Pretty formats: optional non-breaking-space leading padding

[ashrithb]

Changelog category (leave one):

• Improvement

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

Add a session setting output_format_pretty_use_nbsp_for_leading_padding (default false). When enabled, the leading row-number padding and the indent before grid borders in Pretty, PrettyCompact, and PrettySpace formats are rendered with the UTF-8 encoding of U+00A0 NO-BREAK SPACE instead of an ASCII space. The visual width is identical in monospace, but the padding survives copy-paste through tools that compress or trim runs of regular spaces. Only takes effect when output_format_pretty_grid_charset = UTF-8; suppressed under ASCII charset by design.

Documentation entry for user-facing changes

• Documentation is written (mandatory for new features)

---

Motivation

When users copy Pretty/PrettyCompact output through tools that
compress runs of ASCII spaces (Slack, certain markdown renderers, some
terminal pasteboards), the leading row-number padding and the indent
before grid borders get mangled — the columns no longer line up.

Fix

Add a session setting output_format_pretty_use_nbsp_for_leading_padding
(default false). When the setting is enabled AND
output_format_pretty_grid_charset is UTF-8, the leading whitespace
left_blank string in PrettyBlockOutputFormat is rendered with the
UTF-8 encoding of U+00A0 NO-BREAK SPACE (0xC2 0xA0) instead of an
ASCII space (0x20). The visual width is identical in monospace, but
NBSP survives the space-collapsing tools because they do not normalise
non-ASCII whitespace.

The setting is suppressed under ASCII charset by design — the grid is
rendered with ASCII characters, so NBSP would look out of place.

Default false so existing reference output stays bit-identical and no
existing .reference files need to be regenerated. The
SettingsChangesHistory.cpp entry is in the 26.5 block so the new
setting is visible to settings-history consumers.

Test

Adds tests/queries/0_stateless/04123_pretty_nbsp_leading_padding.sh
which uses od -An -v -tx1 to inspect the actual leading bytes of the
output and asserts:

• default: first byte of the column-header line is 0x20 (ASCII space)
• setting on, UTF-8 charset: first two bytes are 0xC2 0xA0 (NBSP)
• setting on, ASCII charset: first byte stays 0x20 (NBSP suppressed)

Closes #95122

[alexey-milovidov]

The second screenshot shows, that it would be better to use it for all type of padding.
Let's rename the setting, and change the behavior accordingly.

[clickhouse-gh]

Workflow [PR], commit [f0f71d6]

Summary: ❌

job_name	test_name	status	info	comment
Stress test (amd_msan)		FAIL
	Logical error: Port is already connected, (header: [A]) (STID: 0996-3f42)	FAIL	cidb
AST fuzzer (amd_debug)		FAIL
	Logical error: Variant A (B) has size C, but expected D (STID: 3068-3371)	FAIL	cidb

---

AI Review

Summary

This PR adds a new setting, output_format_pretty_use_nbsp_for_leading_padding, and wires it through Pretty format rendering plus a stateless test. The implementation change looks consistent, but there is one test-quality issue that should be fixed before merge.

Findings

⚠️ Majors

• [tests/queries/0_stateless/04123_pretty_nbsp_leading_padding.sh:17] The test comment says it validates the "second line" (column header row), but the command uses sed -n '1p'. This mismatch makes the test fragile and can miss regressions in row-prefix padding code paths. Update the sampled line (or comment) so the test verifies the intended row and covers both leading-padding paths.

ClickHouse Rules

Item	Status	Notes
Deletion logging	➖
Serialization versioning	➖
Core-area scrutiny	✅
No test removal	✅
Experimental gate	➖
No magic constants	✅
Backward compatibility	✅
SettingsChangesHistory.cpp	✅
PR metadata quality	✅
Safe rollout	✅
Compilation time	✅
No large/binary files	✅

Final Verdict

• Status: ⚠️ Request changes
• Minimum required actions: align the test’s sampled line with its stated intent (or adjust the intent/comment) and ensure the assertion covers the intended Pretty leading-padding path.

[clickhouse-gh]

sed -n '1p' contradicts the comment above ("second line") and only samples the very first rendered line. That can miss regressions in the data-row prefix path (row_num_string padding), which is generated separately.

Please either switch to 2p (to actually test the column-header line) or update the assertion to explicitly target a data row so both leading-padding paths are covered.

[clickhouse-gh]

LLVM Coverage Report

Metric	Baseline	Current	Δ
Lines	84.00%	84.00%	+0.00%
Functions	91.10%	91.10%	+0.00%
Branches	76.50%	76.50%	+0.00%

Changed lines: 97.56% (40/41) · Uncovered code

Full report · Diff report