codama-idl/codama @codama/node-types@1.10.0

on GitHub

Minor Changes

• #1025 add9be2 Thanks @lorisleiva! - Omit empty array attributes from nodes, and default absent arrays to [] on read.

  Every node array attribute — whether previously required (e.g. programNode.accounts, instructionNode.arguments, structTypeNode.fields) or optional (e.g. instructionNode.discriminators) — is now omitted from the node when empty, and defaults to [] when absent. An absent array and an empty array are semantically identical, so this keeps encoded IDLs small (they are often uploaded on-chain) and makes adding or omitting an array attribute non-breaking. Reading remains backwards-compatible: IDLs that still serialise empty arrays continue to parse, and every visitor and accessor normalises an absent array to [].

  Breaking type change. While runtime reads stay backwards-compatible, this is a breaking change at the type level for external consumers such as renderers and community packages. Array attributes are now typed as optional on the node interfaces (e.g. ProgramNode.accounts is now readonly accounts?: AccountNode[]), so any code that reads an array attribute without guarding — e.g. program.accounts.map(…) — must now coalesce with ?? [] ((program.accounts ?? []).map(…)). Type-parameter constraints also widen from Array<T> to Array<T> | undefined, so indexing into these types (e.g. InstructionNode['arguments'][number]) must be rewritten as NonNullable<InstructionNode['arguments']>[number]. Consumers should expect compile errors on the first unguarded array read after upgrading and add ?? [] guards accordingly.