Filter all clear solution for Magic Square by veky - python coding challenges

Enable Javascript in your browser and then refresh this page, for a much enhanced experience.

Filter all solution in Clear category for Magic Square by veky

def checkio(s):
    n,a=len(s),sum(s,[])
    if all(sum(t)*2==n+n**3 for t in filter(all,s+
        list(zip(*s))+[a[::n+1],a[::n-1][1:-1]])):
        try: i,j=divmod(a.index(0),n)
        except ValueError: return s
        for r in set(range(1,n*n+1)).difference(a):
            s[i][j]=r
            t=checkio(s)
            if t: return t
        s[i][j]=0

July 30, 2013

Comments:

JackSchneider on Oct. 13, 2013, 10:56 p.m.
<pre class='brush: python'> list(zip(*s))+[a[::n+1],a[::n-1][1:-1]])): </pre> tried it and get a generator exit occurred markt after s+ hope you cab tell me how to repair because i want to understand your solution

veky on Oct. 14, 2013, 5:01 a.m.
I have no idea. GeneratorExit shouldn't occur here. Can you tell me the test data you've used, and version of Python you've ran this under? And maybe a complete backtrace?

JackSchneider on Oct. 14, 2013, 9:22 a.m.
I run Python 3.3 in Visual studio Microsoft Visual Studio 2013 Preview Shell (Integrated) Version 12.0.20617.1 PREVIEW Microsoft .NET Framework Version 4.5.50743 Installed Version: IDE Standard Microsoft Python Tools for Visual Studio 2.0.10620.00 Python Tools for Visual Studio provides intellisense, project support, project and item templates, as well as a REPL window for Python development. Microsoft Python Tools for Visual Studio - Django Integration 2.0.10620.00 Provides templates and integration for the Django web framework. Microsoft Python Tools for Visual Studio - Profiling Support 2.0.10620.00 Profiling support for Python projects. SQL Server Data Tools 12.0.30529.0 Microsoft SQL Server Data Tools in order of request. checkio([[1, 15, 14, 4], [12, 0, 0, 9], [8, 0, 0, 5], [13, 3, 2, 16]]) <h1>answer [[1, 15, 14, 4], [12, 6, 7, 9], [8, 10, 11, 5], [13, 3, 2, 16]]</h1> checkio([ [2, 7, 6], [9, 5, 1], [4, 3, 0] ]) <h1>must return [[2, 7, 6], [9, 5, 1], [4, 3, 8]]</h1> checkio([ [0, 0, 0], [0, 5, 0], [0, 0, 0] ]) <h1>can return [[2, 7, 6], [9, 5, 1], [4, 3, 8]] or</h1> <h1>[[4, 9, 2], [3, 5, 7], [8, 1, 6]</h1> Python 3.3 builtins.GeneratorExit occurred Message: <blockquote> <genexpr> in DataModel line 3 Python checkio in DataModel line 3 Python checkio in DataModel line 8 Python checkio in DataModel line 8 Python checkio in DataModel line 8 Python DataModel module line 28 Python </blockquote>

veky on Oct. 14, 2013, 4:57 p.m.
As I said, no idea. You can see at <a href="http://goo.gl/1xfqMd">http://goo.gl/1xfqMd</a> that it produces desired output, and you can step through execution and see what it's doing. The only explanation I have is: "Microsoft. shrug".

nickie on Feb. 26, 2014, 8:07 p.m.
A decryption of this, for all who care. The thing to notice is the "filter(all, ..." and hence the title. <pre class='brush: python'>def checkio(s): # The dimension of the square n = len(s) # A flattened version a = sum(s,[]) # Check if a row, column or diagonal has the right sum def right_sum(t): return sum(t) * 2 == n + n**3 # Checks if all rows, columns and diagonals that do not # contain zeroes have the right sum def good(): rows = s columns = list(zip(*s)) diagonals = [a[::n+1], a[::n-1][1:-1]] full_already = filter(all, rows + columns + diagonals) return all(right_sum(t) for t in full_already) # Now, if it's good so far, solve the rest recursively if good(): try: i, j = divmod(a.index(0), n) # first unknown cell except ValueError: return s # return if none # Try all remaining numbers for the unknown cell remaining = set(range(1, n*n+1)).difference(a) for r in remaining: s[i][j] = r t = checkio(s) if t: return t # Backtrack if no solution was found s[i][j] = 0 </pre>

veky on Feb. 26, 2014, 8:09 p.m.
Now I can put it in Clear with clear consciousness. :-D

Ardillo95 on March 31, 2017, 5:10 a.m.
Hi veky. I am a fan of your solutions and comments but I do not understand why you are not documenting your solutions. I know it is up to everyone and I am not complaining about that, I just would love to see all your solutions with comments as nickie did to your solution of this problem.

