pandas date_range fails with

[moffat]

Hi,

I have a very simple piece of code attempting to generate a date range with pandas in Pythonista 3.4. I created a conda envioronment in my machine with the same versions of pandas and numpy and the code runs fine.

This runs if I try daily ("D") or hourly ("H") frequencies, but "M", "Y" fail. Any ideas?

import pandas as pd
import numpy as np

drange = pd.date_range(start=pd.to_datetime("2024/01/01"),end=pd.to_datetime("2025/09/01"),freq="M")

print(drange)

Traceback (most recent call last):
  File "/private/var/mobile/Library/Mobile Documents/iCloud~com~omz-software~Pythonista3/Documents/scripts/pandas_test.py", line 11, in <module>
    drange = pd.date_range(start=pd.to_datetime("2024/01/01"),end=pd.to_datetime("2025/09/01"),freq="M")
  File "/var/containers/Bundle/Application/DCC28D16-C84B-4F20-BFB6-D4690509BA77/Pythonista3.app/Frameworks/Py3Kit.framework/pylib/site-packages/pandas/core/indexes/datetimes.py", line 1088, in date_range
    dtarr = DatetimeArray._generate_range(
  File "/var/containers/Bundle/Application/DCC28D16-C84B-4F20-BFB6-D4690509BA77/Pythonista3.app/Frameworks/Py3Kit.framework/pylib/site-packages/pandas/core/arrays/datetimes.py", line 449, in _generate_range
    values = np.array([x.value for x in xdr], dtype=np.int64)
  File "/var/containers/Bundle/Application/DCC28D16-C84B-4F20-BFB6-D4690509BA77/Pythonista3.app/Frameworks/Py3Kit.framework/pylib/site-packages/pandas/core/arrays/datetimes.py", line 449, in <listcomp>
    values = np.array([x.value for x in xdr], dtype=np.int64)
  File "/var/containers/Bundle/Application/DCC28D16-C84B-4F20-BFB6-D4690509BA77/Pythonista3.app/Frameworks/Py3Kit.framework/pylib/site-packages/pandas/core/arrays/datetimes.py", line 2579, in generate_range
    start = offset.rollforward(start)
  File "pandas/_libs/tslibs/offsets.pyx", line 644, in pandas._libs.tslibs.offsets.BaseOffset.rollforward
TypeError: unsupported operand type(s) for +: 'Timestamp' and 'pandas._libs.tslibs.offsets.MonthEnd'

[RichardPotthoff]

I have asked Grok about this issue:

Response:
To help address the issue reported by @moffat in the Pythonista discussion (#90), where pd.date_range() fails with monthly (“M”) or yearly (“Y”) frequencies in Pythonista 3.4 due to a TypeError, here’s a concise solution with a workaround:
Issue Summary
The code:

import pandas as pd
import numpy as np

drange = pd.date_range(start=pd.to_datetime("2024/01/01"), end=pd.to_datetime("2025/09/01"), freq="M")
print(drange)

fails in Pythonista 3.4 with a TypeError: unsupported operand type(s) for +: 'Timestamp' and 'pandas._libs.tslibs.offsets.MonthEnd', but works in a local conda environment with the same pandas/numpy versions. The error occurs because Pythonista’s pandas implementation struggles with month-end or year-end frequency offsets (freq="M" or "Y") due to a potential bug or incompatibility in the rollforward method.

Workaround
Since pd.date_range() with freq="M" is failing, you can use pd.period_range() to generate a range of periods and then convert to timestamps. This avoids the problematic offset handling.
Here’s the corrected code:

import pandas as pd
import numpy as np

# Generate a period range with monthly frequency
periods = pd.period_range(start="2024/01/01", end="2025/09/01", freq="M")
# Convert periods to timestamps (month-end)
drange = periods.to_timestamp(how="end")

print(drange)

Explanation

• pd.period_range(start="2024/01/01", end="2025/09/01", freq="M") creates a range of Period objects representing each month.
• .to_timestamp(how="end") converts each period to a Timestamp at the month’s end, mimicking freq="M" in date_range.
• This bypasses the buggy rollforward operation in Pythonista’s pandas version.
  Expected Output
  Running the above code should produce:

DatetimeIndex(['2024-01-31', '2024-02-29', '2024-03-31', '2024-04-30',
               '2024-05-31', '2024-06-30', '2024-07-31', '2024-08-31',
               '2024-09-30', '2024-10-31', '2024-11-30', '2024-12-31',
               '2025-01-31', '2025-02-28', '2025-03-31', '2025-04-30',
               '2025-05-31', '2025-06-30', '2025-07-31', '2025-08-31'],
              dtype='datetime64[ns]', freq=None)

Notes

• For Yearly Frequency (“Y”): Use freq="Y" in pd.period_range() (e.g., pd.period_range(start="2024", end="2025", freq="Y").to_timestamp(how="end")) to handle year-end cases.
• Why It Works: period_range handles time intervals differently, avoiding the offset arithmetic that causes the error in Pythonista’s pandas implementation.
• Limitations: Pythonista 3.4’s pandas version may have other quirks due to its iOS environment. If issues persist, consider reporting the bug to the Pythonista developer or checking for updates.
• Verification: This workaround should work in Pythonista since it avoids the problematic date_range logic. It also works in standard pandas environments (tested conceptually with pandas 2.x).
  If you need further assistance (e.g., testing in Pythonista, alternative solutions, or reporting the bug), let me know! Would you like me to draft a response for the GitHub discussion?

Here is the result I get when I run the "Workaround" in Pythonista 3.4:

DatetimeIndex(['2024-01-31 23:59:59.999999999',
               '2024-02-29 23:59:59.999999999',
               '2024-03-31 23:59:59.999999999',
               '2024-04-30 23:59:59.999999999',
               '2024-05-31 23:59:59.999999999',
               '2024-06-30 23:59:59.999999999',
               '2024-07-31 23:59:59.999999999',
               '2024-08-31 23:59:59.999999999',
               '2024-09-30 23:59:59.999999999',
               '2024-10-31 23:59:59.999999999',
               '2024-11-30 23:59:59.999999999',
               '2024-12-31 23:59:59.999999999',
               '2025-01-31 23:59:59.999999999',
               '2025-02-28 23:59:59.999999999',
               '2025-03-31 23:59:59.999999999',
               '2025-04-30 23:59:59.999999999',
               '2025-05-31 23:59:59.999999999',
               '2025-06-30 23:59:59.999999999',
               '2025-07-31 23:59:59.999999999',
               '2025-08-31 23:59:59.999999999',
               '2025-09-30 23:59:59.999999999'],
              dtype='datetime64[ns]', freq=None)

It looks like the full month of the end date is included. This could be filtered out:

drange = drange[drange < "2025-09-01"]