feat(expr): add jsonb_populate_record(set) function

[wangrunji0408]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

This PR adds the jsonb_populate_record function and its set-returning form:

jsonb_populate_record ( base anyelement, from_json jsonb ) → anyelement
jsonb_populate_recordset ( base anyelement, from_json jsonb ) → setof anyelement

To support jsonb_populate_recordset, some improvements have been made to the #[function] macro, in order to support Option inputs and the Context argument for table functions.

Remaining issue

Calling the scalar function jsonb_populate_record in the FROM clause triggers a panic in the planner.

dev=> select * from jsonb_populate_record(
    null::struct<a int, b int>,
    '{"a": 1, "b": 2}' 
);
ERROR:  Panicked when processing: insert at index 1 exceeds fixbitset size 1

Although calling it in the SELECT list is normal, it seems that the above form is more commonly used.

dev=> select jsonb_populate_record(
    null::struct<a int, b int>,
    '{"a": 1, "b": 2}'
);
 jsonb_populate_record 
-----------------------
 (1,2)

---

This PR also implements a similar function jsonb_to_record(set) on the backend. However, they are not enabled due to lack of frontend support.

These functions are designed to be called in the FROM clause, with type defined by an AS clause.

postgres=# select * from jsonb_to_record('{"a":1,"b":"foo","c":"bar"}')
    as x(a int, b text, d text);
 a |  b  | d 
---+-----+---
 1 | foo | 

It seems that we don't support such usage in both parser and frontend.

-- risingwave
dev=> select * from jsonb_to_record('{"a":1,"b":"foo","c":"bar"}')
    as x(a int, b text, d text);
ERROR:  QueryError

Caused by:
  sql parser error: Expected ), found: int at line:2, column:15
Near "
    as x(a"

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
• My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
• All checks passed in ./risedev check (or alias, ./risedev c)
• My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

• My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

Add 2 functions:

jsonb_populate_record ( base anyelement, from_json jsonb ) → anyelement
jsonb_populate_recordset ( base anyelement, from_json jsonb ) → setof anyelement

See postgres documents for detailed descriptions.

Note that we have different usages compared to pg.

-- risingwave
select (jsonb_populate_record(
    null::struct<a int, b text[], c struct<d int, e text>>,
    '{"a": 1, "b": ["2", "a b"], "c": {"d": 4, "e": "a b c"}, "x": "foo"}'
)).*;
----
1 {2,"a b"} (4,"a b c")

select * from jsonb_populate_recordset(
    null::struct<a int, b int>,
    '[{"a":1,"b":2}, {"a":3,"b":4}]'::jsonb
);
----
1 2
3 4

In Postgres, users need to define a composite type using CREATE TYPE before calling these functions.
In RisingWave, users should use inline struct type instead.

[fuyufjh]

Regarding of the remaining issue: I think it's okay to keep the status quo i.e. do not support this usage now. I can't see any real benefit or use cases. What do you think? cc. @xiangjinwu

---

Did some quick experiments on PG and here is the result:

test=# select * from md5('hello');
               md5
----------------------------------
 5d41402abc4b2a76b9719d911017c592
(1 row) 

-- this is equivalent to
test=# select md5('hello');
               md5
----------------------------------
 5d41402abc4b2a76b9719d911017c592
(1 row)

While operators can't

test=# select * from 1+1;
ERROR:  syntax error at or near "1"

[wangrunji0408]

select * from a_scalar_function_that_returns_record() seems to be useful.
But we support an equivalent way: select (a_scalar_function_that_returns_record()).*;
So I also agree to leave this issue to the future.

[xiangjinwu]

Regarding of the remaining issue: I think it's okay to keep the status quo i.e. do not support this usage now. I can't see any real benefit or use cases. What do you think? cc. @xiangjinwu

Its major usage is to convert from a loosely typed jsonb to strongly typed struct, as we have discussed on the map type issue #13387 (comment) . According to PostgreSQL:

typical use would be to reference a json or jsonb column laterally from another table in the query's FROM clause. Writing json_populate_record in the FROM clause is good practice, since all of the extracted columns are available for use without duplicate function calls.

