https://community.cloudera.com/t5/Support-Questions/Pandas-udf-with-a-tuple-pyspark/m-p/190144#M152233 <P>Hey Bryan thanks so much for taking the time! I think I'm almost there! The hint about the unicode issue helping me get past the first slew of errors. I seem to be running into a length one now however:</P><PRE>@pandas_udf("array&lt;string&gt;") def stringClassifier(lookupstring, first, last): lookupstring = lookupstring.to_string().encode("utf-8") first = first.to_string().encode("utf-8") last = last.to_string().encode("utf-8") #this part takes the 3 strings above and reaches out to another library to do a string match result = process.extract(lookupstring, lookup_list, limit=4000) match_list = [item for item in result if item[0].startswith(first) and item[0].endswith(last)] result2 = process.extractOne(lookupstring, match_list) if result2 is not None and result2[0][1] &gt; 75: answer = pd.Series(list(result2[0])) return answer else: fail = ["N/A","0"] return pd.Series(fail) </PRE><P><STRONG>RuntimeError: Result vector from pandas_udf was not the required length: expected 1, got 2</STRONG></P><P>I'm initially passing three strings as variables to the function which then get passed to another library. The result is a tuple which I covert to a list then to a pandas Series object. I'm curious how I can make a 2 item array object a length of 1 ..? I'm obviously missing some basics here.</P><P>@Bryan C</P> Fri, 13 Jul 2018 05:22:44 GMT alexander_witte 2018-07-13T05:22:44Z