Skip to content

CHANGELOG.md

Latest commit

 

History

History
200 lines (164 loc) · 11.7 KB

CHANGELOG.md

File metadata and controls

200 lines (164 loc) · 11.7 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.7.0] - 2025-09-15

Added

  • Unified IR Pipeline: Complete DAG-based compilation pipeline with symbol tables and dependency graphs
  • Two-Phase Compilation: Analysis phase builds global symbol table, synthesis phase generates code with full dependency knowledge
  • Special Case Registry: TOML-driven configuration system for edge cases, eliminating all hardcoded special logic
  • Module Registry: Centralized module resolution with lexical scoping and cross-module dependency tracking
  • Compilation Unit: Intermediate representation between analysis and synthesis phases for clean separation of concerns
  • Pattern-Based Module Aliases: Intelligent module alias generation (e.g., metav1, corev1, storagev1) to avoid naming conflicts
  • Consistent mod.ncl Structure: Every directory now includes a mod.ncl file that imports all types in that directory

Changed

  • Major Architecture Overhaul: Migrated from ad-hoc processing to structured two-phase compilation with proper symbol tables
  • Module System: Complete rewrite with hierarchical mod.ncl generation and proper import hoisting
  • Import Resolution: All imports now hoisted to module level, eliminating inline imports throughout generated code
  • Import Paths: All import paths now use consistent mod.ncl structure (e.g., ../core/v1/mod.ncl instead of ../core/v1.ncl)
  • Module Alias Generation: Replaced hardcoded if-else chains with pattern-based alias extraction from import paths
  • Special Cases: Moved all K8s and Crossplane special handling from code to declarative TOML configuration files
  • Directory Structure: Eliminated redundant subdirectories in package generation (e.g., crossplane/apiextensions.crossplane.io)
  • Test Infrastructure: Moved test package generation to system temporary directory instead of cluttering examples folder
  • Code Quality: Removed ~300 lines of legacy code, fixed all compilation warnings and clippy lints
  • Error Types: Large PipelineError variants now boxed to reduce stack allocation pressure

Fixed

  • K8s Core Type Imports: Fixed incorrect import paths for TypedLocalObjectReference, PodTemplateSpec, and other core types
  • Duplicate Module Variables: Resolved naming conflicts by using unique module aliases (metav1, corev1) instead of duplicate v1Module declarations
  • Cross-Module References: Properly resolved through two-phase compilation and dependency analysis
  • Import Path Consistency: All imports now use proper relative paths calculated via ModuleRegistry
  • Circular Dependencies: Added detection and proper error reporting via petgraph
  • Test Failures: Fixed naming convention tests, module registry tests, and snapshot tests
  • Memory Safety: Eliminated all unwrap(), expect(), and panic!() calls in favor of proper Result types
  • Memory Optimization: Reduced PipelineError stack footprint from 184 bytes to ~16 bytes by boxing large error variants
  • Clippy Warnings: Fixed all warnings including needless_borrow, manual_flatten, vec_init_then_push, and function argument count issues
  • Code Quality: Fixed all clippy warnings including result_large_err, derivable_impls, and should_implement_trait
  • Error Handling: Improved error message formatting and Display implementations for better debugging experience

Removed

  • Hardcoded Logic: Eliminated all hardcoded K8s and Crossplane special cases (now in TOML configuration)
  • Inline Imports: All imports now properly hoisted to module level
  • Hardcoded Module Aliases: Replaced with pattern-based extraction from import paths

[0.6.4] - 2025-09-01

Added

  • Recursive Type Discovery: Replaced hardcoded namespace lists with intelligent recursive discovery that automatically finds all referenced types
  • Comprehensive Type Coverage: Expanded from 199 to 321 Kubernetes types through recursive discovery including versioned APIs (v1alpha1, v1beta1, v2)
  • Support for unversioned k8s types (e.g., RawExtension, IntOrString) placed in v0 directory to avoid conflicts
  • Reserved keyword escaping for field names starting with $ (like $ref, $schema) in generated Nickel code

Fixed

  • Required Field Usability Issue: Made all fields optional by default to enable gradual object construction (e.g., k8s.v1.LabelSelector & {} now works)
  • Cross-Package Import Resolution: Fixed imports to use full package IDs from manifest configuration instead of bare package names
  • Missing type references (e.g., RawExtension, NodeSelector) now properly discovered and generated
  • Cross-version imports for unversioned runtime types (v0 → other versions)
  • Syntax errors from unescaped special field names in JSON Schema types
  • Reserved keyword escaping in JSON object field names within default values

Changed

  • Breaking: All generated fields are now optional by default instead of required, enabling practical usage patterns
  • k8s type extraction now uses seed-based recursive discovery instead of fixed namespace lists
  • Updated to Kubernetes v1.33.4 schema version (from v1.31.0)
  • Unversioned types are placed in v0 to distinguish from versioned APIs
  • Enhanced import logic handles both v1 core types and v0 unversioned types
  • Package imports now use full package IDs like "github:seryl/nickel-pkgs/pkgs/k8s_io" for consistency

[0.6.3] - 2025-09-01

Added

  • Recursive Type Discovery: Replaced hardcoded namespace lists with intelligent recursive discovery that automatically finds all referenced types
  • Comprehensive Type Coverage: Expanded from 199 to 321 Kubernetes types through recursive discovery including versioned APIs (v1alpha1, v1beta1, v2)
  • Support for unversioned k8s types (e.g., RawExtension, IntOrString) placed in v0 directory to avoid conflicts
  • Reserved keyword escaping for field names starting with $ (like $ref, $schema) in generated Nickel code

