Enum clear solution for Roman Numerals by veky - python coding challenges

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

Enum solution in Clear category for Roman Numerals by veky

from enum import Enum

class Roman(Enum):
    M  = 1000
    CM = 900
    D  = 500
    CD = 400
    C  = 100
    XC = 90
    L  = 50
    XL = 40
    X  = 10
    IX = 9
    V  = 5
    IV = 4
    I  = 1

    @classmethod
    def encode(cls, n):
        for numeral in cls:
            rep, n = divmod(n, numeral.value)
            yield numeral.name * rep

checkio = lambda n: ''.join(Roman.encode(n))

July 12, 2016

Comments:

clamytoe on Sept. 16, 2016, 4:21 p.m.
I haven't a clue how this works but it does! Rest assured that I'm going to be digging into enum! Thank you for introducing me to it.

veky on Sept. 16, 2016, 8:33 p.m.
You say something about happiness about having to create the dict... this is the same, only with a nicer syntax. :-D

GiForce on Sept. 19, 2016, 10:19 p.m.
Wow, that looks clear. Beautiful! I'm so sorry to not understand the whole operation yet.

veky on Sept. 20, 2016, 9:48 a.m.
What is the problem? enum, classmethod, or (godforbid) lambda? :-)

anddam on Oct. 15, 2016, 8:02 a.m.
Love it, my favorite approach so far, in fact I'm changing my iterative approach to one with a generator now. I wouldn't use a class but rather a dict in order to have .items() and explicit "name" and "value" identifiers in the generator, but I have to admit assignment of class attributes is way more readable than items in a dict or a list of tuples.

veky on Oct. 15, 2016, 8:21 a.m.
Well, it's a dict with an attached specialized method. Seems like a class to me. :-)

DahliaSR on Oct. 15, 2016, 7:25 p.m.
Maybe a bit too much for this task but a great example on what is possible. This also is the best approach if you ever want to decode a roman numeral to decimal rep. Just add decode method and your're done. Great!

veky on Oct. 16, 2016, 2:19 a.m.
Of course. As I said many times here, the primary goal isn't to solve the tasks, but to learn Python. But I have to work with the missions that are available. (Yes, I know I can make my own, but that's a whole different part of this game I don't want to go into yet.:) Homework: do write a decode method, and investigate whether you can embed this into Python's <a href="https://docs.python.org/3/library/codecs.html">codecs</a> framework. :-)

DahliaSR on Oct. 16, 2016, 5:04 p.m.
Autumn Holidays just started... so gues what my right hand wright does now :D

mike-weinberg on Dec. 26, 2016, 1:22 a.m.
it seems weird that half of the encoding logic is in the class method (as a generator) and the other half outside. It would have been nice to either add a second method or use nested functions for the encode method, because as-is encode doesn't provide a useful output.

veky on Dec. 26, 2016, 5:43 a.m.
On the contrary, I think the separation of functionality and presentation is critical. Python3 has made that great insight that streams of data (iterators) are fundamental building blocks, and our freezing of them should occur only at the end, for visualization of the result. (: At least that's what theory says. Now let's see the practicality beat purity. :) <hr /> Of course, here it is a bit stretched, since there is not much you can do with the generated values, except `''.join`ing them. In such a case, it's not hard to write a decorator `aggregate` [I have one in my personal library; I think it's a nice homework:] so you can do something like <pre class='brush: python'>@aggregate(''.join) def output(widget): yield '(' if widget.child: yield str(widget.child) yield ')' </pre> [Making it work seamlessly in this case, stacked with `@classmethod`, might be a bit harder, but not impossible.:] Of course, once you make that generalization, you immediately see that inside `@aggregate`, you can write `list`, `frozenset`, `bytearray`,... , or even things like `max`, `sorted` and `any`. And each of them is incredibly useful. `@aggregate` has slim chances of making it to the stdlib, since it fundamentally uses `compose` (functional composition) and `curry` (explicit currying), both of which Guido hates. But for personal uses, it is indispensable. <hr /> (Back to purity-land.) Still, you shouldn't preclude your users' plans. Even if it seems to you that there is no other useful form for results than, say, a `set`, maybe someone would actually want `collections.Counter` there, as a multiset. This is one of rare examples where there is only one possible aggregator - at least I can't see any other - so it might be good to `@aggregate`. I find myself using my own `@aggregate` a lot when writing quick scripts, but when writing libraries, I usually have a vision wide enough to let users decide what form of output they want.

mike-weinberg on Dec. 26, 2016, 6:10 a.m.
I can see why that makes sense, as all sorts of operations would be easier with it in its expanded format. I still think that, in a world where we had to build out a Roman class, having a method for printing the actual Roman numeral makes far more sense than having an unaffiliated global function, and having both methods be exposed is most useful. Ultimately I can't imagine that a roman numeral class is going to be convenient without a decent print method.

