Hacker News Reader

Comment by PythonicAlpha | Hacker News Reader

Yes, Python is slower in execution than some other languages, but:

   ... efficient use of development time ...

That is the reason, why Python counts (not only references). Python has many very good libraries, is a good OOP language, easy to learn, but still very, very powerful. You can express in Python some things in a single line, where you need hundreds of lines in C++ or other languages.

The few percent of running speed that you might loose, are neglect-able in most cases against the win in development speed.

In many applications you don't need the full CPU power, but often times you are hindered by e.g. the disk speed or other factors ... and than you don't lose anything when you are a little bit slower in some minor tasks.

View on HN

> You can express in Python some things in a single line, where you need hundreds of lines in C++ or other languages.

citation needed.

I have none, it is my personal experience, since I programmed in C++ (and other languages) before. So cite me, if necessary.

Example: x = {'foo':34,'bar':[1,2,3],'baz':'quux'}

The headers alone for STL maps are hundreds of LOC.

I agree with the upstream comment about expressiveness, but this isn't a fair example.

Counting headers is rather unfair.

C++11:

map<int, char> x = {{1, 'a'}, {3, 'b'}, {5, 'c'}, {7, 'd'}};

Any not-ancient C++ using boost Assign:

map<int, char> x = map_list_of (1, 'a') (3, 'b') (5, 'c') (7, 'd');

Shamelessly stolen from here:

http://stackoverflow.com/questions/138600/initializing-a-sta...

That's tricky, because C++ is statically typed and you've selected a varying value type, but here's the closest thing in real terms.

    #include <map>
    #include <vector>
    #include <string>
    #include <boost/any.hpp>
    using boost::any;
    using std::map;
    using std::string;
    using std::vector;

    int main() {
        map<string,any> x = {{"foo", 34}, {"bar", vector<int>{1,2,3}}, {"bar", "quux"}};
    }

Syntax overhead amortizes to constant factor ;)

And if you really need to have a piece of your code execute really quickly, you can just make C extension that python can easily invoke.

Python's support for this is awesome, but there is a gap: you lost cross platform. You now have to figure out how to make it compile on every platform that anybody could every possibly want to run it on, including ones not yet invented in the future ... Jython / Java (or other JVM languages) are quite a nice option in that space.

Of course you loose a little platform independence, but that is something you don't have anyway when you program in C++ or that kind. Also you get rather rich support for makefiles that are cross platform. I had no problems till now with C code I wrote that integrates into the Python environment. When you don't need OS specifics and have knowledge how to implement things CPU independent (given a common >=32bit architecture) in C, you will have little to no trouble at all.

Also one little hint: Use Cython for enhancements. Cython is great for implementing small enhancements when you don't want to hassle with refcounts and the normal Python API. Also your enhancements are automatically compatible to Python 2 + Python 3. You can also call C functions from Cython very easily for full speed. Cython is really great, when you don't need to optimize to the last nanosecond.

Reply on news.ycombinator.com