calculate the count by day

Question

I have a dataframe that has columns created_at and entities that look like this

    created_at                         entities
2017-10-29 23:06:28     {'hashtags': [{'text': 'OPEC', 'indices': [0, ...
2017-10-29 22:28:20     {'hashtags': [{'text': 'Iraq', 'indices': [21,...
2017-10-29 20:01:37     {'hashtags': [{'text': 'oil', 'indices': [58, ...
2017-10-29 20:00:14     {'hashtags': [{'text': 'oil', 'indices': [38, ...
2017-10-27 08:44:30     {'hashtags': [{'text': 'Iran', 'indices': [19,...
2017-10-27 08:44:10     {'hashtags': [{'text': 'Oil', 'indices': [17, ...
2017-10-27 08:43:13     {'hashtags': [{'text': 'Oil', 'indices': [0, 4...
2017-10-27 08:43:00     {'hashtags': [{'text': 'Iran', 'indices': [19,.

I want to calculate the count of entities for each day. Basically I want to receive something like

created_at    number_of_entities
2017-10-29           4
2017-10-27           4

How to do that? I have pandas 0.23.4

Answer 1

Given

>>> df
           created_at  entities
0 2017-10-29 23:06:28         1
1 2017-10-29 22:28:20         2
2 2017-10-29 20:01:37         3
3 2017-10-29 20:00:14         4
4 2017-10-27 08:44:30         5
5 2017-10-27 08:44:10         6
6 2017-10-27 08:43:13         7
7 2017-10-27 08:43:00         8

with

>>> df.dtypes
created_at    datetime64[ns]
entities               int64
dtype: object

you can issue:

>>> pd.PeriodIndex(df['created_at'], freq='D').value_counts()
2017-10-29    4
2017-10-27    4
Freq: D, Name: created_at, dtype: int64

A nicer approach without the PeriodIndex constructor has been suggested by jezrael in the comments:

>>> df['created_at'].dt.to_period('D').value_counts()
2017-10-27    4
2017-10-29    4

With some additional renaming to match your output it starts to look suspiciously like jezrael's solution. ;)

>>> datecol = 'created_at'
>>> df[datecol].dt.to_period('D').value_counts().rename_axis(datecol).reset_index(name='number_of_entities')
  created_at  number_of_entities
0 2017-10-27                   4
1 2017-10-29                   4

Alternatively, you can set the index to the dates and then resample:

>>> df.set_index('created_at').resample('D').size()
created_at
2017-10-27    4
2017-10-28    0
2017-10-29    4
Freq: D, dtype: int64

... and if converting to your exact output is necessary:

>>> resampled = df.set_index('created_at').resample('D').size()
>>> resampled[resampled != 0].reset_index().rename(columns={0: 'number_of_entities'})
  created_at  number_of_entities
0 2017-10-27                   4
1 2017-10-29                   4

---

Some more context: resample is especially useful for arbitrary time intervals, for example "five minutes". The following example is taken directly from Wes McKinney's book "Python for Data Analysis".

>>> N = 15
>>> times = pd.date_range('2017-05-20 00:00', freq='1min', periods=N)
>>> df = pd.DataFrame({'time': times, 'value': np.arange(N)})
>>> 
>>> df
                  time  value
0  2017-05-20 00:00:00      0
1  2017-05-20 00:01:00      1
2  2017-05-20 00:02:00      2
3  2017-05-20 00:03:00      3
4  2017-05-20 00:04:00      4
5  2017-05-20 00:05:00      5
6  2017-05-20 00:06:00      6
7  2017-05-20 00:07:00      7
8  2017-05-20 00:08:00      8
9  2017-05-20 00:09:00      9
10 2017-05-20 00:10:00     10
11 2017-05-20 00:11:00     11
12 2017-05-20 00:12:00     12
13 2017-05-20 00:13:00     13
14 2017-05-20 00:14:00     14
>>> 
>>> df.set_index('time').resample('5min').size()
time
2017-05-20 00:00:00    5
2017-05-20 00:05:00    5
2017-05-20 00:10:00    5
Freq: 5T, dtype: int64