veky on Dec. 26, 2016, 10:19 a.m.
It's interesting that you mention printing, since printing the result instead of returning it is a often-made beginner mistake (just see in CiO forum:) stemming from the same kind of "I know better than my caller how the result will be used" thinking. Of course, in that case it's much easier to convince people that returning is more flexible. I'm quite sure you don't think <pre class='brush: python'>class Roman: def print(self): ... Roman(23).print() </pre> would be a good solution, right? :-)

veky on April 21, 2017, 2:05 p.m.
The above was written in jest, but there is a useful insight here: once you understand that classes (even such as the one above) can naturally have instances whose behaviour they define, it's easy to see that what's actually needed is <pre class='brush: python'>class Roman: def __str__(self): return ''.join(self.encode(self.value)) </pre> And both schools of thought are happy. :-)

imtiaz.rahi on Aug. 15, 2019, 10:22 a.m.
I added below code to your original class <pre class='brush: python'> def __str__(self): return ''.join(self.encode(self.value)) if __name__ == "__main__": print(Roman.encode(23)) </pre> Output: <pre class='brush: python'><generator object Roman.encode at 0x7f345a246ba0> </pre> Expected the roman value. What did i miss ?

veky on Aug. 15, 2019, 10:26 a.m.
I'm not quite sure how did you expected the Roman value. The `encode` method doesn't return a `Roman` instance, it returns a generator. I didn't really mean above that you could just add the above lines to the class, some refactoring is needed. But surely you don't "encode" from outer program anymore, the print (via str) does it for you. For example: <pre class='brush: python'>class checkio(NamedTuple): value: int def __str__(self): return ''.join(Roman.encode(self.value)) print(checkio(23)) </pre>

danny.luker_6f2b7a7a6c6340b297 on Jan. 10, 2017, 7:24 p.m.
Beautiful! It took me a while to figure out what was happening here but your solution is very instructive. Thanks!

veky on Jan. 11, 2017, 12:41 p.m.
It's similar to many OrderedDict solutions, only with a nicer syntax. :-D

Ballistik on April 21, 2017, 9:38 a.m.
Very interesting decision)) But why do you use divmod? As i remember, rep in rep, n = divmod(n, numeral.value) will be equal to rep = n // numeral.value . It's not critique, it's only interest)

veky on April 21, 2017, 10:07 a.m.
Yes, and `n` will be as `n %= numeral.value`. I need both. It's a perfect use case for divmod: I need div and mod of same numbers at the same time.

Ballistik on April 21, 2017, 2:24 p.m.
huh, understand) thank you)

txyloveCoding on Sept. 3, 2017, 2:53 a.m.
I think that a return can also do the job of yied ,like <pre class='brush: python'>@classmethod def encode(cls, n): s = [] for i in cls : rep, n = divmod(n, i.value) s.append(rep * i.name) return s </pre> So why do you think that we should use yield instead of return ? just because the code has less lines in that case ?

veky on Sept. 3, 2017, 8:02 a.m.
Only if you * want a list * can afford a list * have only finitely many items * are required to return a list In this case, the last one doesn't hold, since you are in fact required to return a str. So returning a list from one function just to ''.join it in the other is obviously a roundabout way of doing it. It's precisely why Python3 has switched from lists to iterators in almost all its high API: range, reversed, map, filter,... In most cases, you don't want a list as the end result anyway. The true way is to <a href="https://py.checkio.org/mission/roman-numerals/publications/veky/python-3/enum/share/22e044ab85d2e00f7c2bb4b0c7550f41/#comment-49266">separate computation from presentation</a>. BTW it's funny how many times people assume golfing as my main motive. It never was. It's just that Python is so (well) designed that in most cases, the right way to do it is also the shortest. :-)

StefanPochmann on Sept. 3, 2017, 11:08 p.m.
You are getting a list of the pieces created anyway, though, are you not?

veky on Sept. 4, 2017, 4:38 a.m.
Hmm... not from the Python perspective. Yes, the current CPython implementation (after many alternative implementations) is to convert the generator to a list (PySequence_Fast), but that list is not accessible from Python. And it's quite imaginable that e.g. Jython would use a StringBuilder for that purpose. Anyway, even if I did write `''.join(list(` outside, it still wouldn't invalidate my point that it is the user of the function who should say in which form the results are needed, not the implementer of that function.

txyloveCoding on Sept. 4, 2017, 11:42 a.m.
Thank you, that is quite clear!

hao.daniel on May 24, 2019, 3:08 a.m.
Wow! It just blow me away. (in good sense) So much to learn... but life is so short...

veky on May 25, 2019, 6:43 a.m.
The solution is simple: learn only what interests you. :-D

allburov on Nov. 22, 2019, 10:29 a.m.
It's an awesome solution! I forget about `yield`, we don't have to use `result = [] ... result.append[symbol] ... return result`, just `yield symbol`

veky on Nov. 22, 2019, 11:48 a.m.
Of course, it depends on what exactly do you do with the returned value. But even if you want a list, you can always `list` it at the receiving side.

