Ramin Gharib created FLINK-40825:
------------------------------------
Summary: Support casting primitive types to VARIANT
Key: FLINK-40825
URL: https://issues.apache.org/jira/browse/FLINK-40825
Project: Flink
Issue Type: Sub-task
Components: Table SQL / API
Reporter: Ramin Gharib
Assignee: Ramin Gharib
FLIP-521 lists the types that can be cast to and from VARIANT. FLINK-37925 and
FLINK-37926 implemented the direction VARIANT to SQL type. The direction SQL
type to VARIANT is still missing: \{{SqlCastFunction}} rejects every cast to
VARIANT (see the TODO in \{{canCastFrom}}).
This ticket adds \{{CAST}} and \{{TRY_CAST}} from primitive types to VARIANT.
Constructed types (ARRAY, MAP, ROW, STRUCTURED) follow in separate sub-tasks.
h3. Rule
A type can be cast to VARIANT only if a VARIANT kind holds its value without
loss. Any other type is rejected at validation time, never at runtime. The cast
is explicit only.
||Source type||Stored VARIANT kind||
|BOOLEAN|BOOLEAN|
|TINYINT, SMALLINT, INTEGER, BIGINT|TINYINT, SMALLINT, INT, BIGINT|
|FLOAT, DOUBLE|FLOAT, DOUBLE|
|DECIMAL(p, s)|DECIMAL|
|CHAR, VARCHAR|STRING|
|BINARY, VARBINARY|BYTES|
|DATE, TIME, UUID|DATE, TIME, UUID|
|TIMESTAMP(p), TIMESTAMP_LTZ(p)|TIMESTAMP, TIMESTAMP_LTZ, or the nanosecond
kinds when the value has digits below a microsecond|
Rejected: INTERVAL, TIMESTAMP WITH TIME ZONE, RAW, BITMAP, SYMBOL, DESCRIPTOR,
MULTISET, and the constructed types for now.
h3. Semantics
{code:sql}
CAST(CAST(1 AS BIGINT) AS VARIANT) -- 1, stored as BIGINT
CAST(42 AS VARIANT) -- 42, stored as INT
CAST('\{"a": 1}' AS VARIANT) -- the string '\{"a": 1}', not
an object
CAST(CAST(NULL AS INT) AS VARIANT) -- SQL NULL, not a variant null
CAST(CAST('NaN' AS DOUBLE) AS VARIANT) -- NaN, stored as DOUBLE
CAST(INTERVAL '2' DAY AS VARIANT) -- fails at validation
CAST(ARRAY[1, 2] AS VARIANT) -- fails at validation, follow-up
{code}
* An integer keeps the width of its SQL type. \{{PARSE_JSON('1')}} picks the
smallest kind because JSON text has no width, but a cast keeps the declared
type.
* A character string is wrapped as a STRING and never parsed. \{{PARSE_JSON}}
remains the way to parse JSON text.
* NaN and infinity are valid VARIANT values, since the binary format stores
IEEE numbers. \{{PARSE_JSON}} rejects them only because JSON has no literal for
them.
* Every supported type round-trips: \{{CAST(CAST(x AS VARIANT) AS T) = x}}.
* The cast never fails at runtime, so \{{TRY_CAST}} behaves like \{{CAST}}.
h3. Prerequisite change
Casting a VARIANT to FLOAT or DOUBLE rejected every non-finite result as an
overflow. A VARIANT holding NaN, for example from the Avro converter, could
therefore not be read back. The read now rejects only a finite value that does
not fit, such as 1e40 to FLOAT, and keeps a stored NaN or infinity.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
