Hello,

Let say we have an empty table S that represents the schema we want to keep

We have another table T partionned by column A with a schema that depends on the file we have load into. Say :

Now to make T having the same schema as S I do :

SET spark.sql.sources.partitionOverwriteMode=dynamic;
CREATE OR REPLACE TABLE T PARTITIONED BY (A) as SELECT * FROM S WHERE false;

 and the result is, as I wish :

Good. But the fact is I didn't see we can do something like that in databricks doc. Even worse it is said  that overwrite schema option (pyspark option I guess because I don't succeed to use this option with SQL) and dynamic partition don't work together. 

So here's the question : is the behavior described above a bug or a feature ?

Ps : runtime si 16.3