experiments/evaluate.py

"""Evaluate one hypothesis with the existing deterministic symbolic scorer."""

from __future__ import annotations

import argparse
import json
from pathlib import Path
import sys

WORKSPACE_ROOT = Path(__file__).resolve().parent.parent
if str(WORKSPACE_ROOT) not in sys.path:
    sys.path.insert(0, str(WORKSPACE_ROOT))

from experiments import config  # noqa: E402
from symbolic import launch as symbolic_launch  # noqa: E402


def configure_symbolic_evaluation(
    hypothesis,
    depth="metric",
    tracking="tracking",
    input_selection="uniform",
    frame_count=64,
    spatial_code_format="explicit",
):
    """Point symbolic reads and writes at one isolated experiment selection."""
    codes = config.spatial_code_directory(
        hypothesis, depth, tracking, input_selection, frame_count, spatial_code_format
    )
    results = config.result_directory(
        hypothesis,
        "symbolic",
        depth,
        tracking,
        input_selection,
        frame_count,
        spatial_code_format,
    )
    symbolic_run = symbolic_launch.symbolic_run
    symbolic_run.SPATIAL_CODES_DEPTH = depth
    symbolic_run.SPATIAL_CODES_INPUT = input_selection
    symbolic_run.SPATIAL_CODES_TRACKING = tracking
    symbolic_run.SPATIAL_CODES_FRAMES = frame_count
    symbolic_run.SPATIAL_CODES_FORMAT = spatial_code_format
    symbolic_run.SPATIAL_CODES_DIR = str(codes)
    symbolic_run.RESULTS_DIR = str(results)
    symbolic_run.results_dir_for_selection = lambda results_dir=None: str(
        results_dir or results
    )
    return codes, results


def evaluate(
    hypothesis,
    depth="metric",
    tracking="tracking",
    input_selection="uniform",
    frame_count=64,
    scene_ids=None,
    quiet=False,
    errors=False,
    spatial_code_format="explicit",
):
    """Score every available experiment code, or an explicit scene subset."""
    codes, results = configure_symbolic_evaluation(
        hypothesis,
        depth,
        tracking,
        input_selection,
        frame_count,
        spatial_code_format,
    )
    available = symbolic_launch.scenes_with_spatial_codes()
    selected = available if scene_ids is None else list(scene_ids)
    missing = [scene for scene in selected if scene not in available]
    if missing:
        raise FileNotFoundError(
            f"scene(s) have no hypothesis spatial code under {codes}: {missing}"
        )
    if not selected:
        raise FileNotFoundError(f"no hypothesis spatial codes found under {codes}")
    per_scene, combined = symbolic_launch.run_all(selected, quiet=quiet)
    summary = {
        "hypothesis": hypothesis,
        "depth": depth,
        "input": input_selection,
        "tracking": tracking,
        "frames": frame_count,
        "spatial_code_format": spatial_code_format,
        "scenes_run": list(per_scene),
        "combined_aggregate": combined,
    }
    if errors:
        summary["error_analysis"] = {
            question_type: symbolic_launch.error_analysis(per_scene, question_type)
            for question_type in symbolic_launch._ANALYZABLE_TYPES
        }
        symbolic_launch.print_error_analysis(per_scene)
        symbolic_launch.print_mca_breakdown(per_scene)
    results.mkdir(parents=True, exist_ok=True)
    summary_path = results / "_summary.json"
    with summary_path.open("w", encoding="utf-8") as stream:
        json.dump(summary, stream, indent=1)
    return summary, summary_path


def main() -> None:
    parser = argparse.ArgumentParser()
    parser.add_argument("--hypothesis", required=True)
    parser.add_argument("--depth", default="metric", choices=("relative", "metric"))
    parser.add_argument(
        "--input",
        default="uniform",
        choices=("uniform", "selective"),
        dest="input_selection",
    )
    parser.add_argument(
        "--tracking", default="tracking", choices=("tracking", "no tracking")
    )
    parser.add_argument("--frames", type=int, default=64)
    parser.add_argument(
        "--format",
        default="explicit",
        choices=config.SPATIAL_CODE_FORMATS,
        dest="spatial_code_format",
    )
    parser.add_argument(
        "--scenes", default="", help="optional comma-separated scene IDs"
    )
    parser.add_argument("--quiet", action="store_true")
    parser.add_argument("--errors", action="store_true")
    args = parser.parse_args()
    if args.frames < 1:
        parser.error("--frames must be positive")
    scenes = (
        [scene.strip() for scene in args.scenes.split(",") if scene.strip()]
        if args.scenes
        else None
    )
    summary, path = evaluate(
        args.hypothesis,
        args.depth,
        args.tracking,
        args.input_selection,
        args.frames,
        scenes,
        args.quiet,
        args.errors,
        args.spatial_code_format,
    )
    print("\nCOMBINED AGGREGATE")
    for key, value in summary["combined_aggregate"].items():
        print(f"  {key}: {value}")
    print(f"\nwrote experiment summary to {path}")


if __name__ == "__main__":
    main()