to_featureset() is not handling fields with null date values correctly

[tbrobin]

I am using to_featureset() to create a featureset from a spatially enabled dataframe and then using edit_features() to edit a feature layer on AGOL. I was having this issue #1693 so I rolled back to a previous version of the API. Now that 2.3.0 is out, I was going to try my luck again.

It seems there is a new bug with to_featureset(). Only some of my features manage to successfully make it to the feature layer using edit_features(). And I am getting this message when they fail.

{'addResults': [{'objectId': 1,
'uniqueId': 1,
'globalId': None,
'success': False,
'error': {'code': 1000,
'description': 'The incoming tabular data stream (TDS) remote procedure call (RPC) protocol stream is incorrect. Parameter 21 ("@applicant_closeout_date"): The supplied value is not a valid instance of data type float. Check the source data for invalid values. An example of an invalid value is data of numeric type with scale greater than precision.'}}],
'updateResults': [],
'deleteResults': []}

That field is not a float. It is a date with null values mixed in. I did some troubleshooting and it seems to_featureset() is treating that field differently than other date fields in the data with no nulls.

Here is how it treats that field if I remove the null values, and these values will work in a bulk edit using edit_features(adds=featureset.features).

Here is how it treats that field if it has null values, and it won't work using edit_features(adds=featureset.features).

[nanaeaubry]

@tbrobin Thanks for reporting this, sorry for the delayed response. I will take a look and see where this new error is coming from.
Ill post any updates!

[nanaeaubry]

@tbrobin Can you provide us with some reproducible data and code?

[tbrobin]

Dates.xlsx
``
import pandas as pd
from arcgis.features import GeoAccessor, GeoSeriesAccessor

#path to Dates excel
dates_excel=

nonulls=pd.read_excel(dates_excel,sheet_name='NoNulls')
withnulls=pd.read_excel(dates_excel,sheet_name='WithNulls')

nonulls_fs=nonulls.spatial.to_featureset()
withnulls_fs=withnulls.spatial.to_featureset()

#test feature set with nulls
print(withnulls_fs.features)

#test feature set without nulls
print(nonulls_fs.features)
`

[tbrobin]

Thanks for looking into this!

[nanaeaubry]

@tbrobin Ok I see the issue, Pandas interprets NaN values as float and not int. Hence why the entire column gets changed to float values when null dates are present.
A workaround for now would be to take all the values resulting from the feature set and type cast them as int.
We are adding a fix in the code for this, thanks for reporting it!

[tbrobin]

Great! So will this be fixed in the next ArcGIS API Release?

[nanaeaubry]

@tbrobin yes!