The eval method applies the column expression even if the function is not valid.

Even though you get an error when adding a function that doesn't exist to a CAS table object, it still adds the computation to the object. This causes an error moving forward. You either have to reset the CAS table object, or remove just the bad expression.

I feel that if you get an error when using a SAS function it should not add the expression to the CASTable object.

Example below:

Create demo DataFrame

df = pd.DataFrame({
'SASDate':[18260,18212, 18290, 18201, 18158],
'StrDate':['2009-12-29', '2009-11-11', '2010-01-28', '2009-10-31', '2009-09-18']
})

Upload DataFrame to CAS

castbl = conn.upload_frame(df,
casout = {'name':'date_table',
'caslib':'casuser',
'replace':True})

View params of CASTable object

castbl.params
[output]: {'name': 'DATE_TABLE', 'caslib': 'CASUSER(Peter)'}

Table preview

castbl.head()

Eval method bad function

castbl.eval('NewCol = badfunction(SASData)')

View params

castbl.params
[output]
{'name': 'DATE_TABLE',
'caslib': 'samples',
'computedvars': ['NewCol'],
'computedvarsprogram': 'NewCol = badfunction(SASData); NewCol = NewCol; '}

Reproduce the issue with the Python DataFrame and CASTable.eval example in the report, then inspect the CASTable eval implementation and its existing tests. Done means an invalid SAS function raises an error without leaving the computed variable or expression in castbl.params, while valid expressions continue to work.