计算发生次数后，如何在列表中的每一天进行输入

Question

I have two columns of dates in two separate csv files. I am reading them into python, and plan to plot in matplotlib

One is for invoices:

5/1/2015
5/1/2015
5/1/2015
5/2/2015
5/2/2015
5/2/2015
5/2/2015
5/3/2015
5/3/2015
5/3/2015
5/3/2015
5/4/2015
5/4/2015
5/4/2015
5/4/2015
5/4/2015
5/4/2015
5/5/2015
5/5/2015
5/5/2015
5/5/2015
5/7/2015

And the other is for disputes:

5/1/2015
5/1/2015
5/2/2015
5/2/2015
5/3/2015
5/5/2015
5/5/2015

I want to make a list with the number of disputes divided by the number of invoices per day. So for May 1, 2015 the percent is 2/3. For May 2, 2015 2/4. May 3, 2015 is 1/4. May 4, 2015 is 0. May 5, 2015 is 2/4. There are no invoices or disputes on May 6, 2015 so the percent should be zero. Thus the list should be [.66, .5, .25, 0, .5, 0, 0]

Then I am going to graph the percents on the y axis and the date as the x-axis.

I have tried to use df.index.day but then I get groups for the first day of each month, the second, etc. I was using value_counts to count up the occurrences of each date and then dividing between two lists but I was missing numbers where I didn't have invoices or disputes, and I want there to be a value for every day.

Does anyone know a simple way to do this?

Answer 1

One easy way to get the counts is to use collections.Counter :

from collections import Counter
with open('invoice_dates') as f:
    invoice_count = Counter(line.strip() for line in f)

and similarly for dispute_count . You can then get a dictionary mapping dates to dispute percentages by

from __future__ import division # in case you are on Python 2.x
dispute_percentage = {date: dispute_count.get(date, 0) / invoices
                      for date, invoices in invoice_count.items()}

Use iteritems() instead of items() in the last line if you are on Python 2.x.

Answer 2

You should have made it more clear that you were using pandas -- there are built-in tools to aid you to do what you want to do. In this case, you can use value_counts on your f and f2 . With your example data:

>>> f = pd.to_datetime(f)
>>> f2 = pd.to_datetime(f2)
>>> f.value_counts()/f2.value_counts()
2015-05-01    0.666667
2015-05-02    0.500000
2015-05-03    0.250000
2015-05-04         NaN
2015-05-05    0.500000
2015-05-07         NaN
dtype: float64
>>> (f.value_counts()/f2.value_counts()).fillna(0.0)
2015-05-01    0.666667
2015-05-02    0.500000
2015-05-03    0.250000
2015-05-04    0.000000
2015-05-05    0.500000
2015-05-07    0.000000
dtype: float64