🚀 T v0.54.2 is Live! This release added a new built-in package, called testcraft, which will make it much easier to verify the results produced by your reproducible pipelines! Whether you’re writing unit tests manually or building AI-driven workflows that auto-fix broken pipelines, this release has something for you. Here are the highlights: --- 🛠️ Introducing `testcraft`: A First-Class Assertion Framework Testing complex data structures, pipelines, and types should be intuitive and resilient. `testcraft` introduces 40 dedicated `expect_*` functions with built-in safety features: * Three-Tier Expectation System: Distinguishes between passed (`Expect_pass`), clear failure (`Expect_stop`), and indeterminate cases (`Expect_hold` for handling missing/`NA` values gracefully). * Deep VDict Comparison: Order-insensitive, recursive structural comparison of nested dictionaries with crash-safe evaluation. * Comprehensive Coverage: Native assertions covering equality, type checks, dataset/table structures, pipeline nodes, error handling, and string matching. * Inline Node Assertions: Use `check()` directly inside pipeline nodes to enforce invariants while preserving error signals. --- 🤖 Built for the AI Era: Agentic Testing in `t test` I designed the latest test runner enhancements so both developers and AI Agents can inspect, parse, and act on test results seamlessly: * 🤖 Structured JSON Output (`t test --json`): Emits schema v1 JSON containing per-file statuses, execution duration (`duration_ms`), and error details, ideal for feeding directly into LLM agent tools or CI automation loops. * 📊 REPL `t_test()` Returns a DataFrame: No more plain text logs when testing interactively! `t_test()` now returns a native DataFrame, allowing you to filter and inspect test results programmatically using T's pipeline operators: ```t t_test() |> filter($status == "failed") ``` --- ⚡ Modern Test Runner Power Features * Targeted Execution: Flexible filtering with `--only "stats"` and `--not "slow"`, plus `.tignore` file support. * Pipeline Auto-Evaluation: `t test` automatically evaluates embedded pipeline definitions and reports node failures inline. * CI & Observability Ready: Native JUnit XML output (`--format junit`), `--failfast` execution, execution timeouts (`--timeout`), and code coverage reporting (`--coverage`). --- 💡 Try it today! tstats-project.org/getting-s… #Reproducibility #RStats #Python #Julia