Investigate `lark` as an alternative to `antlr4-python3-runtime` in `openqasm3.parser`

[TheGupta2012]

Background

The current OpenQASM3 parser is a known performance bottleneck, especially for large or complex QASM files. Rather than a full migration, this issue proposes a focused experiment: fork the repo, implement a Lark-based parser in parser.py for a subset of the grammar, and evaluate feasibility, performance, and compatibility.

Goals

• Evaluate if Lark can serve as a drop-in replacement for the current parser logic.
• Maintain API and object compatibility with existing code.
• Benchmark correctness and performance on OpenQASM3 files.

---

Proposed Steps

1. Fork and Prepare

• Fork the OpenQASM repository.
• Create a feature branch (e.g., lark-parser-prototype).

2. Analyze parser.py

• Identify key functions/classes responsible for:
  • Parsing input strings/files
  • Emitting AST or domain objects
  • Error handling and reporting

3. Develop Lark Grammar

• Extract a minimal but representative subset of the OpenQASM3 grammar (e.g., statements, declarations, a few gate definitions).
• Write the equivalent Lark EBNF grammar.

4. Implement Lark Parser

• In parser.py, add a new parser class or function using Lark.
• Use a Transformer to convert the Lark parse tree into the same AST or objects as the current parser.
• Add a feature flag or environment variable to switch between the old and new parsers for testing.

5. Ensure API Compatibility

• Match all public methods, return types, and error messages to the existing API.
• Write or adapt unit tests to verify that the Lark-based parser produces identical results.

6. Benchmark and Report

• Compare parsing speed and memory usage between the original and Lark-based parser.
• Document any edge cases, limitations, or incompatibilities encountered.

7. Share Findings

• Summarize results in a markdown file or GitHub issue.
• If successful, outline next steps for incremental migration.
• Eventually, we want this change to be merged upstream into openqasm3 if the maintainers are willing. If not, we can merge this as a parser component within pyqasm itself.

---

References

• Lark Documentation
• OpenQASM3 Parser (parser.py)