dr_product()

Define a data delivery by name, its expected columns and its quality rules. A product is a reusable specification. Bind a source with dr_add_source(), or supply data when executing a workflow. Use dr_run() before writing.

Usage

dr_product(
  id,
  data = NULL,
  contract = NULL,
  version = "1.0.0",
  owner = NULL,
  description = NULL,
  code_version = NULL,
  source_name = NULL,
  execution = NULL,
  contracts = NULL,
  table = NULL
)

Arguments

id: Product identity, unique within a dependency graph.
data: Optional table, path, source adapter, product, or successful run. Successful lake runs are pinned to their exact published release.
contract: Optional contract, named type vector or prototype list.
version: Optional immutable definition version. When omitted, lake publication derives a technical version from the definition.
owner, description: Optional metadata, otherwise inherited from contract.
code_version: Optional code and dependency version. Required only when explicitly reusing a previously published lake release with cache = TRUE.
source_name: Optional name for data. Defaults to the product ID for ordinary inputs, or the upstream product ID for a nested product.
execution: Optional connection-free dr_execution_config() stored on this definition. Used when it is the root of dr_run(), dr_publish() or dataraft.lake::dr_ingest(). An explicit execution argument overrides these defaults.
contracts: Named table contracts when data is a dm model.
table: Table name to select from a successful model trial or publication. Published selections retain the exact member release.

Value

A product, ready for composition, inspection and execution.

Details

Add named sources, ordinary transformation functions and optional checks or a target. Use dr_run() to try it, or dr_publish() to save checked output. dr_run() executes the full configuration, including writers. Products can be sources of other products; shared dependencies run once per execution.

Table models

A table input defines a table product. A dm input defines a model product: its tables and declared relationships are checked together and published in one transaction. dr_collect() returns a dm; select a member explicitly with dr_product("report", result, table = "policies"). Model execution materializes all tables in memory. Table contracts are supplied through contracts.

Examples

orders <- dr_product("orders", data.frame(id = 1:2, amount = c(25, 75))) |>
  dplyr::mutate(amount = round(amount, 2)) |>
  dr_add_quality(~ amount >= 0)
orders |> dr_run(write = FALSE, stop_on_failure = FALSE) |> dr_collect()

Documentation source snapshot · View source at d9cb454 ↗
Code examples are documented source examples; they are not executed in this website build.