Re: remaining sql/json patches
jian he <jian.universality@gmail.com>
Commits
GET /api/v1/messages/:b64id/commits
the thread's linked commits as JSON, with link sources.
API reference →
-
SQL/JSON: Various improvements to SQL/JSON query function docs
- ce416fadb4b6 17.0 landed
- 42de72fa7b80 18.0 landed
-
SQL/JSON: Fix some obsolete comments.
- 290a6d800d90 17.0 landed
- 7768b6569de9 16.4 landed
- 3a8a1f3254b2 18.0 landed
-
SQL/JSON: Fix issues with DEFAULT .. ON ERROR / EMPTY
- c0fc0751862d 17.0 landed
-
JSON_TABLE: Add support for NESTED paths and columns
- bb766cde63b4 17.0 landed
-
Fix JsonExpr deparsing to emit QUOTES and WRAPPER correctly
- f6a2529920cf 17.0 landed
-
Fix typo introduced in 6185c9737
- 2f6e78b0619a 17.0 landed
-
Add basic JSON_TABLE() functionality
- de3600452b61 17.0 landed
-
Avoid splitting errmsg string to span multiple lines
- 085e759e9da7 17.0 landed
-
Add SQL/JSON query functions
- 6185c9737cf4 17.0 landed
-
Implement various jsonpath methods
- 66ea94e8e606 17.0 cited
-
Add soft error handling to some expression nodes
- aaaf9449ec6b 17.0 landed
- 7fbc75b26ed8 17.0 landed
-
Adjust populate_record_field() to handle errors softly
- 1edb3b491bee 17.0 landed
-
Refactor code used by jsonpath executor to fetch variables
- faa2b953ba3b 17.0 landed
-
Test EXPLAIN (FORMAT JSON) ... XMLTABLE
- 752533d40fd5 17.0 landed
-
Simplify productions for FORMAT JSON [ ENCODING name ]
- d3fe6e90bab5 17.0 landed
-
Add trailing commas to enum definitions
- 611806cd726f 17.0 cited
-
doc: add missing <returnvalue> and whitespace
- e055b6be7ebb 17.0 landed
-
Add more SQL/JSON constructor functions
- 03734a7fed7d 17.0 landed
-
Rename a nonterminal used in SQL/JSON grammar
- 254ac5a7c31f 17.0 landed
-
Some refactoring to export json(b) conversion functions
- b22391a2ff7b 17.0 landed
-
Don't include CaseTestExpr in JsonValueExpr.formatted_expr
- 66a9003e2e3e 16.0 landed
- b6e1157e7d33 17.0 landed
-
Code review for commit b6e1157e7d
- 7c7412cae3ea 17.0 landed
-
Pass constructName to transformJsonValueExpr()
- 7825a1b01e40 16.0 landed
- 785480c9533d 17.0 landed
-
Unify JSON categorize type API and export for external use
- 3c152a27b063 17.0 landed
-
Make some indentation in gram.y consistent
- 5edf438eeb00 17.0 landed
- 01f1f789df56 16.0 landed
-
Allow most keywords to be used as column labels without requiring AS.
- 06a7c3154f5b 14.0 cited
-
Reduce size of backend scanner's tables.
- 7f380c59f800 13.0 cited
-
Use perfect hashing, instead of binary search, for keyword lookup.
- c64d0cd5ce24 12.0 cited
On Mon, Mar 11, 2024 at 11:30 AM jian he <jian.universality@gmail.com> wrote:
>
> On Sun, Mar 10, 2024 at 10:57 PM jian he <jian.universality@gmail.com> wrote:
> >
> > one more issue.
>
> Hi
> one more documentation issue.
> after applied V42, 0001 to 0003,
> there are 11 appearance of `FORMAT JSON` in functions-json.html
> still not a single place explained what it is for.
>
> json_query ( context_item, path_expression [ PASSING { value AS
> varname } [, ...]] [ RETURNING data_type [ FORMAT JSON [ ENCODING UTF8
> ] ] ] [ { WITHOUT | WITH { CONDITIONAL | [UNCONDITIONAL] } } [ ARRAY ]
> WRAPPER ] [ { KEEP | OMIT } QUOTES [ ON SCALAR STRING ] ] [ { ERROR |
> NULL | EMPTY { [ ARRAY ] | OBJECT } | DEFAULT expression } ON EMPTY ]
> [ { ERROR | NULL | EMPTY { [ ARRAY ] | OBJECT } | DEFAULT expression }
> ON ERROR ])
>
> FORMAT JSON seems just a syntax sugar or for compatibility in json_query.
> but it returns an error when the returning type category is not
> TYPCATEGORY_STRING.
>
> for example, even the following will return an error.
