Why Python Programmers Should Learn Python | Hacker News Reader

Why Python Programmers Should Learn Python

wordaligned.org

56 pointsbdfh4213 comments

Unfortunately, the example still contains duplicated information (and effort). I would do:

  words = ["zero", "one", "two", "three", "four",
           "five", "six", "seven", "eight", "nine"]

  numbers = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

  number_to_word = dict(zip(numbers, words))
  word_to_number = dict(zip(words, numbers))

The rest, of course, can remain the same. I think adding this would also help to show how Python idioms like zip can help simplify code.

That said, I still voted the submission up. I've seen several articles like this which point out how people use new languages in the same braindead way they used their old languages, yet I feel this one featured a very simple example with a paradigm that maps well to Python.

EDIT: Whoops, I got distracted and didn't realize that I didn't finish reading the article. It turns out that the approach is modified later on to do something even better, but I'll leave you to check out the article for that.

numbers = range(10) is mucho easier than typing the numbers :)

    words = ["zero", "one", "two", "three", "four",
           "five", "six", "seven", "eight", "nine"]
    number_to_word = dict(zip(range(10), words))
    word_to_number = dict(zip(words, range(10)))

The last approach with lots of testing is good.

Well, here's the approach that was settled on in the article:

  import string
  words = "zero one two three four five six seven eight   nine".split()
  number_to_word = dict(zip(string.digits, words))
  word_to_number = dict(zip(words, range(10)))

I'm not sure why string.digits was used in the first zip but not the second. As kirubakaran pointed out, though, the number_to_word dictionary isn't even needed. The best approach, in my opinion, becomes:

  number_to_word = "zero one two three four five six seven eight nine".split()
  word_to_number = dict(zip(number_to_word, range(10))

This would serve to display several cool Python idioms: the simplicity of string splitting, the power of zip, the simplicity of producing a list of numbers, and the use of the accessor on different types of data structures. If memory recalls, none of those operations are that easy to do in Java.

It can be made a little neater, show off a little more of Python and be more versatile by replacing range(10) with count() from the iterator module [http://docs.python.org/lib/itertools-functions.html]

    from itertools import count
    words = "zero one two three four five six seven eight nine".split()
    word_to_number = dict(zip(words, count()))

[http://codepad.org/naCDzno7]

I like itertools as much as the next programmer (although a large subset of it will be subsumed when Python3k switches to iterators on operations like map and zip), but I feel like using an imported module to (apparently) produce a list of numbers might actually give the wrong idea about Python. In several other languages, you can't even write to standard output without importing a module or five, whereas in Python, you can get a lot done without even touching the standard library (although you will likely want to rather quickly). Seeing an import statement at the top might give the Python newbie the idea that core Python doesn't have "basic" functionality.

I think a good later lesson would be to explain how iterators and the standard library works and then revisit this code to make it look like your example, but in the beginning, I think range(10) would be simpler and would better serve the purpose of the lesson. Even in this case, xrange (at least in Python 2.x) is obviously the better choice, but introducing that complexity at this point doesn't gain much, in my opinion.

Won't range behave like xrange soon anyway?

Or just:

    words = "zero one two three four five six seven eight nine ten".split()
    word_to_number = dict((j,i) for i,j in enumerate(words))

Yes. Also, number_to_word dictionary is not needed as words[0] is already "zero"

Good point. That idea was not mentioned in the article, and it would obviously take up less space and be a faster lookup.

Real programmers can write Fortran in any language.

Schemers who want to borrow the zip idiom can do this:

  (define (zip . lists)
    (apply map list lists))

There's nothing more maddening than seeing someone write Java in Python. I see it all over the place here, with giant utility classes that should just be modules, instance variables instead of locals (yes, that means `self.foo` where `foo` would work exactly the same).

Just a note, Java has split and replace for Strings, too...

Reply on news.ycombinator.com