lixuchu on Nov. 29, 2019, 2:22 a.m.
I spent some time to do some homework. Now I can understand your code. It's amazing. Your code always show me something new. Thank you!

vlad.bezden on Feb. 14, 2020, 3:05 p.m.
Very smart to derive from Enum. I love it, just beautiful! Just wondering if you could achieve the same effect using dictionary?

veky on Feb. 14, 2020, 3:11 p.m.
Absolutely. See <a href="https://py.checkio.org/mission/roman-numerals/publications/smilicic/python-3/what-have-the-ordereddicts-ever-done-for-us/share/dfb2b6cc81377109e761657637a58c4d/">this solution</a> for example. And my comment of course. :-)

not_real on April 3, 2021, 7:53 a.m.
🤯 Really elegant use of <a href="https://docs.python.org/3/library/enum.html">enum</a>

r_tchaik on April 18, 2021, 1:08 p.m.
Short and very elegant solution. I like it very much.

rybld2 on May 16, 2021, 10:55 a.m.
it is indisputably a virtuoso exercise. Why not turn it into a lesson with a few comments ?

veky on May 16, 2021, 1:02 p.m.
I did, see above. :-) What do you need to know?

_Chico__5d82cee4f5234e78a5c65e on May 17, 2021, 12:32 p.m.
OMG what's that?!

veky on May 17, 2021, 1:04 p.m.
A solution. :-D

_Chico__5d82cee4f5234e78a5c65e on May 17, 2021, 1:04 p.m.
Man, i now you would say something like that xD

veky on May 17, 2021, 1:10 p.m.
Then why did you ask? :-]

_Chico__5d82cee4f5234e78a5c65e on May 17, 2021, 1:12 p.m.
good question...i don't even know the answer :-)

veky on May 17, 2021, 1:12 p.m.
Let me try to offer one: because CiO gives you some points for every comment that you make here? :-]

jsg-inet on Aug. 13, 2021, 9:57 a.m.
Great! Lot of learning for me here: Enum, defining a class to solve the task,... I have to study more this solution to squeeze all knoledge inside it!

Adrian_Sanz_Wallace on Oct. 20, 2021, 8:18 a.m.
I like this solution, but maybe I would have put it in creative because of how weird it is. Very Nice!! +5

veky on Oct. 20, 2021, 9:23 a.m.
We had this discussion before. "Weird" is not the same as "creative". :-] And if you look closer at the other adjectives, you'll see "educational" in the "Clear" category description. I'm sure you'll agree this qualifies as educational. ;-)

tssrkt777 on Dec. 6, 2021, 2:07 a.m.
Very nice solution. :)

TrangOul on Dec. 6, 2021, 10:46 a.m.
A clever approach with defined 4 and 9 (and higher powers of 10). It'd be arguably better if you started with the single "official" symbols only (I, V, X...) and generated the required combinations in the code instead of hardcoding them.

veky on Dec. 6, 2021, 7:14 p.m.
"Arguably" is relative. :-) For a long time through history, subtractive rules <a href="https://www.encyclopedia.com/science/encyclopedias-almanacs-transcripts-and-maps/roman-numerals-their-origins-impact-and-limitations">weren't so uniform</a> as they are today. E.g. for a long time, 4 was IIII. This can be achieved in my solution by simply deleting line 15, while it would be much harder with your approach. I must say I also have <a href="https://py.checkio.org/mission/roman-numerals/publications/veky/python-3/history/share/39d8049cdda5f405fa95586c7752ff7a/">a solution implementing your idea</a>, though it's there only because it has an important museological value. :-D

H0r4c3 on July 10, 2022, 6:13 p.m.
Great solution, great explanations!

Sinv on Aug. 1, 2022, 10:57 a.m.
The solution is impressive. Short and beautiful.

veky on Aug. 1, 2022, 4:17 p.m.
There are much shorter ones. But beautiful it is. :-)

sMiTeOne on Dec. 10, 2022, 9:49 a.m.
A good example of using the module Enum!

veky on Dec. 11, 2022, 4:48 a.m.
Well, I don't think this is a good example, but it surely is an interesting one. :-)

DDela on Feb. 2, 2023, 7:33 a.m.
Code-review should be at least 30 symbols long and it should be a commment for the solution (not a reply to another comment).

DDela on Feb. 8, 2023, 10:39 a.m.
Code-review should be at least 30 symbols long and it should be a commment for the solution (not a reply to another comment).

DDela on April 26, 2023, 9:56 a.m.
Code-review should be at least 30 symbols long and it should be a commment for the solution (not a reply to another comment).

Pavellver on Oct. 5, 2023, 9:10 a.m.
I like the use of yield and overall how short the code is.

GitHub

0 %

You ClassRoom Demo is Ready With the new tool CheckiO ClassRoom, you will be able to analyze your students' progress, have your own leaderboard, change missions order on the map, and so on.