Show HN: Python to C++14 transpiler

Show HN: Python to C++14 transpiler(github.com)

96 points by morgenkaffee 10 years ago | 44 comments

pjmlp 10 years ago |

It is called a compiler even it outputs code in another language, transpiler is some neologism from JavaScript developers without a background in compiler design.

chrisseaton 10 years ago | |

I'd defend the word 'transpiler' and I did my masters and PhD on language implementation, and that's where I work professionally now, so I'm not ignorant of compilers.

I think it implies a translation from one high level language to another (not that 'high level' is well defined either), with only desugaring and maybe type-checking - no real lowering or optimisations. That's a useful subset of compilers, so can have its own word I believe.

pjmlp 10 years ago | | |

So for you a C compiler that follows the initial workflow is a transpiler?

Or an Eiffel implementation, as another example.

poizan42 10 years ago | |

Why are people giving this word so much flak? Yes it's a neologism, but it's a useful one. A decompiler is a compiler to, it just goes from a low level language to a high level one. Do you suggest that we stop using the word decompiler as well?

kitd 10 years ago | |

'Transpiler' goes back long before the Javascript/CoffeScript world. Pascal/C and FORTRAN/C have a history of 'transpilation' products.

It's just short-hand for source-to-source compiler and IMHO has been used enough to warrant its inclusion in a programmer's dictionary.

pjmlp 10 years ago | | |

Yes they have, and we called them compilers.

Can you dig me a 70's paper with the transpiler word on it?

srean 10 years ago | |

> It is called a compiler even it outputs code in another language

Indeed and most compilers do exactly that. Cant help getting a little annoyed whenever I hear this 'transpiler' word. I guess the ship has sailed, oh well !

nimitkalra 10 years ago | |

From what I've noticed, these "transpilers" output code that is readable (the code itself is written as though a human wrote it) where as "compilers" output code that has been optimized and show effects of name mangling in the code itself, etc. Just an observation.

I think it makes sense to use a different term for this "compiler"-esque behavior. For example, I might edit the output of CoffeeScript generated Javascript whereas I wouldn't know how to modify the output of gcc.

pjmlp 10 years ago | | |

The output of gcc can be straight readable Assembly, once upon a time a normal language to write business applications in.

Also C compilers used to generate Assembly text files, which were then piped into the Assembler.

So no, transpiler doesn't make any sense.

omaranto 10 years ago | |

People believe some languages are "higher level" than others and call a tool that translates programs written in A to programs written in B a compiler if A is clearly higher level than B and a transpiler if B is of a level higher than or roughly equal to A.

You can certainly think this distinction doesn't merit using a different word, but you shouldn't think that people who use the word "transpile" use it as a synonym for "compiler".

lmm 10 years ago | |

The distinction is meaningful and valuable. Language evolves. No need for the attacks.

morgenkaffee 10 years ago | |

You are right. C++ is quite different to Python and so it should be called a compiler.

But it fits the project since I don't have any background or experience in compiler design.

tormeh 10 years ago | |

Transpilers are a subcategory of compilers. It's just a more specific word.

jmgao 10 years ago | |

Alternatively, translator, the word that CFront used.

Sharlin 10 years ago | | |

And "translation" is still the term of art for the compilation process at least in the C and C++ standards.

drvortex 10 years ago | |

Strictly you are right. However, by the strict definition of a compiler, an interpreter is nothing but a sort-of "real time compiler".

But a compiler has come to mean a program that translates source-code into machine code.

By this new understanding, a transpiler is a combination of a compiler and a decompiler with different input and output languages.

So po-tay-toes, po-taa-toes. No one cares, we know what is meant by transpiler.

programmer_man 10 years ago | |

The word transpiler was used before we had web browsers, much less javascript.

Point taken that it is still a compiler.

agumonkey 10 years ago | |

or pretty printing

throwaway999888 10 years ago | |

I for one am just waiting for the cispiler.

Symmetry 10 years ago | | |

An optimizing cispiler sounds really interesting. Something that applies all the optimizations that the compiler would do when translating your C to machine code but showing you the results in C. Of course not all optimizations are representable in C (calling conventions for instance) but many are.

haberman 10 years ago |

Key sentence: "The goal is to showcase the power of C++14 templates and not to create a fully functional transpiler."

Viewed through that lens, this is a really novel and cool demonstration.

Joky 10 years ago | |

This seems to me to be very close to what Pythran [0] is doing (since 2011), except that Pythran includes some type inference and bridge with Python code.

So view through that lens I'm not really seeing the novelty right now?

[0]: http://github.com/serge-sans-paille/pythran/

anon4 10 years ago | | |

There is also nuitka - http://nuitka.net/

Sharlin 10 years ago |

Should be noted that this only works on a "statically typeable" subset of Python where every variable has a de facto static type inferred at the first assignment. For instance, the following valid Python code would output invalid C++:

  var = []
  var = 2

morgenkaffee 10 years ago | |

Yes you would have to program in a subset of python. But type annotations are not always needed.

For the array I do a little hackery. You can define the array without an initial value in the container and I can guess the value type.

  arr = []
  arr.append(1)

it will spit out

  std::vector<decltype(1)> arr{};
  arr.push_back(1);

exprx 10 years ago | |

Well, you could always begin another scope with each assignment.

    auto var = // this type can't be inferred because it's not used
    {
      auto var = 2;
    }

I'm too exhausted to think why this may not be applicable.

marvy 10 years ago | | |

Python:

    x = 2
    if stringy: x = '2'
    print x+x # 4 or 22???

C++:

    int x = 2;
    if(stringy) string x = "2";
    print x+x; // 4 (string x is out of scope)

Fede_V 10 years ago |

There's a few similar process. Shedskin, Nuitka, Pythran, etc. They are all pretty cool projects, and worth looking into just to learn new techniques.

srean 10 years ago | |

... and unpython, spyke .... all very cool and with their own spin. What is a little worrisome is the rate at which these get abandoned and how many of them target some specific subset of Python. I particularly liked unpython's take.

seivan 10 years ago |

Would someone be kind to explain how the transpiler/compiler knows that num of T1 generic(?) type can be used with the <= operator? Or that something the user themselves have to define?

Wouldn't be something like "T1 where T1 is Numeric"?

Thanks!

inglor 10 years ago | |

This is how C++ templates work. They are a form of polymorphism through how you use the object and not based on its type.

If you pass a type that cannot be used with the <= operator it will error in compile time.

Being able to do this is part of why C++ templates are much more powerful than Java/C# generics and why they enable a different (and alternative) form of polymorphism to inheritance and explicit interfaces.

seivan 10 years ago | | |

I see. In e.g Swift, you'd have to define before hand that T1 could be used with <= while here I guess it adds in the method for "each" T1 that could be used with it.

I assume this just duplicates the method for each type at compile time instead of at runtime try to figure it out?

Damnit, now I wanna rewrite back to C++.

morgenkaffee 10 years ago | |

It is all magic of Clang or GCC. Once you call the template with T1 the C++ compiler will enforce that your passed type supports the <= operator.

techdragon 10 years ago |

Writing something like this for Python 2 is like throwing a urine filled water balloons at all the progressive developers working hard to get the Python community transitioned to Python 3.

Don't have enough reasons to stick with shitty old Python 2, well then here's another anchor for your boat!

Edit: The first pull request was for Python 3 support, hooray.

anc84 10 years ago | |

No need to phrase it so nastily.

marktangotango 10 years ago | | |

You say nasty, I say colorful. No need to characterize it so prejudicially.