Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Choosing a target

There is no universally “best” target — the right destination depends on the shape of the pressure you are feeling. Use the matrix below to map a signal to an ecosystem, then open the matching playbook.

Target-selection matrix

Pressure / needRecommended targetNotes
General N-D numeric arrays, dense matmul, axis reductions at scalendarrayThe general Rust N-D array production path; BLAS-backed matmul available.
Small/mid dense vectors & matrices, decompositions, solvers (LU/QR/SVD), eigenvaluesnalgebraThe dense linear-algebra path.
Group-by, joins, pivots, query-style dataframe analyticsPolars (Rust) / Pandas (Python)matten-data is an ingestion on-ramp only; it will not grow these.
Autodiff, training loops, GPU/device executionCandle (Rust) / framework of choicematten is not an ML framework.
Existing Python scientific stack, NumPy interopNumPy (Python)Manual/conceptual hand-off; no automatic bridge.
Small numeric work, ingestion, glue, learning/PoCstay with mattenMigrating here would add dependencies for no real gain.

A quick decision path

  1. Is the bottleneck a dense numeric kernel (matmul, matrix–vector, operator application, axis reductions) that you have measured as hot? → ndarray (general N-D) or nalgebra (if it is fundamentally small/mid dense linear algebra needing decompositions).
  2. Do you need linear-algebra results matten does not provide (LU/QR/SVD, solvers, eigenvalues)? → nalgebra.
  3. Is the real need tabular (group-by/join/pivot/query)? → Polars (Rust) or Pandas (Python). Not matten-data.
  4. Is it ML (autodiff/training/GPU)? → Candle or another ML framework.
  5. Are you already in Python? → NumPy/Pandas, with matten as the upstream Rust producer if useful.
  6. None of the above, or the work is small? → stay with matten.

Playbooks

Full per-target playbooks are available for every destination above: ndarray, nalgebra, Polars/Pandas, Candle, and NumPy. The two Rust array/linalg targets carry task-scoped positioning notes from the accepted RFC-049 peer comparison; the dataframe, ML, and Python targets are different paradigms with no such benchmark (see each playbook).