xiangfu0 opened a new pull request, #19263:
URL: https://github.com/apache/pinot/pull/19263
## Depends on
- #19247
This is intentionally a draft stacked follow-up to keep PostgreSQL parser
compatibility separate from the core BYTES_ARRAY feature. Until #19247 merges,
GitHub's diff against master also includes that parent PR. After #19247 lands,
this branch will be rebased so only the PostgreSQL parser commit remains.
## Summary
- Accept PostgreSQL hex bytea constants in both '\x0102'::bytea and
CAST('\x0102' AS BYTEA) forms.
- Normalize supported bytea constants to Pinot's existing binary-literal
representation before either the single-stage or multi-stage engine plans the
query.
- Preserve the existing SQL-standard X'0102' form.
- Cover scalar literals, BYTES array construction, ingested multi-value
BYTES predicates, planner type inference, and runtime execution.
## Usage
PostgreSQL infix-cast form:
~~~sql
SELECT '\x0102'::bytea;
~~~
Standard CAST spelling:
~~~sql
SELECT CAST('\x0102' AS BYTEA);
~~~
Construct and query a BYTES array:
~~~sql
SELECT ARRAY['\x00'::bytea, '\x0102'::bytea, '\xFF'::bytea];
SELECT id, byte_values
FROM events
WHERE ARRAYS_OVERLAP(
byte_values,
ARRAY['\x0102'::bytea, '\xCAFE'::bytea]
);
~~~
Result values remain hex strings, and an array result reports BYTES_ARRAY
metadata.
## Compatibility scope
This change supports quoted PostgreSQL hex-format bytea constants. Values
must start with \x; upper- and lowercase hex are accepted, as is whitespace
between complete byte pairs.
It does not add a general expr::type cast operator, dynamic STRING-to-BYTES
conversion, or PostgreSQL's historical octal escape-format bytea input.
Normalizing only quoted constants keeps both query engines aligned and avoids
per-row decoding.
No Thrift or protobuf schema changes are introduced here. Supported
constants normalize to the same scalar binary literals used by X'...' syntax.
## Validation
- Parser/request/Thrift tests: 64/64 passed.
- Single-stage literal-only broker tests: 9/9 passed.
- Multi-stage bytea type-inference test: 1/1 passed.
- BinaryTypes runtime/H2 queries: 4/4 passed across both optimizer variants.
- BytesMvTypeTest: 14/14 passed for real Avro array<bytes> ingestion and
both query engines.
- Full integration-test reactor: 63/63 modules and 956 goals passed under
JDK 25.
- spotless:apply, checkstyle:check, license:format, and license:check passed
for all five affected modules.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
