moonbitlang / core / 5143

91%

push

github

perf(bigint): build to_string digits in one pass instead of prepending (#3709)

* perf(bigint): build to_string digits in one pass instead of prepending

The base-10 `BigInt::to_string` emitted each decimal digit with
`ret = digit.to_string() + ret`, prepending into a freshly allocated
string on every digit. That is quadratic in the digit count and
allocates a one-char string plus a full copy of the accumulator per
digit.

Emit the digits least-significant-first into a pre-sized
`FixedArray[Char]` and materialize the result string once. The
limb->decimal base-conversion loop above this is a separate O(n^2) cost
and is left untouched; this only removes the quadratic string building
in the digit-emission phase.

Bench (BigInt::to_string, native):
    40 digits:  649.9 ns -> 148.1 ns  (4.4x)
   400 digits:    9.73 µs ->  2.76 µs (3.5x)
  4000 digits:  410.4 µs -> 194.5 µs  (2.1x)

Adds bigint/to_string_bench_test.mbt to cover the path.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* test(bigint): guard make_decimal against non-positive digit counts

Addresses review feedback: make_decimal claimed to produce exactly
`digits` digits but silently misbehaved for digits <= 0 (it still built
"1" and passed a negative size_hint to StringBuilder). Fail loudly on
misuse instead.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

9 of 9 new or added lines in 1 file covered. (100.0%)

15628 of 17132 relevant lines covered (91.22%)

191916.19 hits per line