[BUG] "Date" must be type "object"

[kampai-shp]

Expected Behavior

I was able to use the examples to get daily OHLCV data to work, but could not intraday data to chart. After a couple of days of going through code and testing various scenarios I think I got to the core issue. "Date" must be of type "object".

Problem arises using Polars to read a text file in CSV format. Pandas used to support "infer_datetime_format" but this was deprecated in version 2.0.0. Polars has "try_parse_dates" which returns a datetime64[us] type for intraday data. This type is most common for intraday analysis of stock prices.

So inputting data in Polars, then applying "to_pandas()" results in a datetime64[us] "Date" format, which fails in lightweight_charts. Applying:
df_polars["Date"] = df_polars["Date"].astype("object") works.

NOTE: This is not a "Polars problem". The issue is lightweight_charts not handling ["Date"] inputs that are not of type "object". There is code in _df_datetime_format in abstract/SeriesCommon that manipulates the Date or Time column names so I am using "Date" as the generic name.

Current Behaviour

The examples for daily & intraday (examples 1 & 3) provide an Excel .csv file. When this is read by Pandas, the type for the datetime is "object". This works, as referenced by running the examples. A simple "print(df.dtypes)" shows the data type.

Running the attached CSV-formatted text file in the reproducible example below shows how lightweight charts fails if the datetime format is not "object".

MP_test_data.txt

I believe the issue may be in abstract/SeriesCommon:
def _set_interval(self, df: pd.DataFrame):
if not pd.api.types.is_datetime64_any_dtype(df['time']):
df['time'] = pd.to_datetime(df['time'])

_set_interval wants an object type and if given another type, it skips this, which causes downstream issues.

Reproducible Example

import pandas as pd
import polars as pl
from lightweight_charts import Chart

if __name__ == '__main__':

    chart = Chart()

    df_pandas = pd.read_csv("MP_test_data.txt", header=None, names=["Date", "Open", "High", "Low", "Close", "Volume"])
    # infer_datetime_format was deprecated in Pandas 2.0.0
    print(df_pandas)
    print(df_pandas.dtypes) # Date is object

    df_polars = pl.read_csv(
        "MP_test_data.txt",
        has_header=False,
        new_columns=["Date", "Open", "High", "Low", "Close", "Volume"],
        try_parse_dates=True,
    ).to_pandas()
    print(df_polars)
    print(df_polars.dtypes) # Date is datetime64[us]

    # uncomment to make polars version work
    # df_polars["Date"] = df_polars["Date"].astype("object")

    # uncomment to show pandas works
    chart.set(df_pandas)

    # polars path - fails unless line above is uncommented
    # chart.set(df_polars)

    chart.show(block=True)

Environment

- OS: Windows 11
- Library: 2.1