src/cl_macros/schema.py

"""Dataset schema for CL macro transformation pairs — using stdlib dataclasses."""

from __future__ import annotations

import json
from dataclasses import dataclass, field
from enum import Enum
from typing import Optional


class MacroCategory(str, Enum):
    CONTROL_FLOW = "control-flow"
    DSL = "dsl"
    ANAPHORIC = "anaphoric"
    CAPTURE_MANAGEMENT = "capture-management"
    CODE_WALKING = "code-walking"
    RECURSIVE_EXPANSION = "recursive-expansion"
    DISPATCH = "dispatch"
    READ_MACRO = "read-macro"
    INDIRECTION = "indirection"
    SCOPE = "scope"
    EFFICIENCY = "efficiency"
    COMPILER_MACRO = "compiler-macro"


class Complexity(str, Enum):
    BASIC = "basic"
    INTERMEDIATE = "intermediate"
    ADVANCED = "advanced"


class Source(str, Enum):
    LET_OVER_LAMBDA = "let-over-lambda"
    SYNTHETIC = "synthetic"
    CL_UTILITIES = "cl-utilities"
    ALEXANDRIA = "alexandria"
    ON_LISP = "on-lisp"
    SERAPEUM = "serapeum"
    ITERATE = "iterate"
    ANAPHORA = "anaphora"
    TRIVIA = "trivia"
    ARROW_MACROS = "arrow-macros"
    MODF = "modf"
    ACCESS = "access"
    FOR = "for"
    CL_INTERPOL = "cl-interpol"
    SCREAMER = "screamer"
    COALTON = "coalton"
    NHOOKS = "nhooks"
    GENERIC_CL = "generic-cl"
    OTHER_LIBRARY = "other-library"


class MacroTechnique(str, Enum):
    NESTED_BACKQUOTE = "nested-backquote"
    GENSYM = "gensym"
    ONCE_ONLY = "once-only"
    RECURSIVE_EXPANSION = "recursive-expansion"
    CODE_WALKING = "code-walking"
    DLAMBDA = "dlambda"
    ANAPHOR = "anaphor"
    READER = "reader"
    COMPILER_MACRO = "compiler-macro"
    MACROLET = "macrolet"
    SYMBOL_MACROLET = "symbol-macrolet"
    DEFSETF = "defsetf"
    TAGBODY = "tagbody-go"
    PARSER = "parser"


@dataclass
class TransformationExample:
    """A single macro transformation pair for fine-tuning."""

    id: str
    before_code: str
    problem_pattern: str
    macro_definition: str
    after_expansion: str
    macro_category: MacroCategory
    technique: list[MacroTechnique] = field(default_factory=list)
    source: Source = Source.LET_OVER_LAMBDA
    source_chapter: Optional[str] = None
    complexity: Complexity = Complexity.BASIC
    has_capture_risk: bool = False
    requires_gensyms: bool = False
    library_name: str = ""
    macro_name: str = ""
    call_sites: list[str] = field(default_factory=list)
    macroexpand_1_result: Optional[str] = None
    is_verified: bool = False
    formulation: str = "macro-from-usage"
    nl_description: Optional[str] = None
    is_synthetic: bool = False
    quality_score: Optional[float] = None
    commentary: Optional[str] = None


@dataclass
class DatasetRecord:
    """Full dataset record with instructions for fine-tuning."""

    instruction: str
    input: str
    output: str
    category: str = ""
    technique: str = ""
    complexity: str = ""
    quality_score: Optional[float] = None

    def model_dump_json(self) -> str:
        return json.dumps(
            {
                "instruction": self.instruction,
                "input": self.input,
                "output": self.output,
                "category": self.category,
                "technique": self.technique,
                "complexity": self.complexity,
                "quality_score": self.quality_score,
            }
        )


@dataclass
class RawExtractionRecord:
    """Intermediate record produced by Phase 1 extraction."""
    id: str
    library_name: str
    system_name: str = ""
    source_file: str = ""
    macro_name: str = ""
    macro_definition: str = ""
    form_type: str = "defmacro"
    docstring: Optional[str] = None
    args: str = ""
    extracted_at: str = ""
    status: str = "extracted"