veky on March 31, 2017, 10:56 a.m.
Well, I can answer this in many ways, but I don't know which answer you'd prefer. So I'll try them all. :-D First, try to reflect on why are you a fan of my solutions? I'd say much of that reason is because they are fun to understand. Like a magic trick, there is value in getting it yourself instead of having it explained to you. Only, in a magic trick, the value is entertainment, and here, it's education. Second, there is a lot that goes on in my code. And believe it or not, not everyone has trouble with same things. I try to draw attention to some particular feature in the title sometimes (like in this solution), but people's confusion is all over the map (you see it often if you read comments people leave). If I explained everything that goes on up there, the explanation would be at least thrice bigger than the code. Third, I cannot even convince myself I'd know what would the problems be. Many times, I was <a href="https://py.checkio.org/mission/three-words/publications/veky/python-3/for-else/share/cc9483a7db9cabc4f132f0dfcfa37b46/#comment-28027">surprised</a> by the extent of misunderstanding people have about Python. Fourth, it's important to shake people up a little, so they don't get too comfortable, accepting suboptimal design constraints in the process. If through all your life you only see classes with <a href="https://py.checkio.org/mission/haunted-house/publications/veky/python-3/4d/#comment-49535">`init`</a> methods defined, you're missing a lot, and seeing how I write classes here might open your eyes. And I surely ain't gonna start documenting stuff I intentionally didn't write. :-D Fifth, I strongly believe that when you have a nice enough programming language, code is the best documentation - and I'm trying to prove that Python is nice enough for that purpose. Almost every thing you might express in a comment or a docstring, you can express in code - and for a few places where you cannot, Guido is <a href="https://www.python.org/dev/peps/pep-0526/">willing to change his language</a> just so you could. Sixth, I do explain my code in natural language (as much as I can manage). Everytime I've been asked about something in my code someone found confusing, I've answered. It's just that I believe these comment boxes are a much more natural place to do it, than the frozen code editor above them. Here we have interactivity, and it is the best conduit for explanations. And seventh, of course: why would I document my code when other people already do that (as you pointed out for nickie)? :-D

Ardillo95 on April 1, 2017, 2:47 a.m.
I like your solutions because they use the potential Python has, the things that make me love the language. And it is quite interesting how you solve a problem using less than a quarter of the amount of code I use. Hahahahaha now I need to solve Haunted House to see your class based solution. I also do not understand why it is so strange to see class based solutions, I actually like this kind of solutions (but maybe the reason is the same as the reason why it is also strange for me to make a class based solution). Thank you very much for your time.

veky on April 1, 2017, 3:54 a.m.
"Less than a quarter of the amount of code" is mostly because I don't like duplication. And documenting the code in an additional layer above, as I explained, I view as just unnecessary duplication. :-) About classes: yes, you need to solve the mission to see my solution, but here is just the relevant excerpt of conversation: <blockquote> <blockquote> <blockquote> The second innovation is to use the class without an `init` method defined. </blockquote> It’s really sad that that’s considered an innovation. I’m writing a banking software of about 10k lines. Almost none of my classes have `init` defined. Dunder init was meant as a quick&dirty shortcut for Java-nostalgic people who wanted to call their classes like functions (only without writing `new` all the time:) but couldn’t be trusted to get metaclasses (or `new`, which is just a pragmatic hook into meta`.call`) right. Real Pythonic classes that model complex entities should have custom explicit constructors as classmethods. (Those that model simple entities should just use inheritance.:) Besides, in actual classes you need professionally, you often need more than one constructor. And it always makes me sad when people say "because of `init`, Python’s classes can only have one constructor". Same kind of error as wanting “multiline lambdas” - people learn a shortcut, and think it’s the only way - and noone tells them afterwards that it’s just a shortcut, not the real professional way to do things. :-/ </blockquote> Many thanks for such detailed explanation to my guesses! Some of the things you taught here are nothing but subversive in a good way. The point you raised on the `init` method is a fine example of this. The `init` method always bothers me, when I try to construct a class, I need to shift my thinking into an instance of this class and bind everything again into that instance. I tried to hack around it such as updating the `self.dict` with everything in the `**kwargs`, but fully understood it would be hard on the eyes. A `namedtuple` is actually what I need 90% of the time. Also I see that you note that professional developers don’t really use `init` as much for simpler objects. That is REALLY good to know - but all the tutorials and books I read are seriouslly into the `init` stuff, but none teaches the further steps for a developer to grow. </blockquote>

Codeship

0 %

PyCharm The Python IDE for Professional Developers. Enjoy productive Python, web and scientific development with PyCharm. Download it now!