perrygeo / jenks Goto Github PK

Supply wheels for Windows to make installation easier, since getting Cython and a C compiler set up on Windows is prohibitively difficult for most users.

There are project templates which can do the majority of the work of using AppVeyor for this, e.g. https://github.com/ionelmc/cookiecutter-pylibrary.

It seems to me that we should include variance_combinations in the output of jenks.jenks() instead of discarding the data. This would allow one to compute the goodness of variance fit, for example.

I suggest:

(...)
lower_class_limits, variance_combinations = jenks_matrices(data, n_classes)
(...)
return {'breaks':kclass, 'variance_combinations':variance_combinations}

Hi,

I tried to install it in Windows 10 and it gives the following message:

Obtaining jenks from git+https://github.com/perrygeo/jenks.git#egg=jenks
  Cloning https://github.com/perrygeo/jenks.git to c:\users\irene\src\jenks
Requirement already satisfied: Numpy in c:\users\irene\appdata\local\conda\conda\envs\envpy34\lib\site-packages (from jenks)
Installing collected packages: jenks
  Running setup.py develop for jenks
    Complete output from command C:\Users\irene\AppData\Local\conda\conda\envs\envpy34\python.exe -c "import setuptools, tokenize;__file__='C:\\Users\\irene\\src\\jenks\\setup.py';f=getattr(tokenize, 'open', open)(__file__);code=f.read().replace('\r\n', '\n');f.close();exec(compile(code, __file__, 'exec'))" develop --no-deps:
    running develop
    running egg_info
    writing requirements to jenks.egg-info\requires.txt
    writing dependency_links to jenks.egg-info\dependency_links.txt
    writing top-level names to jenks.egg-info\top_level.txt
    writing jenks.egg-info\PKG-INFO
    warning: manifest_maker: standard file '-c' not found

    reading manifest file 'jenks.egg-info\SOURCES.txt'
    writing manifest file 'jenks.egg-info\SOURCES.txt'
    running build_ext
    cythoning jenks.pyx to jenks.c
    building 'jenks' extension
    **error: Microsoft Visual C++ 10.0 is required. Get it with "Microsoft Windows SDK 7.1":** www.microsoft.com/download/details.aspx?id=8279

Naturally, after downloading 20GB of Microsoft stuff (many SDK, Visual Studio, and other things I didnt ask for...) it keeps giving the same message. I also tried with MinGW, but keeps giving this message. Is there any workaround for this? Do you know why is this error happening?

Thanks!

Line 97 should be k = data.shape[0] instead of k = data.shape[0] - 1 (remove the -1).

Simple test: clustering this data set [1, 2, 3, 100] currently yields [1, 2, 100] but should yield [1, 3, 100]. Basically the large outlier 100 should be isolated into its own cluster.

I think it is important to note that this is a version of Jenks which does not provide an optimal solution, i.e. not Fisher-Jenks, not optimal Jenks.

import jenks
import stats

from hypothesis import example, given, assume
from hypothesis import strategies as st


@st.composite
def list_choose_clusters(draw, max_size=None, unique=False):
    xs = draw(st.lists
            ( st.integers(min_value=-10**3, max_value=10**3)
            , min_size=3, max_size=max_size, unique=unique
            ))
    k = draw(st.integers(min_value=2, max_value=len(xs)-1))
    return (xs, k)

@given(list_choose_clusters(max_size=15, unique=True))
def test_jenks_breaks(t):
    tf = lambda cs: sum(stats.variance(c, stats.mean(c)) for c in cs)
    v1 = tf(stats.jenks(*t)[1])
    v2 = tf(stats.break_list(t[0], jenks.jenks(*t)))
    assert v1 == v2

$ pytest perrygeo-tests.py
...
Falsifying example: test_jenks_breaks(
    t=([0, 1, 3, -2], 2),
)

An optimal solution is [[-2], [0, 1, 3]] with a sum variance of 1.55, not [[0, -2], [1, 3]] with a sum variance of 2.

It seems like all Python ports are derived from the same algorithm, so see mthh/jenkspy#22 for more information including documentations/links, sources and full tests.