avoid unnecessary interface conversions for JSONWrappers

[jycor]

No description provided.

[github-actions]

	Main	PR
covering_index_scan_postgres	1941.92/s	1939.09/s	-0.2%
groupby_scan_postgres	131.93/s	131.64/s	-0.3%
index_join_postgres	682.59/s	684.54/s	+0.2%
index_join_scan_postgres	871.88/s	873.81/s	+0.2%
index_scan_postgres	25.52/s	25.51/s	-0.1%
oltp_delete_insert_postgres	881.72/s	854.31/s	-3.2%
oltp_insert	754.01/s	750.16/s	-0.6%
oltp_point_select	3088.25/s	3082.02/s	-0.3%
oltp_read_only	3102.69/s	3087.88/s	-0.5%
oltp_read_write	2377.10/s	2360.05/s	-0.8%
oltp_update_index	778.27/s	783.47/s	+0.6%
oltp_update_non_index	828.98/s	826.39/s	-0.4%
oltp_write_only	1828.11/s	1810.96/s	-1.0%
select_random_points	1958.88/s	1969.71/s	+0.5%
select_random_ranges	1131.74/s	1126.48/s	-0.5%
table_scan_postgres	24.40/s	24.27/s	-0.6%
types_delete_insert_postgres	825.37/s	818.64/s	-0.9%
types_table_scan_postgres	8.67/s	${\color{lightgreen}10.57/s}$	${\color{lightgreen}+21.9\%}$

[github-actions]

	Main	PR
Total	42090	42090
Successful	18275	18276
Failures	23815	23814
Partial Successes1	5335	5335

	Main	PR
Successful	43.4189%	43.4212%
Failures	56.5811%	56.5788%

${\color{lightgreen}Progressions (1)}$

subselect

QUERY: select count(*) from tenk1 t
where (exists(select 1 from tenk1 k where k.unique1 = t.unique2) or ten < 0);

Footnotes

1. These are tests that we're marking as Successful, however they do not match the expected output in some way. This is due to small differences, such as different wording on the error messages, or the column names being incorrect while the data itself is correct. ↩

[itoqa]

Ito Diff Report — a8665b3 → 2bc960f: 10 test cases ran, 0 regressions ❌, 0 new failures ❌, 0 still failing ❌, 0 fixed ✅, 10 passing ✅, 0 additional findings ⚠️.

Diff Summary

This diff run surfaced

Coverage focused on JSON output behavior across text and binary responses, nested and mixed-value query contexts, and upgraded legacy data to confirm valid, consistent serialization in real read paths. It also covered edge conditions where isolated bad JSON retrievals occur among valid rows, showing explicit deterministic failures without broader result corruption, including repeat and concurrent execution patterns.

Safe to merge — the exercised paths remained stable across normal behavior, compatibility scenarios, and failure-handling cases, with no test regressions observed. Given the broad pass across the JSON output and serialization surfaces touched by this PR, current merge risk is low.

Tests run by Ito

Result	State	Severity	Type	Description
✅	Passing	—	Document	json_out/jsonb_out correctly route sql.JSONWrapper values through jsonWrapperToFormattedString and return normalized JSON text output.
✅	Passing	—	Document	Nested JSONB values in records, arrays, path extraction, concatenation, and COALESCE serialized as valid text JSON without io output type errors.
✅	Passing	—	Format	The JSON text wire-format test passed and returned compact payload bytes that match the updated expectations for object and array rows.
✅	Passing	—	Format	The same JSON query returned different raw bytes between text and binary wire formats, and repository code plus wire assertions show this representation difference is intentional rather than an application defect.
✅	Passing	—	Lazy	Wire-protocol SELECT on JSON columns returned compact text payloads for both object and nested-array rows, matching JSONBytes GetBytes(ctx) output.
✅	Passing	—	Lazy	json_out propagates JSONBytes GetBytes errors directly and deterministically, with no opaque fallback behavior or result-shape drift across repeated and concurrent execution.
✅	Passing	—	Legacy	AnyWrapper-based legacy JSON values, including large-object fixtures, were read and serialized successfully through json_out and jsonb_out without compatibility errors.
✅	Passing	—	Legacy	Code-backed verification confirms unsupported AnyWrapper-unwrapped payloads fail with explicit "unexpected type" errors while supported JSONDocument and string payloads still serialize correctly in the same run.
✅	Passing	—	Legacy	Round-trip, reserialization, and live SQL checks all passed, with equivalent legacy and modern JSON rows producing identical text and JSONB output.
✅	Passing	—	Output	Text-format and binary-format reads returned valid JSON for equivalent values, with only expected representation-specific spacing differences.

Tip

Reply with @itoqa to send us feedback on this test run.

[fulghum]

Good optimization to avoid the extra JSON marshalling where we can.