Allocate dump() indent buffer only for pretty-print

[22elix3r]

Summary

Every dump() constructed a 512-byte indent_string even for compact output (indent = -1), where the buffer is never read. That per-call heap allocation showed up as ~17% overhead on dump-many-small-values.

Related Issue

Fixes #5413

Changes Made

• Default-construct indent_string empty
• Grow it on demand in ensure_indent() using indent_char (the previous doubling path filled with ' ', which was wrong for a custom indent character)
• Tests for compact dump, dump(2), and dump(1, '\\t')

The make_shared output-adapter allocation from the issue is left for a follow-up; it is a larger adapter API change.

Testing

Commands executed:

• python3 tools/amalgamate/amalgamate.py -c tools/amalgamate/config_json.json -s .
• cmake -S . -B build -DJSON_BuildTests=ON -DCMAKE_BUILD_TYPE=Debug
• cmake --build build --target test-serialization_cpp11
• ./tests/test-serialization_cpp11 --no-skip -tce='*downloaded*'
• ./tests/test-serialization_cpp11 --no-skip -tce='*downloaded*' -tc='*5413*'

Results:

• full serialization: 8 passed, 111 assertions
• issue dump() performs avoidable per-call heap allocations (eager 512-byte indent_string + shared_ptr output adapter) — ~17% overhead on the dump-many-small-values pattern #5413 case: 1 passed, 4 assertions

Notes

Pretty-print still allocates a 512-byte (or larger) indent buffer the first time a nested indent is needed. Compact dump allocates none.

• The changes are described in detail, both the what and why.
• If applicable, an existing issue is referenced.
• The Code coverage remained at 100%. A test case for every new line of code.
• If applicable, the documentation is updated.
• The source code is amalgamated by running make amalgamate.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[gregmarr]

FYI, there is an open PR that removes this buffer entirely.

[22elix3r]

Thanks — that is #5285 (and the related overflow work in #5186). This PR is a smaller, independent change: skip the 512-byte indent_string allocation on compact dump(). If #5285 lands first this becomes redundant and can be closed; until then it still helps compact dump() on current develop.

[github-actions]

🔴 Amalgamation check failed! 🔴

The source code has not been amalgamated and/or formatted correctly.

📎 A ready-to-apply patch is attached to the failed workflow run as the amalgamation-patch artifact. Download it, then apply it locally from the repository root with:

git apply amalgamation.patch

This does not require installing astyle yourself.

[nlohmann]

@22elix3r CI is red: the build fails to compile with serializer.hpp:992:27: error: no member named 'assign' in 'alt_string' (the custom-string-type test), and Fuzzing, CodeQL, and the amalgamation check are also failing.

[nlohmann]

Thanks for this — and good catch on the indent-buffer regrowth using a hardcoded space fill instead of the configured indent_char past 512 bytes.

This turns out to be superseded by already-in-progress maintainer work: #5285 (part of an active, currently-green stack) removes the indent_string allocation entirely rather than just deferring it, replacing it with a stack-resident write buffer — so the bug this PR fixes can't recur there, and we don't need a separate fix for it. We closed our own equivalent PR (#5483) for the same reason and are moving forward with #5285 instead. Appreciate you tracking down the fill-character issue regardless.

— closed by Claude Code on behalf of @nlohmann