Change empty-match exit from break 2 to break

[JanTvrdik]

Summary

• Changes break 2 to break on empty match in PatternIterator::getIterator() for simpler control flow that is easier to reason about.
• Adds a regression test (testZeroLengthMatchMidBufferLoadsMoreData) demonstrating a case where break 2 would incorrectly abort: when an empty match occurs mid-buffer because a fixed-width alternative needs more characters than remain, loading the next chunk resolves it into a positive-length match.

Context

With break 2, an empty match mid-buffer with the stream still active would immediately exit both loops and throw. With break, the iterator returns to the outer loop, loads the next chunk, and retries — which can turn the empty match into a successful one when more data is available.

Since the library's SQL tokenizer patterns never produce empty matches in practice, the break 2 fail-fast offered no real benefit while being harder to reason about.

Test plan

• vendor/bin/tester tests/ — all 65 tests pass

[Copilot]

Pull request overview

This PR adds a regression test for the fail-fast behavior of PatternIterator when encountering an empty regex match. The test verifies that the break 2 statement (which exits both inner and outer loops) prevents unnecessary stream consumption when an empty match occurs mid-buffer, since loading more data cannot help resolve the situation.

Changes:

• Adds testEmptyMatchFailsEarlyWithoutConsumingEntireStream() to verify that only 2 chunks are touched (1 processed + 1 prefetched) when an empty match triggers early exit, rather than draining all 4 chunks.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.