feat: support Codama root-node IDLs and variable discriminators

[adiman9]

Summary

• add fallback parsing for Codama rootNode IDLs in arete-idl
• convert Codama programs, accounts, instructions, errors, and defined types into IdlSpec
• fix macro codegen/parser generation for variable-length and embedded account discriminators
• add targeted tests for Codama parsing and discriminator handling
• keep the arete-a4-sdk changes formatting-only

Why

Some Steel/Codama-generated IDLs are emitted as Codama root-node JSON rather than the legacy format arete-idl expects today. Those IDLs also use discriminator patterns that are not always 8-byte Anchor-style discriminators. This branch teaches the IDL layer and generated parsers to accept those layouts without regressing the existing path.

What Changed

arete-idl

• parse_idl_content now falls back to deserializing a Codama root node when direct IdlSpec parsing fails
• maps Codama program metadata, instructions, accounts, defined types, and errors into the internal IdlSpec
• derives account discriminators from the accountDiscriminator enum when present
• derives instruction discriminants from the synthetic discriminator argument default value and removes that field from emitted args
• supports number, public key, string, defined type links, arrays, fixed-size types, structs, and fieldless enums in the Codama conversion path
• records Codama metadata as spec: "codama" and origin: "codama-rootNode"

arete-macros

• generated account parsers now use starts_with(...) instead of assuming an 8-byte discriminator slice
• account matching is sorted by discriminator length descending so shorter prefixes do not mask longer ones
• generated account decoders detect when the payload itself starts with an embedded discriminator field and deserialize from the full buffer instead of skipping bytes first

rust/arete-a4-sdk

• rustfmt-only reformat of the auth-related files included in the branch
• no functional behavior changes in the SDK portion of this PR

Constraints

• Codama enums with payload fields are still rejected because arete-idl only models fieldless enums today
• the Codama conversion path currently expects root-node shaped JSON and looks for account discriminators in a defined type named accountDiscriminator

Testing

• cargo test -p arete-idl -p arete-macros -p arete-a4-sdk

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
arete-docs	Ready	Preview, Comment	Jun 17, 2026 9:34pm

[greptile-apps]

Greptile Summary

This PR adds Codama rootNode IDL parsing to arete-idl and fixes macro code generation for variable-length and embedded account discriminators. It also brings gRPC connection management improvements, negative caching for token resolvers, and type name deduplication to the interpreter generators.

• arete-idl: adds a fallback parse path that deserializes Codama root-node JSON and maps it to IdlSpec; errors from both parsers are now surfaced together; CodamaTypeNode/CodamaCountNode gain #[serde(other)] catch-alls so unknown node kinds produce clear errors.
• arete-macros: account parser switches from a fixed 8-byte slice match to starts_with(...) sorted by discriminator length; code gen detects embedded discriminator structs and skips the skip-bytes step; generated gRPC connection loop gains exponential backoff, keepalive, and a semaphore-gated resolver concurrency cap.
• interpreter: Rust/TypeScript generators add name-collision deduplication for resolved struct types; runtime_resolvers gains per-path profiling and negative-cache handling for token metadata misses.

Confidence Score: 5/5

The change is safe to merge. All correctness concerns raised in earlier review rounds have been addressed — discriminant extraction now errors on missing defaults, account discriminator lookups return errors when the map is non-empty but an account is absent, and both parser errors are surfaced together. Resolver mutations are still combined with primary mutations in a single batch before emission.

The IDL parsing, macro codegen, and runtime changes are well-structured, each new code path is covered by targeted tests, and the previous blocking concerns have been resolved. The embedded-discriminator heuristic is sound for Steel/Codama naming conventions, and the gRPC backoff and resolver semaphore additions improve resilience without introducing new data-path risks.

arete-macros/src/codegen/vixen_runtime.rs — largest surface area of new generated runtime code; worth a second read on the ManagedYellowstoneGrpcSource SourceTrait implementation and the async_resolver_order counter lifetime.

