Matrix Runs

Expand a single step into multiple parallel instances across a parameter grid.

Defining a matrix

- id: fit-model
  script: models/fit.R
  matrix:
    algo: [lm, glm, gam]
    dataset: [train, full]
  args: ["--algo", "{{ .Matrix.algo }}", "--data", "{{ .Matrix.dataset }}"]

This expands into 6 step instances (3 algorithms x 2 datasets), running in parallel.

Environment variables

Each matrix instance receives its parameters as environment variables:

• DAGGLE_MATRIX_ALGO – lm, glm, or gam
• DAGGLE_MATRIX_DATASET – train or full

Format: DAGGLE_MATRIX_<KEY> (uppercased).

Limiting parallelism

By default, all matrix instances run in parallel. Use max_parallel to cap concurrency:

- id: fit-model
  script: models/fit.R
  matrix:
    algo: [lm, glm, gam, rf, xgb]
  max_parallel: 2    # run at most 2 at a time

Template access

Use { .Matrix.key } in any string field:

- id: render
  quarto: reports/template.qmd
  matrix:
    region: [north, south, east, west]
  args: ["-P", "region:{{ .Matrix.region }}"]