Struct ProviderConn 

pub struct ProviderConn { /* private fields */ }

A connected provider — the protocol-matched client over a dialed channel.

Implementations

impl ProviderConn

pub fn new(channel: H2Channel, protocol: PluginProtocol) -> Self

Wrap a dialed channel, selecting the client by the handshake’s negotiated protocol (Plugin::handshake().app_protocol).

pub async fn get_schema(&mut self) -> Result<ProviderSchema, ProviderError>

GetProviderSchema (v6) / GetSchema (v5) → provider-config + per-resource implied types.

pub async fn configure( &mut self, config: &DynamicValue, terraform_version: &str, ) -> Result<(), ProviderError>

ConfigureProvider (v6) / Configure (v5) — provider creds/settings.

pub async fn plan_resource_change( &mut self, type_name: &str, prior_state: &DynamicValue, proposed_new_state: &DynamicValue, config: &DynamicValue, ) -> Result<PlannedChange, ProviderError>

PlanResourceChange — the provider’s proposed new state PLUS which attribute paths (if any) force a destroy+create instead of an in-place update. See PlannedChange’s doc for why the latter matters: it is the ONLY authoritative source for that decision, and was silently discarded here before PlannedChange existed.

pub async fn apply_resource_change( &mut self, type_name: &str, prior_state: &DynamicValue, planned_state: &DynamicValue, config: &DynamicValue, ) -> Result<DynamicValue, ProviderError>

ApplyResourceChange — execute the change. Returns the new state.

pub async fn read_resource( &mut self, type_name: &str, current_state: &DynamicValue, ) -> Result<Option<DynamicValue>, ProviderError>

ReadResource — read the resource’s ACTUAL current state from the provider (the refresh primitive). Returns Ok(None) when the provider reports the resource no longer exists (new_state is cty-null), so callers drop stale / phantom entries from state; Ok(Some(dv)) with the refreshed wire state when it still exists.

pub async fn read_data_source( &mut self, type_name: &str, config: &DynamicValue, ) -> Result<Option<DynamicValue>, ProviderError>

ReadDataSource — evaluate a data block by querying the provider, returning its result state. The apply engine reads data sources up front so ${data.<type>.<name>.<attr>} references resolve; without it those strings leaked verbatim to managed-resource RPCs (the rio-drive Cloudflare 400). Ok(None) if the provider returned cty-null.

pub async fn import_resource_state( &mut self, type_name: &str, id: &str, ) -> Result<Option<DynamicValue>, ProviderError>

ImportResourceState — adopt a resource that EXISTS in the cloud but is absent from magma’s state, by its provider-native import id (e.g. the repo name for github_repository). Returns the imported wire state (Ok(Some(dv))) or Ok(None) if the provider imported nothing / returned cty-null. This is the read/import half of the protocol that powers import-on-create-conflict (422 already-exists → observe → adopt)

• magma import. Mirrors apply/read’s V5/V6 dispatch + msgpack decode.

pub async fn upgrade_resource_state( &mut self, type_name: &str, stored_version: i64, raw_json: &[u8], ) -> Result<DynamicValue, ProviderError>

UpgradeResourceState — migrate a StateInstance’s raw attribute JSON (persisted under an older stored_version of the provider’s schema for type_name) forward to the CURRENT schema. The terraform plugin protocol requires this to run before a stored instance is fed into ReadResource/PlanResourceChange/ ApplyResourceChange whenever stored_version is older than the provider’s live ProviderSchema::resource_version — decoding old-schema JSON straight against the new implied type (skipping this call) risks a marshal mismatch or provider-side crash/misparse on any resource type whose schema evolved. raw_json is the instance’s raw attribute bytes as stored (magma persists state attributes as JSON, never the legacy flatmap format, so only RawState.json is populated). Returns the upgraded value, decodable via DynamicValue::to_json against the CURRENT implied type.