Write parser output only after successful generation

[ydah]

Command rendered the parser output before validating the computed states. When %expect did not match the actual conflict count, Lrama exited with status 1 only after writing a complete output file, allowing build tools to treat an unvalidated parser as up to date.

Validate the states before rendering the parser output. Render into an in-memory buffer and write the target file only after rendering completes, preserving an existing output file when rendering raises an exception.

Add command-level regression coverage for %expect mismatches producing no parser output and rendering failures preserving existing output.

[Copilot]

🟢 Approval recommended

The change directly addresses the reported failure mode and includes targeted regression tests that cover both %expect mismatches and render exceptions without introducing obvious behavioral regressions.

Pull request overview

This pull request ensures Lrama only writes the generated parser output file after the computed states have been successfully validated, preventing build tools from treating an invalidated parser as up to date. It also adds regression coverage to confirm that %expect mismatches and rendering failures do not clobber parser outputs.

Changes:

• Reorders the command workflow to validate computed states before rendering/writing the parser output.
• Buffers rendered parser output in-memory and writes the output file only after rendering completes successfully.
• Adds command-level specs covering %expect mismatch behavior (no output file created) and render exceptions preserving existing output.

File summaries

File	Description
spec/lrama/command_spec.rb	Adds regression tests ensuring no output is written on %expect mismatch and existing output is preserved on render failure.
lib/lrama/command.rb	Validates states before rendering and buffers output in StringIO before writing to the target file.

Review details

• Files reviewed: 2/2 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.