OK, I'll try. On Apr 30, 2015 06:54, "Reynold Xin" <r...@databricks.com> wrote:
> We added ExpectedInputConversion rule recently in analysis: > https://github.com/apache/spark/blob/master/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercion.scala#L647 > > With this rule, the analyzer automatically adds cast for expressions that > inherit ExpectsInputTypes. We can make all string functions inherit > ExpectsInputTypes and specify input types, so the casts are added > automatically. Would you like to submit a PR? > > > > On Wed, Apr 29, 2015 at 2:06 PM, Olivier Girardot <ssab...@gmail.com> > wrote: > >> I guess you can use cast(id as String) instead of just id in your where >> clause ? >> >> Le mer. 29 avr. 2015 à 12:13, lonely Feb <lonely8...@gmail.com> a écrit : >> >> > Hi all, we are transfer our HIVE job into SparkSQL, but we found a >> litter >> > difference between HIVE and Spark SQL that our sql has a statement like: >> > >> > select A from B where id regexp '^12345$' >> > >> > in HIVE it works fine but in Spark SQL we got a: >> > >> > java.lang.ClassCastException: java.lang.Long cannot be cast to >> > java.lang.String >> > >> > Can this statement be handled with Spark SQL? >> > >> > >