I will give an example on how it could help if we decided to use jsonb for proto map<string, Location>:

select (lookup(col1, 'SFO')).latitude from t; -- if we had a `lookup` function on native `map` or `struct[]`
select col1 -> 'SFO' ->> 'latitude' from t; -- if we used `jsonb`

select
  latitude
from
  t,
  json_populate_record(null::struct<latitude float8, longitude float8>, col1 -> 'SFO');
-- if we used `jsonb`, and `json_populate_record` helps use convert inner `jsonb` back to strongly typed struct `Location`.

It is true that we can do (...).* in select-list as well. But the lateral-from syntax is more convenient if we want to process further immediately, for example select tand(latitude) * sind(longitude) from ...

That being said, it is always possible to implement a functionality that can only be used in limited cases, and then improved in future iterations.

[github-actions]

This PR has been open for 60 days with no activity. Could you please update the status? Feel free to ping a reviewer if you are waiting for review.

[gitguardian]

⚠️ GitGuardian has uncovered 1 secret following the scan of your pull request.

Please consider investigating the findings and remediating the incidents. Failure to do so may lead to compromising the associated services or software components.

🔎 Detected hardcoded secret in your pull request

GitGuardian id	GitGuardian status	Secret	Commit	Filename
9425213	Triggered	Generic Password	505790f	ci/scripts/regress-test.sh	View secret

🛠 Guidelines to remediate hardcoded secrets

1. Understand the implications of revoking this secret by investigating where it is used in your code.
2. Replace and store your secret safely. Learn here the best practices.
3. Revoke and rotate this secret.
4. If possible, rewrite git history. Rewriting git history is not a trivial act. You might completely break other contributing developers' workflow and you risk accidentally deleting legitimate data.

To avoid such incidents in the future consider

• following these best practices for managing and storing secrets including API keys and other credentials
• install secret detection on pre-commit to catch secret before it leaves your machine and ease remediation.

---

<sup>🦉 GitGuardian detects secrets in your source code to help developers and security teams secure the modern development process. You are seeing this because you or someone else with access to this repository has authorized GitGuardian to scan your pull request.

Our GitHub checks need improvements? Share your feedbacks!</sup>

[github-actions]

license-eye has totally checked 5010 files.

Valid	Invalid	Ignored	Fixed
2150	1	2859	0

Click to see the invalid file list

• src/expr/impl/src/scalar/jsonb_record.rs

[TennyZhuang]

LGTM

[xiangjinwu]

Just wondering the following:

• How is frontend agg name jsonb_populate_recordset mapped to the proto enum, by the macro?
• How does frontend infer the return type is same as first argument for jsonb_populate_record?

[wangrunji0408]

How is frontend agg name jsonb_populate_recordset mapped to the proto enum, by the macro?

by the PbType::from_str_name function generated by prost:

risingwave/src/prost/src/lib.rs

Lines 171 to 177 in 86ac793

	impl FromStr for crate::expr::table_function::PbType {
	type Err = ();
	fn from_str(s: &str) -> Result<Self, Self::Err> {
	Self::from_str_name(&s.to_uppercase()).ok_or(())
	}
	}

How does frontend infer the return type is same as first argument for jsonb_populate_record?

The #[function] macro generates a type inference function for this function.

risingwave/src/expr/macro/src/gen.rs

Lines 113 to 117 in 86ac793

	} else if self.ret == "struct" {
	if let Some(i) = self.args.iter().position(|t| t == "struct") {
	// infer as the type of "struct" argument
	return Ok(quote! { |args| Ok(args[#i].clone()) });
	}

and frontend will call this function at the end of infer_type.

risingwave/src/frontend/src/expr/type_inference/func.rs

Line 72 in 86ac793

let return_type = (sig.type_infer)(&input_types)?;