Problem plotting historical time series

[elysiumfield]

I have recently updated plotly to ver 3.5.0 and am having problems plotting historical time series. The test code is as follows:

import numpy as np
import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
dt = pd.date_range(start='1969-12-31', periods=270, freq='Q')
t1 = []
t1.append(dict(x=dt, y=np.ones(270), name='cape', type='scatter'))
layout = go.Layout(title="test", width=1800, height=930, font=dict(size=27), titlefont=dict(size=42))
fig = go.Figure(data=t1, layout=layout)
pyo.plot(fig, filename='test.html', auto_open=False, show_link=False)

This code returns the error:

  File "C:\ProgramData\Anaconda3\lib\site-packages\plotly\offline\offline.py", line 645, in plot
    '100%', '100%', global_requirejs=False)

  File "C:\ProgramData\Anaconda3\lib\site-packages\plotly\offline\offline.py", line 336, in _plot_html
    jdata = _json.dumps(figure.get('data', []), cls=utils.PlotlyJSONEncoder)

  File "C:\ProgramData\Anaconda3\lib\json\__init__.py", line 238, in dumps
    **kw).encode(obj)

  File "C:\ProgramData\Anaconda3\lib\site-packages\plotly\utils.py", line 168, in encode
    encoded_o = super(PlotlyJSONEncoder, self).encode(o)

  File "C:\ProgramData\Anaconda3\lib\json\encoder.py", line 199, in encode
    chunks = self.iterencode(o, _one_shot=True)

  File "C:\ProgramData\Anaconda3\lib\json\encoder.py", line 257, in iterencode
    return _iterencode(o, 0)

  File "C:\ProgramData\Anaconda3\lib\site-packages\plotly\utils.py", line 233, in default
    return encoding_method(obj)

  File "C:\ProgramData\Anaconda3\lib\site-packages\plotly\utils.py", line 297, in encode_as_datetime
    obj = obj.astimezone(pytz.utc)

OSError: [Errno 22] Invalid argument

However, defining dt = pd.date_range(start='1970-03-31', periods=270, freq='Q'), everything works fine. I am positive that previous versions of plotly could plot datetimeindex that was older.

[jonmmease]

Hi @elysiumfield, thanks for the report.

Could you add the versions of Python and pandas that you're using?

I'm not able to reproduce the error with Python 3.6, pandas 0.23, and and plotly 3.5.0 on MacOS. I get the following figure:

Thanks!

[jonmmease]

Also, have you seen this Stack Overflow thread? https://stackoverflow.com/questions/37494983/python-fromtimestamp-oserror

Could you try the code example from this comment https://stackoverflow.com/a/45372194/4551895:

import datetime
import platform
print(
    "Running on Python ver.{} on {} {}\n" \
    .format(
        platform.python_version(),
        platform.system(),
        platform.release()
        )
)
for timestamp in range(1, 100000000):
    try:
        dt = datetime.datetime.fromtimestamp(timestamp)
    except:
        pass
    else:
        break
print(
    "Smallest accepted Unix timestamp by {}: '{}' ({})" \
    .format(platform.system(), timestamp, dt)
)

On OS X I get

Running on Python ver.3.6.8 on Darwin 18.0.0

Smallest accepted Unix timestamp by Darwin: '1' (1969-12-31 19:00:01)

[elysiumfield]

Hi @jonmmease, I think you are on the right track. I am using Python 3.7 and Pandas 0.23.4. Running the code you posted, I got:

Running on Python ver.3.7.1 on Windows 7
Smallest accepted Unix timestamp by Windows: '1' (1970-01-01 01:00:01)

I also get the same problem as in the link you posted, but it doesn't say how to solve it.

[elysiumfield]

I have tested a bit more. Tried my test code on another other computer and it worked. That one is having:

Running on Python ver.3.6.4 on Windows 7
Smallest accepted Unix timestamp by Windows: '86400' (1970-01-02 01:00:00)

I updated Plotly to ver 3.5.0 and it still works. But datetime.datetime.fromtimestamp(-43201, pytz.utc) gives [Errno 22] error.

Not sure that my plot problem is related to datetime in Windows after all. The main difference I know of between my test on the two different computers is python 3.7 vs 3.6, but there are probably more differences due to dependencies on other packages.

[mprostock]

Problem can be changed by a 1-character-fix in the code ;-)
See PR : #1501

[elysiumfield]

OK great, thanks.

[jonmmease]

Fixed by #1501