[mono-move] String natives

[georgemitenkov]

Description

Implements the remaining 0x1::string natives for mono-move (the v2 VM). Only internal_check_utf8 was implemented before; this adds the other three so Move code that calls string::sub_string and string::index_of can run:

• internal_is_char_boundary(v: &vector<u8>, i: u64): bool
• internal_sub_string(v: &vector<u8>, i: u64, j: u64): vector<u8>
• internal_index_of(v: &vector<u8>, r: &vector<u8>): u64

Each native mirrors the legacy VM logic so both VMs produce identical results. internal_sub_string copies the result off the VM heap before allocating the return vector, since new_byte_vector can trigger a GC that relocates the source bytes.

How Has This Been Tested?

Renamed the existing check_utf8.move differential test to string.move and added cases for the new natives, run on both the legacy MoveVM and mono-move with their outputs compared:

• index_of: match found, first of multiple occurrences, no match (returns length), empty needle, multibyte content.
• sub_string: basic ranges, empty range, full string, out-of-range and reversed indices (abort), multibyte boundary success.
• is_char_boundary has no public wrapper, so it is exercised through sub_string's boundary asserts: the multibyte success cases require it to return true, the mid-codepoint cases require false (abort).

cargo build -p mono-move-natives
RUST_MIN_STACK=1073741824 cargo test -p mono-move-testsuite --test differential

All 193 differential cases pass.

Key Areas to Review

• Argument order in the natives matches the Move signatures (arg(0), arg(1), arg(2)).
• GC safety: byte slices are consumed before any allocation, and the substring is copied to an owned Vec before new_byte_vector.

string::insert is not covered end-to-end: it calls vector::append, which needs the 0x1::vector::move_range native that mono-move does not implement yet. That is out of scope here, and is_char_boundary is fully covered through sub_string instead.

Type of Change

• New feature

Which Components or Systems Does This Change Impact?

• Move/Aptos Virtual Machine

Checklist

• I have read and followed the CONTRIBUTING doc
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I identified and added all stakeholders and component owners affected by this change as reviewers
• I tested both happy and unhappy path of the functionality
• I have made corresponding changes to the documentation

---

Note

Medium Risk
Changes VM string execution and GC-sensitive native code; behavior is constrained by differential tests against the legacy VM but substring/index logic is user-visible on-chain.

Overview
Adds three mono-move natives for 0x1::string so string::index_of and string::sub_string work in the v2 VM alongside the existing internal_check_utf8. New implementations are internal_is_char_boundary, internal_sub_string, and internal_index_of, registered in make_all_string_natives.

Shared from_utf8_checked treats invalid UTF-8 on already-constructed strings as a VM invariant violation. internal_sub_string aborts with code 1 when j < i; out-of-range and non-boundary indices are still enforced by Move wrappers. Substring bytes are copied to an owned Vec before new_byte_vector to avoid GC relocating source slices.

Differential coverage moves from check_utf8.move to string.move, keeping utf8 cases and adding index_of / sub_string scenarios (including multibyte boundaries via sub_string asserts).

^{Reviewed by Cursor Bugbot for commit ed7b4ce. Bugbot is set up for automated code reviews on this repo. Configure here.}

[georgemitenkov]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• [mono-move] String natives #20163 👈 (View in Graphite)
• [mono-move] Storage slot natives #20160
• [mono-move] Add secp256k1 natives #20159
• [mono-move] Small benchmark improvements #20158
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[github-actions]

mono-move benchmark gate

2 regression(s) beyond ±3% noise band

3 ok · 2 improved · 0 new · 0 absent (threshold T = ±3%, criterion mean CI vs main)

Benchmark	mean Δ	95% CI	median (main → PR)	Verdict
fib/mono	-0.7%	[-1.7%, +0.2%]	8.07ms → 8.07ms	ok
nested_loop/mono	+19.3%	[+19.1%, +19.5%]	11.85ms → 14.10ms	regression
merge_sort/mono	+4.1%	[+3.7%, +4.4%]	1.22ms → 1.27ms	regression
bst/mono	+3.0%	[+2.8%, +3.2%]	3.63ms → 3.74ms	ok
match_sum/mono	-6.5%	[-7.3%, -5.9%]	25.72ms → 24.33ms	improved
int_arith_loop/mono_u64	-0.8%	[-1.9%, +0.1%]	408.83µs → 405.35µs	ok
int_arith_loop/mono_i64	-13.7%	[-14.1%, -13.2%]	678.34µs → 585.60µs	improved

Improvements are not failures. main rebaselines on merge, so the next PR compares against the faster code automatically.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit ed7b4ce. Configure here.}

[cursor]

Verbose multi-line header block in test file

Low Severity

The 10-line header block narrates which natives are tested, how they're exercised (through wrappers), and provides byte-level encoding details for "héllo." Per the MonoMove review convention, verbose multi-line header blocks in test files narrating what tests exercise are flagged — brief per-function comments suffice. The per-function comments already present in the file (e.g., // Returns the index of the first occurrence.) are fine.

 

^{Triggered by learned rule: MonoMove: flag verbose AI-generated comments in test and code files}

^{Reviewed by Cursor Bugbot for commit ed7b4ce. Configure here.}

[vgao1996]

Are we safely checking if i and j are char boundaries here?

[vgao1996]

Nit: or v.len() if absent