test(jsonpointer): expand is_relative_pointer coverage for magnitude, digits, and the tail

[vtushar06]

I added 77 tests to test/jsonpointer/jsonpointer_is_relative_pointer_test.cc. Core already passes all of them, so this is coverage, not a fix.

The existing file has 54 tests and covers the integer prefix well - 0, 120, 1234567890, the leading-zero rejections, 0#, and a few token characters. Reading it against draft-handrews-relative-json-pointer-01 section 3 and the RFC 6901 section 3 tail, the branches with no test at all were integer magnitude, the non-ASCII digit families beyond the single Bengali case, and most of the unescaped range.

What is new

• integer magnitude (9) - the ABNF puts no upper bound on non-negative-integer, so 4294967296, 9223372036854775808#, 18446744073709551616 and a 100-digit prefix are all valid. The old file stopped at 1234567890. This is the branch where a fixed-width integer parse shows up: JsonPointer.Net 7.0.1 throws OverflowException at exactly 4294967296 because it parses the prefix with uint.Parse.
• non-ASCII digit families (10) - the file had one Bengali digit. I added Arabic-Indic, fullwidth, superscript, circled, mathematical-bold (astral), Ethiopic, Devanagari, and the two positional variants. I enumerated every codepoint carrying a Unicode digit property (1862 of them) and Core rejects all 1862; these ten are the per-family representatives.
• index manipulation (9) - 0+1, 1-2, 0+1/foo, 2+10#. There is no index-manipulation production in handrews-01; it exists only in the later draft-bhutton-relative-json-pointer-00. These pin Core to the draft the validation spec actually references, which matters because implementations are split: JsonPointer.Net 7.0.1, @hyperjump/json-schema-formats and opis/json-schema 2.6.0 all accept 0+1.
• octothorpe placement (6) - the "#" alternative is the whole remainder, so 1#/foo/bar, 0#/ and 0##/foo are invalid while 120# is valid, and # is an ordinary character inside a token (0/a#b).
• integer not followed by / or # (4) - 2foo/bar, 0 /foo, %30/foo.
• reference-token alphabet (9) - unescaped = %x00-2E / %x30-7D / %x7F-10FFFF, so DEL (%x7F), } (%x7D), . (%x2E), astral characters, U+10FFFF, control characters, and a raw newline or tab inside a token are all valid. Only / and ~ are excluded.
• embedded NUL (3) - NUL is inside %x00-2E so 0/foo<NUL>bar is valid. I also added 0/foo<NUL>~, invalid because of the dangling ~; a validator treating the input as a C string would truncate at the NUL and wrongly accept it.
• escapes through the wrapper (8), terminators (9), leading zero (7), empty reference tokens (3).

How I checked the expected values

I wrote the expected verdict for each input from the ABNF, then built the sourcemeta_core_jsonpointer_unit target with the modified file: 785 passed, 0 failed. clang-format is idempotent on the result.

The three NUL cases use an explicit-length std::string_view and carry a static_assert(sizeof(value) - 1 == N). That is load-bearing rather than decorative: a \xNN escape is greedy in C++, so "0/foo\x00bar" is the single escape \x00b (0x0B) followed by ar - eight bytes, not the nine intended. The first version of these tests had exactly that defect and still passed, because the shortened literal happened to give the same verdict. The static_assert turns it into a compile error.

RFC references

• draft-handrews-relative-json-pointer-01 section 3 (Syntax): https://datatracker.ietf.org/doc/html/draft-handrews-relative-json-pointer-01#section-3
• RFC 6901 section 3 (reference-token, unescaped, escaped): https://www.rfc-editor.org/rfc/rfc6901#section-3

[Copilot]

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

[augmentcode]

🤖 Augment PR Summary

Summary: This PR expands unit-test coverage for sourcemeta::core::is_relative_pointer (Relative JSON Pointer validation).

Changes:

• Adds 77 new test cases in test/jsonpointer/jsonpointer_is_relative_pointer_test.cc and includes <string_view> for explicit-length inputs.
• Covers very large (unbounded) integer prefixes, including values beyond 32/64-bit ranges.
• Adds representative non-ASCII digit-family prefixes to confirm only ASCII digits are accepted in the integer prefix.
• Adds negative tests for index-manipulation syntax (e.g. 0+1) to pin behavior to draft-handrews-relative-json-pointer-01.
• Extends coverage for # placement, “integer then junk” cases, and reference-token character boundaries (control chars, DEL, astral, U+10FFFF).
• Adds explicit embedded-NUL token cases using std::string_view lengths with static_assert to avoid C++ \xNN greediness pitfalls.

Technical Notes: The added tests are coverage-only (core already passes them) and exercise edge cases around parsing/termination and RFC 6901 escape handling.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. No suggestions at this time.

Comment augment review to trigger a new review at any time.

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}