Solve a model with the Quicopt service — solve

Encodes the model, sends it, and returns the parsed result. The first keyless call mints an API key, remembered for the rest of the R session; pass api_key to authenticate with a key you already hold (used as-is, never remembered).

Usage

solve_model(
  m,
  base_url = DEFAULT_BASE_URL,
  api_key = NULL,
  project = NULL,
  config = NULL,
  gzip = FALSE,
  timeout = 60,
  transport = NULL
)

# S3 method for class 'quicopt_model'
solve(a, b, ...)

Arguments

m: A model(), a program(), or already-encoded bytes.
base_url: The service to talk to; defaults to DEFAULT_BASE_URL.
api_key: A key you hold, or NULL to mint and reuse a free-tier key.
project: A project tag for per-project invoicing, or NULL.
config: Named list of extra query parameters; a source_language here overrides the automatic tag.
gzip: Compress the request body.
timeout: Seconds to wait for the service.
transport: The HTTP layer, replaceable for testing: a function(req) taking list(method, url, headers, body, timeout) and returning list(status, headers, body).
a: The model (the argument is named a to match the solve generic).
b: Unused; supplying it is an error.
...: Passed on to solve_model().

Value

The service's answer as a quicopt_result: a list with status, objective, feasible, solution (a named numeric vector), model_class (the class the service read the model as, e.g. "milp"), and the ready-to-print display. Printing the result prints display.

Details

solve(m) and solve_model(m) are the same function — the former extends the base::solve generic (as solve.qr does), the latter is the unambiguous name for pipes and for reading. For a long-running model, or a first call against a cold worker, submit() queues the same request asynchronously instead of blocking.