Envless.variables.apply

def apply(
  project,
  environment,
  changes = nil,
  product: nil,
  timeout: nil,
  **fields
)

def apply: (
  String project,
  String environment,
  ?(apply_changes | Hash[Symbol | String, untyped])? changes,
  ?product: (String | Symbol)?,
  ?timeout: seconds?,
  **untyped
) -> apply_result

Runs creates, updates and deletes against an environment inside one transaction, behind a per environment advisory lock. This is the method a sync client should use instead of a loop of single calls: the environment is never left half written and two concurrent applies cannot interleave.

baseline is required and is how drift is detected. Send every live variable as you last read it, as a Hash of uniqueId, name and updatedAt filled from the id, name and updatedAt of a read. Before touching anything the server compares your baseline against the live rows and refuses the whole request with 409 variable.baseline_drifted when a live variable is missing from your baseline, a baseline entry no longer exists, or an updatedAt differs by a single millisecond. A partial baseline is not a partial check, it is guaranteed drift. The drifted names appear only in the error message, so on a 409 re-read the environment and rebuild the change set.

Apply is value only and name only. creates write name and value and take the column defaults for everything else, updates write value alone, and there is no rename and no defaultValue path here, use update or update_by_id for those. Every creates and updates value goes through the same v1:/v2: prefix check create enforces, and one plaintext value refuses the whole change set with 400 variable.ciphertext_invalid before anything is written.

Every creates name must match ^[A-Za-z_][A-Za-z0-9_]*$ and be 1 to 100 characters, and is stored with exactly the casing you send. Names are compared without regard to case, so two creates entries differing only in case are a repeat, and creating api_key while API_KEY is live and not in deletes is a conflict. baseline names are not pattern checked, send each one back exactly as you read it so a legacy name still round-trips.

Example#

require "envless/sdk"

passphrase = ENV.fetch("ENVLESS_PASSPHRASE")
workspace_id = Envless.me.get[:workspaceId]
current = Envless.variables.list_all("api", "production")

baseline = current.map do |variable|
  {
    uniqueId: variable[:id],
    name: variable[:name],
    updatedAt: variable[:updatedAt] || variable[:createdAt]
  }
end
legacy = current.select do |variable|
  variable[:name].start_with?("LEGACY_")
end
flag = Envless.encrypt_value("true", passphrase, workspace_id)

result = Envless.variables.apply(
  "api",
  "production",
  baseline:,
  creates: [{name: "FEATURE_FLAG", value: flag}],
  deletes: legacy.map { |variable| {uniqueId: variable[:id]} }
)

puts "#{result[:added]} #{result[:updated]} #{result[:removed]}"

Parameters#

Returns#

apply_result with added, updated and removed counts for what the transaction wrote.

Worth knowing#

Required scopes#

variables:write

Underlying endpoint#

In TypeScript#