dr_contract_update()

Adds or replaces declared column types without reading data or executing rules. New columns are optional unless explicitly included in required. Change id for a derived product, or version for a revised definition of the same contract. The original specification is unchanged.

Usage

dr_contract_update(
  x,
  id = x$id,
  version = x$version,
  columns = NULL,
  remove = character(),
  ...
)

Arguments

x: A confirmed contract specification.
id, version: Identifier and definition version. At least one must differ from x; version ordering is not inferred.
columns: Named type vector or zero-length prototypes to add or replace, as in dr_contract(). Omit to retain the declared types.
remove: Character vector of declared column names to remove.
...: Named contract fields to replace in full, such as required, key, grain, rules, owner or column_metadata.

Value

A new contract specification. Use dr_contract_diff() to review it.

Details

Contracts describe final output expectations. Deferred dplyr transformations do not rewrite them. After renaming, remove the old column and add the new one here. After aggregation, explicitly set the new grain and key. A changed grain requires an explicit key, including character() when no key is claimed. Removing or changing a key column also requires an explicit key. Removing a required column requires an explicit required vector.

Existing rules must be explicitly replaced or reaffirmed with rules after a removal, type change or grain change. Rules may contain arbitrary R code, so their dependencies cannot safely be inferred. This review prevents silent inheritance of guarantees, but is not proof that revised rules are correct. Metadata for removed columns is discarded; other metadata is inherited.

Examples

orders <- dr_contract("orders", columns = c(id = "integer"), key = "id")
enriched <- orders |>
  dr_contract_update(id = "enriched_orders", columns = c(channel = "character"))
dr_contract_diff(orders, enriched)