Important Files Changed

Filename	Overview
arete-idl/src/parse.rs	Adds Codama root-node fallback parsing with all previous-thread concerns addressed: both errors propagated, Unknown catch-alls added, discriminator misses now return errors, discriminant extraction validates numeric default_value.
arete-macros/src/idl_parser_gen.rs	Replaces fixed 8-byte discriminator slice match with starts_with() sorted by length descending; min-length check relaxed to data.is_empty(); well-tested with the new parsers test module.
arete-macros/src/idl_codegen.rs	Adds embedded-discriminator detection via field-name/type heuristic and routes borsh/bytemuck deserialization to the full buffer for those accounts; tests confirm both detection and decode path.
arete-macros/src/codegen/vixen_runtime.rs	Large addition: ManagedYellowstoneGrpcSource, exponential backoff, keepalive, resolver semaphore, and async_resolver_order counter; resolver mutations are still combined with primary mutations in a single batch preserving the no-duplicate-frames guarantee.
interpreter/src/runtime_resolvers.rs	Adds per-stage profiling, negative-cache handling for token metadata misses, and UpdateContext propagation; logic is clean and the context is always restored after apply.
interpreter/src/vm.rs	Adds retry tracking, per-entry TTL, and negative caching for resolver entries; in_flight/queued state transitions look consistent across take, restore, and schedule_retry paths.
interpreter/src/rust.rs	Type-name deduplication via build_resolved_type_name_map; field_type_to_rust still returns base type (underscore-prefixed map param intentionally unused); tests confirm no regression on existing generated types.
interpreter/src/typescript.rs	Mirrors the Rust name-deduplication pattern; resolved_type_to_interface_name rebuilds the name map on every call, though impact is limited to code generation.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[parse_idl_content] -->|try| B[serde_json: IdlSpec]
    B -->|Ok| C[IdlSpec returned]
    B -->|Err: IdlSpec error| D[try Codama fallback]
    D -->|serde_json: CodamaRoot| E{kind == rootNode?}
    E -->|No| F[Return error]
    E -->|Yes| G[Extract accountDiscriminator\nfrom defined_types]
    G --> H[codama_account_to_idl\nfor each account]
    G --> I[codama_instruction_to_idl\nfor each instruction]
    G --> J[codama_defined_type_to_idl\nfor each defined type]
    H --> K[Lookup discriminator\nfrom HashMap]
    K -->|found| L[vec discriminator byte]
    K -->|not found, map non-empty| M[Return Err - missing variant]
    K -->|map empty| N[vec empty - Anchor fallback]
    I --> O[Find discriminator arg\nwith numberValueNode default]
    O -->|found| P[SteelDiscriminant with value]
    O -->|missing default| Q[Return Err - cannot determine]
    L & P & J --> R[IdlSpec assembled]
    D -->|Err: Codama error| S[Both errors surfaced]

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
    A[parse_idl_content] -->|try| B[serde_json: IdlSpec]
    B -->|Ok| C[IdlSpec returned]
    B -->|Err: IdlSpec error| D[try Codama fallback]
    D -->|serde_json: CodamaRoot| E{kind == rootNode?}
    E -->|No| F[Return error]
    E -->|Yes| G[Extract accountDiscriminator\nfrom defined_types]
    G --> H[codama_account_to_idl\nfor each account]
    G --> I[codama_instruction_to_idl\nfor each instruction]
    G --> J[codama_defined_type_to_idl\nfor each defined type]
    H --> K[Lookup discriminator\nfrom HashMap]
    K -->|found| L[vec discriminator byte]
    K -->|not found, map non-empty| M[Return Err - missing variant]
    K -->|map empty| N[vec empty - Anchor fallback]
    I --> O[Find discriminator arg\nwith numberValueNode default]
    O -->|found| P[SteelDiscriminant with value]
    O -->|missing default| Q[Return Err - cannot determine]
    L & P & J --> R[IdlSpec assembled]
    D -->|Err: Codama error| S[Both errors surfaced]

Loading

_{Reviews (10): Last reviewed commit: "fix: surface errors for unsupported Coda..." | Re-trigger Greptile}