Describe the enhancement requested
The SqlInfo enum in FlightSql.proto already captures a useful slice of SQL dialect metadata (identifier quoting, null ordering default, supported grammar, etc.), but leaves several axes uncovered that clients routinely need when compiling SQL for a specific backend. Although it's unrealistic to describe entire dialect grammar with sqlinfo flags, the goal is to expose just enough flags that will allow query engines to generate correct pushdown sql queries (predicate pushdown, projection pushdown, ...). Since adbc also adopts the same SqlInfo flags in it's spec the changes naturally apply to both flight sql and adbc.
The immediate use case for these particular flags is to enable adbc data source (apache/spark#54603) in spark w/o having to pre-configure dialects in spark code like it's currently done with jdbc. The necessary identified codes so far are:
SQL_LIMIT_OFFSET_SYNTAXSQL_NULLS_ORDERING_SYNTAXSQL_BOOLEAN_LITERAL_SYNTAXSQL_DATETIME_LITERAL_SYNTAX
Component(s)
Format, C++
Describe the enhancement requested
The
SqlInfoenum inFlightSql.protoalready captures a useful slice of SQL dialect metadata (identifier quoting, null ordering default, supported grammar, etc.), but leaves several axes uncovered that clients routinely need when compiling SQL for a specific backend. Although it's unrealistic to describe entire dialect grammar with sqlinfo flags, the goal is to expose just enough flags that will allow query engines to generate correct pushdown sql queries (predicate pushdown, projection pushdown, ...). Since adbc also adopts the same SqlInfo flags in it's spec the changes naturally apply to both flight sql and adbc.The immediate use case for these particular flags is to enable adbc data source (apache/spark#54603) in spark w/o having to pre-configure dialects in spark code like it's currently done with jdbc. The necessary identified codes so far are:
SQL_LIMIT_OFFSET_SYNTAXSQL_NULLS_ORDERING_SYNTAXSQL_BOOLEAN_LITERAL_SYNTAXSQL_DATETIME_LITERAL_SYNTAXComponent(s)
Format, C++