carsonfarmer / streamhist Goto Github PK

Hi,
I found a problem when I was using it.

>>> from streamhist import StreamHist
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.6/site-packages/streamhist/__init__.py", line 9, in <module>
    from histogram import StreamHist
ModuleNotFoundError: No module named 'histogram'

I tried to install ,But I got No matching distribution found for histogram.
I found this package on pypi(histogram), but I can't download it, returned 404.

How to solve this problem?

Thank you!

Hey Carson!

Unfortunately, I noticed that the current version of your library has some serious bug in the sum() function. Running your examples leads (soon or later) to the following result:

I've dug a little bit into it and, most likely, this is the commit that causes it: a937507

I know that @belm0 wrote it to fix the function and quantiles, but it looks like tests do not cover everything needed.

On the other hand, using an earlier version, e.g. 7dc5169, leads to:

u = (bin_i.value + (bin_i1.value - bin_i.value) * z) -> TypeError: unsupported operand type(s) for *: 'float' and 'NoneType'

in _compute_quantile(). So it's something that @belm0 probably fixed, but it turns out that there was a price for it ;) I know that you don't actively support the repository, but I think it's important for everyone to be aware of the problem.

would you make a git tag corresponding to the pypi release?

probably:

git tag 0.1.0 28fac98
git push --tags

What's blocking release to pip?

quantiles output can be far out of the histogram input range:

>>> h = streamhist.StreamHist()
>>> h.update((5, 7, 70))
>>> h.quantiles(.5, .9, .95)
[-24.5, 70.0, 70.0]

Hello,
Since it seems that this repository is not maintained anymore, and I need such a library, I made another implementation. It is available on pypi and is also hosted on github:
https://github.com/maki-nage/distogram

This may interest other people looking for a python implementation of this algorithm.

Unhandled exception seen in tests and real world use:

======================================================================
ERROR: streamhist.test.test_histogram.test_median_mean
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/.../python3.7/site-packages/nose/case.py", line 198, in runTest
    self.test(*self.arg)
  File "streamhist/streamhist/test/test_histogram.py", line 97, in test_median_mean
    assert about(h.median(), 0, 0.05)
  File "streamhist/streamhist/histogram.py", line 164, in median
    return self.quantiles(0.5)[0]
  File "streamhist/streamhist/histogram.py", line 407, in quantiles
    qq = _compute_quantile(target_sum, bin_i, bin_i1, prev_sum+1)
  File "streamhist/streamhist/histogram.py", line 462, in _compute_quantile
    u = (bin_i.value + (bin_i1.value - bin_i.value) * z)
TypeError: unsupported operand type(s) for *: 'float' and 'NoneType'

Can I add weighted data such as tuple of (value, weight) to the histogram?