EvalStatement throws an AnalysisException when query includes non-existing field #47

[kortemik]

ID <433>
Reporter <@q22u>

If a DPL query contains a non-existing field, spark throws an AnalysisException.

For example:
index=index_A | eval a = typeof(12) | eval b = typeof("string") | eval c = typeof(1==2) | eval d = typeof(badfield)

should return "Invalid" instead for ... | eval d = typeof(badfield)

results in

org.apache.spark.sql.AnalysisException: cannot resolve 'badfield' given input columns: [a, partition, host, index, _raw, _time, b, sourcetype, offset, c, source];;
'Project [_raw#28, _time#29, host#30, index#31, offset#32L, partition#33, source#34, sourcetype#35, a#87, b#108, c#131, 'typeOf('badfield) AS d#143]
+- Project [_raw#28, _time#29, host#30, index#31, offset#32L, partition#33, source#34, sourcetype#35, a#87, b#108, UDF((1 = 2)) AS c#131]
+- Project [_raw#28, _time#29, host#30, index#31, offset#32L, partition#33, source#34, sourcetype#35, a#87, UDF(string) AS b#108]
+- Project [_raw#28, _time#29, host#30, index#31, offset#32L, partition#33, source#34, sourcetype#35, UDF(12) AS a#87]
+- Filter index#31 LIKE index_A
+- Relation[_raw#28,_time#29,host#30,index#31,offset#32L,partition#3

should return "Invalid" instead for ... | eval d = typeof(badfield)

Developer Note:
Perhaps when visitEvalFieldType() is called, the existence of given field should be checked?

#95

[kortemik]

Has a disabled test in EvalTest.java called "testEvalTypeofInvalid". Going to be looked at in #295