> `
> CREATE TYPE regtest_comptype AS (b text);
> SELECT JSON_QUERY(jsonb '{"a":{"b":"c"}}', '$.a' RETURNING
> regtest_comptype format json);
> `
>
> seems only types in[0] will not generate an error, when specifying
> FORMAT JSON in JSON_QUERY.
>
> so it actually does something, not a syntax sugar?
>
SELECT * FROM JSON_TABLE(jsonb'[{"aaa": 123}]', 'lax $[*]' COLUMNS
(js2 text format json PATH '$' omit quotes));
SELECT * FROM JSON_TABLE(jsonb'[{"aaa": 123}]', 'lax $[*]' COLUMNS
(js2 text format json PATH '$' keep quotes));
SELECT * FROM JSON_TABLE(jsonb'[{"aaa": 123}]', 'lax $[*]' COLUMNS
(js2 text PATH '$' keep quotes)); -- JSON_QUERY_OP
SELECT * FROM JSON_TABLE(jsonb'[{"aaa": 123}]', 'lax $[*]' COLUMNS
(js2 text PATH '$' omit quotes)); -- JSON_QUERY_OP
SELECT * FROM JSON_TABLE(jsonb'[{"aaa": 123}]', 'lax $[*]' COLUMNS
(js2 text PATH '$')); -- JSON_VALUE_OP
SELECT * FROM JSON_TABLE(jsonb'[{"aaa": 123}]', 'lax $[*]' COLUMNS
(js2 json PATH '$')); -- JSON_QUERY_OP
comparing these queries, I think 'FORMAT JSON' main usage is in json_table.
CREATE TYPE regtest_comptype AS (b text);
SELECT JSON_QUERY(jsonb '{"a":{"b":"c"}}', '$.a' RETURNING
regtest_comptype format json);
ERROR: cannot use JSON format with non-string output types
LINE 1: ..."a":{"b":"c"}}', '$.a' RETURNING regtest_comptype format jso...
^
the error message is not good, but that's a minor issue. we can pursue it later.
-----------------------------------------------------------------------------------------
SELECT JSON_QUERY(jsonb 'true', '$' RETURNING int KEEP QUOTES );
SELECT JSON_QUERY(jsonb 'true', '$' RETURNING int omit QUOTES );
SELECT JSON_VALUE(jsonb 'true', '$' RETURNING int);
the third query returns integer 1, not sure this is the desired behavior.
it obviously has an implication for json_table.
-----------------------------------------------------------------------------------------
in jsonb_get_element, we have something like:
if (jbvp->type == jbvBinary)
{
container = jbvp->val.binary.data;
have_object = JsonContainerIsObject(container);
have_array = JsonContainerIsArray(container);
Assert(!JsonContainerIsScalar(container));
}
+ res = JsonValueListHead(&found);
+ if (res->type == jbvBinary && JsonContainerIsScalar(res->val.binary.data))
+ JsonbExtractScalar(res->val.binary.data, res);
So in JsonPathValue, the above (res->type == jbvBinary) is unreachable?
also see the comment in jbvBinary.
maybe we can just simply do:
if (res->type == jbvBinary)
Assert(!JsonContainerIsScalar(res->val.binary.data));
-----------------------------------------------------------------------------------------
+<synopsis>
+JSON_TABLE (
+ <replaceable>context_item</replaceable>,
<replaceable>path_expression</replaceable> <optional> AS
<replaceable>json_path_name</replaceable> </optional> <optional>
PASSING { <replaceable>value</replaceable> AS
<replaceable>varname</replaceable> } <optional>, ...</optional>
</optional>
+ COLUMNS ( <replaceable
class="parameter">json_table_column</replaceable> <optional>,
...</optional> )
+ <optional> { <literal>ERROR</literal> | <literal>EMPTY</literal> }
<literal>ON ERROR</literal> </optional>
+ <optional>
+ PLAN ( <replaceable class="parameter">json_table_plan</replaceable> ) |
+ PLAN DEFAULT ( { INNER | OUTER } <optional> , { CROSS | UNION } </optional>
+ | { CROSS | UNION } <optional> , { INNER | OUTER }
</optional> )
+ </optional>
+)
based on the synopsis
the following query should not be allowed?
SELECT *FROM (VALUES ('"11"'), ('"err"')) vals(js)
LEFT OUTER JOIN JSON_TABLE(vals.js::jsonb, '$' COLUMNS (a int PATH
'$') default '11' ON ERROR) jt ON true;
aslo the synopsis need to reflect case like:
SELECT *FROM (VALUES ('"11"'), ('"err"')) vals(js)
LEFT OUTER JOIN JSON_TABLE(vals.js::jsonb, '$' COLUMNS (a int PATH
'$') NULL ON ERROR) jt ON true;