Fixed

  • Required Field Usability Issue: Made all fields optional by default to enable gradual object construction (e.g., k8s.v1.LabelSelector & {} now works)
  • Cross-Package Import Resolution: Fixed imports to use full package IDs from manifest configuration instead of bare package names
  • Missing type references (e.g., RawExtension, NodeSelector) now properly discovered and generated
  • Cross-version imports for unversioned runtime types (v0 → other versions)
  • Syntax errors from unescaped special field names in JSON Schema types
  • Reserved keyword escaping in JSON object field names within default values

Changed

  • Breaking: All generated fields are now optional by default instead of required, enabling practical usage patterns
  • k8s type extraction now uses seed-based recursive discovery instead of fixed namespace lists
  • Updated to Kubernetes v1.33.4 schema version (from v1.31.0)
  • Unversioned types are placed in v0 to distinguish from versioned APIs
  • Enhanced import logic handles both v1 core types and v0 unversioned types
  • Package imports now use full package IDs like "github:seryl/nickel-pkgs/pkgs/k8s_io" for consistency

[0.6.2] - 2025-09-01

Fixed

  • Fixed missing cross-version imports in k8s_io packages (e.g., v1alpha1/v1beta1 types now properly import ObjectMeta, Condition, etc. from v1)
  • Consolidated duplicate handle_k8s_core_import implementations between lib.rs and main.rs

Added

  • Test suite for cross-version k8s type imports to prevent regression

[0.6.1] - 2025-09-01

Added

  • Smart workspace dependency management tool (workspace-deps) for switching between local and remote dependencies
  • Python-based version bump tool (version-bump) for reliable version management
  • Python-based tooling infrastructure in nix/packages/ for better maintainability
  • SmartError pattern for actionable error messages in development tools

Changed

  • Replaced hardcoded dev-mode script with dynamic workspace dependency discovery
  • Updated release script to use new Python-based version-bump tool
  • Improved development environment by removing unnecessary cargo tools (cargo-workspaces, cargo-release)
  • Fixed Python linting issues by using shell wrapper approach

Fixed

  • Test fixture inclusion in Nix builds with custom source filter
  • Workspace dependency publishing workflow for crates.io
  • Release script version bumping now works correctly

[0.6.0] - 2025-09-01

Changed

  • BREAKING: Changed package output directory from packages/ to pkgs/
  • BREAKING: Updated base package ID to github:seryl/nickel-pkgs/pkgs to match nickel-pkgs structure
  • Updated manifest to use latest stable versions: Kubernetes v1.33.4, Crossplane v2.0.2
  • Validation now exclusively uses the Nickel CLI binary, not the library
  • nickel-with-packages is exposed in flake for external users

Fixed

  • Removed nickel-lang-core dependency to avoid API instability issues
  • Simplified Nickel validation tests to use CLI instead of library API
  • Fixed compilation errors with external projects using amalgam
  • Enhanced fingerprinting to properly detect version changes in manifests
  • URL sources now properly lock to specific versions instead of tracking main branch

[0.5.1] - 2025-09-01

Added

  • --version flag to display amalgam version information
  • Workspace metadata for crane to silence warnings
  • CI can now be disabled via GitHub repository variable ENABLE_CI (disabled by default)

Fixed

  • Updated GitHub Actions to use latest versions (upload-artifact v4, install-nix-action v30)
  • Resolved all clippy warnings and formatting issues
  • Added missing k8s-openapi dependency to amalgam-daemon
  • Fixed redundant guards and iterator usage in dependency analyzer
  • Corrected file extension checking to use is_some_and

Changed

  • Made CLI command optional to support version flag without subcommand

[0.5.0] - 2025-09-01

Added

  • GitHub Actions workflow for Nix flake CI/CD with devshell integration
  • Support for --package-path option in validation command for local package resolution
  • Generic package dependency resolution using manifest-based registry
  • Git ref tracking for URL-based package sources (e.g., v1.17.2)
  • Package fingerprinting with SHA256 for intelligent change detection
  • Comprehensive test suite for manifest generation
  • Library interface (lib.rs) for better testing support

Changed

  • BREAKING: Package dependencies now always use Index format instead of Path
    • All packages reference upstream repositories via 'Index { package = "...", version = "..." }
    • Removed special-case handling for k8s_io, crossplane, and other packages
  • Updated reqwest from 0.11 to 0.12 (workspace-wide)
  • Improved manifest generation with metadata preservation in comments
  • Validation now uses absolute paths for nickel binary resolution
  • CI pipeline now uses ci-runner command from devshell

Fixed

  • Duplicate hyper and h2 dependency versions resolved
  • Reserved keyword escaping in generated Nickel code
  • Import resolution for k8s core types
  • Package validation with proper NICKEL_IMPORT_PATH environment variable

Removed

  • Hardcoded package ID generation for specific packages
  • Local Path dependencies in favor of Index dependencies
  • Unused test directories and duplicate files in examples/

[0.4.1] - 2025-08-28

Initial Features

  • Core amalgam functionality for generating Nickel packages from:
    • Kubernetes CRDs
    • OpenAPI specifications
    • Go type definitions
  • Manifest-based package generation system
  • Incremental compilation with fingerprinting
  • Support for complex type mappings and imports
  • Package validation with Nickel typecheck
  • Vendor management system