Polars 2.0 Migration Cheat Sheet and Regex Codemod Pack

Polars 2.0 removes melt, with_row_count, fetch, profile and a dozen more APIs, and the default engine is now streaming. This pack finds every break in your repo, rewrites the mechanical ones with a dry-run diff, and tells you which ones need a person. Tested against Polars 2.0.0 on 8 October 2026.

What You Get

• Migration cheat sheet. Every rename, removal, behaviour and dtype change from the official upgrade guide in tables, with the real 2.0.0 error messages so you can search your logs.
• Rule-driven codemod. A standard-library Python script with 17 rewrite rules in rules.json. It prints a unified diff by default and writes only with --write.
• Scoped, pandas-aware edits. Keyword renames stay inside the call. Files that import pandas are reported, never rewritten, and positional melt calls are left alone.
• Pre-upgrade audit script. Finds all 36 patterns across a repo, ranks them break, behavior, order or info, flags pins like polars<2, and exits non-zero for CI.
• 83 pytest tests. Every rule runs its before and after examples, its negative cases and an idempotence check, so you can trust the rewrite.
• Streaming row-order checklist. Six sections covering how to find code that depends on order, which fix to pick, and how to test with both engines.
• Before and after scripts. A Polars 1.x script that fails on 2.0.0, the codemod output that runs, and a demo of three ways to get stable order.
• Runbook and CI gate. Seven steps with rollback, a GitHub Actions workflow and a pre-commit snippet that block new removed-API usage.

How to Use

1. Run python3 -m pytest codemod/tests -q to confirm the tooling works.
2. Run the audit on your repo and read the break findings first.
3. Run the codemod without flags and read the diff it prints.
4. Run it again with --write, commit the mechanical edits on their own, then apply the review rules.
5. Fix the remaining audit items by hand and work through the row-order checklist.
6. Upgrade to Polars 2, run your tests under both engines, then add the CI gate.

Who This Is For

• Data engineers with Polars 1.x pipelines that must survive the 2.0 upgrade.
• Analytics teams whose reports depend on row order or on hash values.
• Tech leads who need an auditable list of what breaks before approving the upgrade.
• Consultants migrating several client repos who want one repeatable process.

Everything runs on the Python standard library, so you can drop it into any repo without touching dependencies. Browse more in the developer tools collection, or try JSON to CSV converter.