[BUG] converting Pandas DataFrame into TimeSeries object

[Nikodemski2002]

Describe the bug
Hi, I have a problem with converting Pandas DataFrame into Darts TimeSeries object namely, my Pandas DataFrame consists of the datetime64 index and 4 columns of float64 data I want to have in the TimeSeries object. After executing below code, I get an error: coordinate component has dimensions ('_field',), but these are not a subset of the DataArray dimensions ('_time', 'component', 'sample')

To Reproduce
TSDF = TimeSeries.from_dataframe(modeldf)

Expected behavior
Getting a Darts TimeSeries object.

System (please complete the following information):

• Python version: 3.9
• darts version 0.23.1

Additional context

Kind regards
Nikodem

[alexcolpitts96]

It is hard to infer what is going on from the provided info; however, here is a minimal example that should work:

import darts
import pandas as pd
import numpy as np

df = pd.DataFrame(
            data=np.random.randint(0, 100, size=(100, 4)), columns=list("ABCD"),
            index=pd.date_range(start=0, periods=100, freq='1D'),
        )

ts = darts.TimeSeries.from_dataframe(df)

If that works and your code is still not working, you should provide more information about your code. I also will recommend reinstalling darts or creating a new venv, conda env, etc. To avoid having to create a new environment, you can force reinstall darts using

pip install darts --ignore-installed

[hrzn]

+1, the code TSDF = TimeSeries.from_dataframe(modeldf) is not reproducible. Please provide a fully self-contained code snippet.

[mhadi4194]

I ran into same issue.
looks like issue is because of column name index.
here is code to reproduce:

df2 = pd.DataFrame(
            data=np.random.randint(0, 100, size=(100, 1))
            ,index=pd.date_range(start=0, periods=100, freq='1D'),columns=['y']
        ).reset_index()
df2 = df2[df2['index'].dt.day != 2] # data has incomplete date
df2['id'] = list("ABCD") * 25 # multiple time series
df2 = df2.pivot(columns='id',values='y',index='index').fillna(value=0).reset_index()

series = TimeSeries.from_dataframe(df2
                                   ,time_col='index'
                                   ,fill_missing_dates=True
                                   ,freq='1D'
                                   
                                   ,fillna_value=0
                                  )

Error message:

ValueError: coordinate component has dimensions ('id',), but these are not a subset of the DataArray dimensions ('index', 'component', 'sample')

solutions:
remove column index name:

df2.columns.name = None

[madtoinou]

Thank you @mhadi4194 for pinpointing the source of the error and suggesting a solution, I opened a PR with a fix.