop shrink-wrap application.
It is a large application, with a code base of over 400,000
files. The majority of which is C++, although there is some
JavaScript in the mix in some of the corners of the UI. The code
base is about 30 years old... so you probably can imagine that
the code has room for improvement.
For an existing project of that size, it makes no sense to try to
throw it out and start over. That would be recklessly
ill-advised.
I was fantasizing: if the app were to be written today, what
would be suitable languages?
C++17, is a shoe in, since that corresponds to reality (if looked
at through a carnival mirror, since the actual code is smeared
across C++89 - C++98/03 - C++11/14 - C++17 like cities built on
top the ruins of cities, and its turtles all the way down).
Since I'm a D fanboy, I put D on the list. And I think these as
well: C#, F#, Go*, C, Lua (with some core functionality in C),
Java, Groovy*, Scala*, Swift. And maybe, at a stretch, Ada*,
Clojure*, Kotlin*.
* I've no practical experience in these languages.
A friend of mine said I should add Rust* to the list. I'm not
familiar with Rust, so today is my "dive into Rust" day. That's
how I got here. (Hi Walter! Hi Bearophile! I've been away for
a long while.)
In my fantasy, the biggest challenge with a large multi-platform
application is the tooling and support.
So some of the questions that are meta-language related are...
What are the IDEs like? People want to have bother-free version
control, enjoyable debugging experience, works on the target
platforms, manage very large projects. C++ has excellent support
by Visual Studio, and very good support by Xcode.
What do the standard libraries offer? C++ has a reasonably nice
standard library.
What off-the-shelf libraries can be leveraged? C++ has Boost,
and many others. A vibrant ecosystem.
Can the language "do the job"? Part of "doing the job" is
interoperating with the target platform. In this fantasy case,
Win32 with GDI/GDI+ (possibly .NET as a viable alternative), and
Cocoa. If that involves writing a C-API bridging layer from
Language X to Host Platform, that is friction. A fantasy project
this size can absorb that kind of friction.
Can we hire 100+ programmers willing to use that language?
Those things are important, even though they are outside the
domain of the language itself. Toolchains. IDEs. Tool vendor
stability. Debugging. Large project management. Version
control. Suitability. Community & ecosystem. Human resources.
Since my love-hate relationship with C++ is well known, in my
do-over fantasy world, I hope there is a better suited language.
And "make your own language" is not on the table. I've read
Aho's dragon book... I've tried that... it is VERY HARD to make
your own general purpose language, and will consume 10+ years of
your life. Bjarne Stroustrup said, "If you do anything useful it
will haunt you forever after, and if you have a major success you
get decades of hard manual labor - meaning you have to work on
the manual." *hugs* to Walter & Andrei.
A cautious person would probably say "Use Swift on Mac, use C# or
F# on .NET on Windows. Write the code twice. Boom, done."
That's been tried, turns out maintaining parity that way does not
work out so well. Case in point: check out Office for Windows,
and Office for Mac.
Step back...
Why am I sharing this on this forum?
I think these meta-language issues need to be thought about to
help D gain more traction. In addition to the core language
issues that Andrei has already discussed in depth, and were
mentioned again in this thread.
And I really love D. It speaks to me. It is the language I wish
C++ was. If I could have created a language, I would compare it
against D as my gold standard.
I look at the history of RemObjects, and how they had a
distributed object technology... but to support it they had to
become a language vendor (of Oxygene, their version of Delphi,
which is Apple/Wirth's Object Pascal), and then had to create
their Elements compiler, and extend it to support multiple target
platforms, and then extend it more to support multiple languages
(Object Pascal, C#, Swift, Java) in their ecosystem. And now
they've jumped into the IDE game. It is interesting to note that
where they started and how they got to where they are is an
interesting journey of necessity.
