Bug: pct_change with frequency set as 'BM' throws value error

[min2bro]

Code Sample, a copy-pastable example if possible

import pandas as pd
import random
import numpy as np


Creating the time-series index 
n=60
index = pd.date_range('01/13/2020', periods = 70,freq='D') 
  
Creating the dataframe  
df = pd.DataFrame({"A":np.random.uniform(low=0.5, high=13.3, size=(70,)), 
                   "B":np.random.uniform(low=10.5, high=45.3, size=(70,)),  
                   "C":np.random.uniform(low=70.5, high=85, size=(70,)), 
                   "D":np.random.uniform(low=50.5, high=65.7, size=(70,))}, index = index) 


df.pct_change(freq='BM')

Problem description

For a time-series data, the df.pct_change(freq='BM') doesn't works and throws following error.

ValueError: cannot reindex from a duplicate axis

Expected Output

df.asfreq('BM').pct_change()

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.6.3.final.0
python-bits : 64
OS : Windows
OS-release : 10
machine : AMD64
processor : Intel64 Family 6 Model 142 Stepping 9, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : None.None

pandas : 0.25.1
numpy : 1.15.4
pytz : 2017.2
dateutil : 2.6.1
pip : 18.0
setuptools : 39.1.0
Cython : 0.26.1
pytest : 5.1.2
hypothesis : None
sphinx : 1.6.3
blosc : None
feather : None
xlsxwriter : 1.0.2
lxml.etree : 4.1.0
html5lib : 0.9999999
pymysql : None
psycopg2 : 2.7.5 (dt dec pq3 ext lo64)
jinja2 : 2.9.6
IPython : 6.1.0
pandas_datareader: None
bs4 : 4.6.0
bottleneck : 1.2.1
fastparquet : None
gcsfs : None
lxml.etree : 4.1.0
matplotlib : 2.1.0
numexpr : 2.6.2
odfpy : None
openpyxl : 2.4.8
pandas_gbq : None
pyarrow : 0.8.0
pytables : None
s3fs : None
scipy : 1.0.0
sqlalchemy : 1.1.13
tables : 3.4.2
xarray : 0.10.7
xlrd : 1.1.0
xlwt : 1.3.0
xlsxwriter : 1.0.2

[dongho-jung]

Hello, I have found that this happens when the offset which is made of freq is anchored. (offset.isAnchored)

so I propose that when the freq gets anchored, make the data resampled first, before the next line
rs = data.div(data.shift(periods=periods, freq=freq, axis=axis, **kwargs)) - 1

--- a/pandas/core/generic.py
+++ b/pandas/core/generic.py
@@ -10413,6 +10413,8 @@ class NDFrame(PandasObject, SelectionMixin):
         else:
             data = self.fillna(method=fill_method, limit=limit, axis=axis)

+        if freq and to_offset(freq).isAnchored():
+            data = data.asfreq(freq)
         rs = data.div(data.shift(periods=periods, freq=freq, axis=axis, **kwargs)) - 1
         rs = rs.reindex_like(data)
         if freq is None:

[dongho-jung]

However, I couldn't run tests yet because, in the latest commit c4489c, it spewed out a lot of fails.

[min2bro]

@0xF4D3C0D3 Can you change the label of this issue as Bug ?

[dongho-jung]

No, I think I can't do that because I'm not a contributor nor member.

[dongho-jung]

Oh, the c4489c was innocent, the fails were on my own mistake. So I have run the tests with the above change.

and now I'm running pytest pandas now

[dongho-jung]

good! the tests are passed. I'll create the PR