Add order_by parameter to pandas.Series.value_counts()

[gamesbook]

If possible, it would be really useful to allow for an extra parameter that can override the default behaviour of the pandas.Series.value_counts() function to allow for sorting by by another series. we are looking to push the result out to a bar plot, and it makes sense to have the category (x-axis) displayed in a particular order.

Thanks!

[jreback]

can you give an example of you want?

does order/sort_index not do what you want?

[gamesbook]

Here is an example of the code we are trying to run:

import pandas
import matplotlib.pyplot as plt
import numpy as np

dat = pandas.Series([1,1,1,2,2,3,3,3,3,4,4,4,5,5,5,5,6])
fig = plt.figure(figsize=(6,6))
freq= dat.value_counts()
width = 1.0
plt.ylabel('Frequency')
plt.xlabel('Items')
ax = plt.axes()
ax.set_xticks(np.arange(len(freq)) + (width / 2))
ax.set_xticks(freq)
freq.plot(kind='bar')
plt.show()

You will see that items appear on the x-axis in order 5,3,4,1,2,6 - as opposed to 1,2,3,4,5,6.

Maybe its possible to somehow incorporate the collections module code into the pandas value_counts() routine; the values() and keys() from collections are exactly what is needed for the y and x axes respectively.

import collections
a = [1,1,4,4,2,3,4,1,5,1,4,2,2,4,2,2,3,1,2,2,3,3,4,5,5]
a.sort()
print a
counter=collections.Counter(a)
print(counter) # ({2: 7, 4: 6, 1: 5, 3: 4, 5: 3})
print(counter.keys()) # [1, 2, 3, 4, 5]
print(counter.values()) # [5, 7, 4, 6, 3]
print(counter.most_common(3)) # [(2, 7), (4, 6), (1, 5)]

[jtratner]

Why can't you just sort the results yourself after getting the value counts? That's going to be exactly the same as an internal function...

eg.

dat = pandas.Series([1,1,1,2,2,3,3,3,3,4,4,4,5,5,5,5,6])
fig = plt.figure(figsize=(6,6))
freq= dat.value_counts().sort_index()

Bam, now it's sorted 1-6

[gamesbook]

Yes, that works,. My apologies - I was told by the programmer that they had tried it and it did not work.