why am i getting TypeError: dtype datetime64[ns] cannot be converted to timedelta64[ns]?-CodePudding

The code:

import pandas as pd
import datetime as dt
d1 = pd.to_datetime(df.dates, errors='coerce')
d1 = pd.to_timedelta(d1).dt.days

i have lots of dates in a column which i want to convert to number of days but i simply keep getting errors. im new to pandas so sorry if the question is silly

but why am i getting errors and how to fix them.

TypeError                                 Traceback (most recent call last)
<ipython-input-119-298bf748984e> in <module>()
      1 d1 = pd.to_datetime(df.dates, errors='coerce')
----> 2 d1 = pd.to_timedelta(d1).dt.days

2 frames
/usr/local/lib/python3.7/dist-packages/pandas/core/tools/timedeltas.py in to_timedelta(arg, unit, errors)
    122         return arg
    123     elif isinstance(arg, ABCSeries):
--> 124         values = _convert_listlike(arg._values, unit=unit, errors=errors)
    125         return arg._constructor(values, index=arg.index, name=arg.name)
    126     elif isinstance(arg, ABCIndex):

/usr/local/lib/python3.7/dist-packages/pandas/core/tools/timedeltas.py in _convert_listlike(arg, unit, errors, name)
    171 
    172     try:
--> 173         td64arr = sequence_to_td64ns(arg, unit=unit, errors=errors, copy=False)[0]
    174     except ValueError:
    175         if errors == "ignore":

/usr/local/lib/python3.7/dist-packages/pandas/core/arrays/timedeltas.py in sequence_to_td64ns(data, copy, unit, errors)
   1018     else:
   1019         # This includes datetime64-dtype, see GH#23539, GH#29794
-> 1020         raise TypeError(f"dtype {data.dtype} cannot be converted to timedelta64[ns]")
   1021 
   1022     data = np.array(data, copy=copy)

TypeError: dtype datetime64[ns] cannot be converted to timedelta64[ns]

Thank you

CodePudding user response：

To get the number of days to now, use:

today = pd.Timestamp.today().normalize()
d1 = (today - pd.to_datetime(df['dates'], errors='coerce')).dt.days
print(d1)

# Output
0    2910
1    4993
2    4471
3    3973
4    7293
5    3183
Name: dates, dtype: int64

CodePudding user response：

@Corralien's solution is great! I just wanted to add some context for learning purposes:

Notice that it returns a TypeError because the method timedelta is meant to be applied to a range of time, a delta. In Pandas "linguo" it is a datetimelike value. So if you pass regular dates to it, the error you got is somewhat expected.

From the docs, you can see that the TypeError is mentioned right at the bottom.

Try this out to see the types you get:

d1 = pd.to_datetime(df['dates'], errors='coerce')
print(d1.dtype)
print(type(d1))
print(type(d1.iloc[0]))

Then this should be the result:

datetime64[ns]
<class 'pandas.core.series.Series'>
<class 'pandas._libs.tslibs.timestamps.Timestamp'>

Thus:

TypeError: dtype datetime64[ns] cannot be converted to timedelta64[ns]