Pandas support

[wimgielis]

Hello all,

So I installed Pythonista and I am generally very happy with the result.

When I am now trying to use my 4K lines script in Python, while it works fine on the laptop, I get issues on Pandas timezone conversions (or so it seems). I am pretty sure it has to do with older versions of Pandas whereas it is more recent on my laptop.

Is this something that others have struggled with as well? Any solution I could adopt?

The script is about a hike and different stops along the way. I calculate the difference in time between successive stops. Pythonista and Pandas for some reason do not like it. On the laptop works well.

[wimgielis]

I get:
TypeError: Cannot compare tz-naive and tz-aware datetime-like objects

ChatGPT said:
That error is actually a much stronger clue than the original one.

TypeError: Cannot compare tz-naive and tz-aware datetime-like objects

This means that somewhere in your data you're mixing:

timezone-aware datetimes (e.g. 2026-07-21 20:02:23+02:00)
timezone-naive datetimes (e.g. 2026-07-21 20:02:23 with no tzinfo)
Pandas 2.3.3 often normalizes these cases more gracefully. Pandas 1.4.4 is much less forgiving, and the earlier 'NoneType' object has no attribute 'total_seconds' may simply be a consequence of that older implementation.

[wimgielis]

Still working on this currently. Maybe it has to do with the code, which is exposed in a different way in older Pandas, and treated gracefully in newer pandas. I will come back on this one.

[wimgielis]

This is the decisive clue. Your data is not the problem.

Both environments have:

datetime64[ns, Europe/Brussels]

and both have:

<class 'pandas._libs.tslibs.timestamps.Timestamp'>

So the Series is already a proper timezone-aware pandas datetime column.

The failure happens when Pythonista's pandas 1.4.4 tries to interpret the timezone metadata for Europe/Brussels:

pandas/_libs/tslibs/timezones.pyx", line 266, in get_dst_info
AttributeError: 'NoneType' object has no attribute 'total_seconds'

This is coming from the older pandas timezone handling code, not from your subtraction.

Why the subtraction triggers it

This line:

_df["RAW_cumulative time"] = (
_df["RAW_scan time"] - _df["RAW_scan time"].iloc[0]
)

forces pandas to align and operate on the timezone-aware datetime array. During that operation, pandas 1.4.4 asks the timezone object for DST transition information.

The timezone is:

zoneinfo.ZoneInfo(key="Europe/Brussels")

and pandas 1.4.4 was not fully compatible with Python's zoneinfo objects in all cases. It expected older pytz-style timezone objects, where DST metadata was available in the way pandas expected.

Your laptop has pandas 2.3.3, which fixed/improved this area.

Recommended fix: use pytz for pandas 1.4.4 compatibility

In your Utils.now_local() function, change:

return datetime.now(ZoneInfo(timezone))

to:

import pytz

return datetime.now(pytz.timezone(timezone))

Then your timestamps will carry:

tzinfo=<DstTzInfo 'Europe/Brussels' ...>

instead of:

tzinfo=zoneinfo.ZoneInfo(key='Europe/Brussels')

Pandas 1.4.4 handles the former much better.

Alternative: convert before creating the DataFrame

If you want to keep ZoneInfo everywhere else, normalize only at the pandas boundary:

import pytz

_df[column_name] = _df[column_name].apply(
lambda x: x.astimezone(pytz.timezone("Europe/Brussels"))
)

However, I would prefer fixing the timestamp creation point because then all downstream calculations stay consistent.

Longer-term option

If Pythonista allows it, upgrading pandas from 1.4.4 would likely solve this without code changes. But given Pythonista's packaging constraints, switching from zoneinfo to pytz is probably the cleanest compatibility fix.

One small note: zoneinfo is the modern Python approach, so your original implementation was correct for current Python. The change is mainly to accommodate the older pandas stack on iOS.

[wimgielis]

What I observe is the following:

• due to an older Pandas version in Pythonista we need to deal with tzinfo in a different way. ChatGPT helped me, then it worked.
• pandas df.to_markdown() does not work. ImportError: Missing optional dependency 'tabulate'.
• while debugging the script on the iPad, the Pythonista app would crash very often. When I then start the app again it closes immediately since, presumably, it wants to rerun my script which again crashes Pythonista. The only thing that works for me is to delete the Python script from the iCloud folder. When Pythonista starts, it cannot find the py file anymore and gives up trying to execute it. Obviously then I lose my script and need to copy/paste again on my laptop in a OneDrive folder location. Then on the iPad need to import that script again.

Unless I am missing essential skills/knowledge, this was not much fun as one of the first uses of the Pythonista app.