Knowledge Transfer

Migration projects create knowledge silos if not managed deliberately. This chapter covers documentation-driven development, Oracle mode as a knowledge base, and cross-training on Rust idioms.

Documentation-Driven Development

Every migrated module should have a doc comment explaining its origin:

#![allow(unused)]
fn main() {
//! # Data Loader
//!
//! Migrated from `src/data_loader.py`.
//!
//! ## Key Changes
//! - `load_csv()` returns `Result<DataFrame>` instead of raising exceptions
//! - NumPy operations replaced with trueno `Vector`
//! - File I/O uses `BufReader` for 3x throughput improvement
}

Oracle Mode as Knowledge Base

Batuta’s Oracle provides natural language access to stack knowledge:

batuta oracle "How do I load a model with quantization?"
batuta oracle --recipe ml-random-forest --format code
batuta oracle --rag "tokenization pipeline"

Re-index after adding documentation:

batuta oracle --rag-index

Cross-Training on Rust Idioms

Python-to-Rust Mental Model Shifts

Python Concept	Rust Equivalent	Key Difference
`try/except`	`Result<T, E>` + `?`	Errors are values
`None` checks	`Option<T>` + `.map()`	Compiler-enforced null safety
`class`	`struct` + `impl`	No inheritance; use traits
List comprehension	`.iter().map().collect()`	Lazy evaluation
`with` context manager	`Drop` trait	Automatic cleanup on scope exit

Recommended Learning Path

Week 1-2: Rust Book chapters 1-10 (ownership, borrowing, traits)
Week 3-4: Read stack code with pmat query --include-source
Week 5-6: Pair-program on a low-risk migration
Week 7+: Independent migration with mentored review

Knowledge Artifacts

Artifact	Location	Purpose
CLAUDE.md	Project root	Machine-readable project context
Oracle recipes	`batuta oracle --cookbook`	Code patterns with tests
mdBook	`book/src/`	Comprehensive reference
API docs	`cargo doc --no-deps`	Generated from doc comments

Navigate: Table of Contents

Keyboard shortcuts