Skip to content

Template gallery

Every jm object or jm function invocation produces a project shaped to one of a handful of patterns — a preset. This gallery shows what each preset generates before you run it, so you can browse, find the shape that matches your work, and run the exact command shown at the top of the page.

Each page is titled with the CLI line that materialises it. Run it verbatim, then open <comp>_core.c and replace the /* TODO */ markers with your algorithm.

Preset names describe what the component does to data, not what domain you're working in. A "processor" is any 1:1 input→output transform — a DSP filter, a Q15→float converter, and a CSV row re-encoder all fit. Each preset page lists concrete examples across domains so you can recognise your shape.

Presets

Preset Data-flow shape Concrete examples Page
processor input → output (1:1) DSP filter, Q15→float, running-average smoother, byte-to-token jm object NAME --preset processor
generator () → output NCO, LFSR, counter, UUID, queue drainer, tokenizer jm object NAME --preset generator
consumer input → () running mean, integrator, checksum, log writer, metric reporter jm object NAME --preset consumer
reader external source → output file reader, CSV row reader, WAV/PNG loader, TCP socket consumer jm object NAME --preset reader
function free C function (no class) unit conversion, lookup, CRC, format detector, pure transform jm function FN --module MOD
blockwise array input → array output FFT, overlap-save filter, CSV batch transformer, image kernel jm object NAME --preset blockwise

--preset NAME is shorthand for the underlying flag bundle (e.g. --preset generator expands to --arg-type void); you can always pass those flags directly instead.

blockwise (array-in → array-out) uses --preset blockwise. The default element type is float _Complex; override with explicit --arg-type and --return-type flags. See the blockwise page for the generated C/Python shapes and the plan-once/execute-many FFTW pattern.

Need variable-output (zero or more outputs per call — peak detector, event finder, syllable boundary detector)? That's a capability flag, not its own preset — add --variable-output --max-out N to any preset that has output and declare per-event fields with repeatable --result-field name:T.

How to read each page

Every preset page has the same five sections:

  1. Command — the exact jm invocation. Copy, paste, run.
  2. What you get — the generated files (_core.h, _core.c, _ext.c highlights, the test, the Python .pyi). Real output, not pseudocode.
  3. What you fill in — the /* TODO */ line(s) and what the finished body would look like for a typical algorithm of that shape.
  4. Python usage — what import + call sites look like once jm build is done.
  5. Concrete types — the allowlist for each slot the preset exposes (--arg-type, --state, --init-param, etc.). Rows link back to the master Type slots page so you can cross-reference. A type that isn't in a slot's row is rejected by the CLI and by jm bind.

Status

Goal: every preset's command produces a scaffold that compiles and passes jm build && jm test immediately. Fill in the /* TODO */ body with your logic; everything around it stays green.

All six presets — processor, generator, consumer, reader, blockwise, and function — build and test green straight from scaffold. Fill in the /* TODO */ body with your logic; everything around it stays green. Use a real snake_case component name (my_filter, iq_reader) when you run these; the examples use NAME only as a placeholder.

Refreshing a scaffold

The _core.c you edit is sacred — jm apply never overwrites it, and the additive verbs (jm method, computed jm property) only inject a new declaration and append a fresh stub, leaving your existing bodies intact. Adding state with jm add is the exception: it rebuilds the object from the manifest, so keep your body in impl/create_impl so it survives. To deliberately throw the scaffold away and rebuild it from the manifest, run jm regenerate <component> (git stash first — it discards your hand-written _core.c). See Type slots for the full sacred/glue contract.