[nlohmann]

Part of the stack for the zero-copy view (#5295). This change speeds up json::parse on its own.

Summary

get_codepoint() read the four hex digits of a \uXXXX escape with four calls to get(). Each call went through a chain of range comparisons.

For contiguous input, the four digits are now decoded with one table lookup per byte (after yyjson's read_hex_u16). The table maps a byte to its value, or to 0xFF for anything that is not a hex digit, so an invalid digit shows in the OR of the four values. The lexer then skips the four bytes and updates the position counters as four get() calls would.

If a digit is invalid, or fewer than four bytes are left, nothing changes and the existing loop runs. Error messages and positions are therefore exactly as before.

Tests

• The contiguous and the streaming path give the same value, or the same exception message, for:
  • valid escapes and surrogate pairs;
  • truncated input at every distance from the end;
  • an invalid digit at each of the four positions;
  • 3,000 seeded random escapes.
• The existing lexer, parser, and Unicode tests pass unchanged.

Benchmarks

json::parse in a separate process, best of 5, Apple M1 Max:

file	change
twitterescaped.json (every non-ASCII character as \u)	−13.6%
citm_catalog, twitter, mesh.pretty, poet, github_events, update-center	within ±0.3%

This PR originally also compared literals (true/false/null) with memcmp. That made citm_catalog, twitter, and mesh.pretty 2–3% slower, so it was dropped.

Public API

No change.

---

Written by Claude Code.

🤖 Generated with Claude Code

[nlohmann]

The zero-copy view stack (#5295) was rebased onto develop and regrouped from 20 PRs into 10, one squashed commit each. The changes of this PR are now in #5738; its review comments stay here for reference.

---

Written by Claude Code.