scripts/make_jsonl.py

import csv
import json
import os

METADATA_DIR = "metadata"
OUTPUT_JSONL = "data/test.jsonl"


def read_csv(path):
    with open(path, "r", encoding="utf-8") as f:
        return list(csv.DictReader(f))


def build_metadata():
    entries = []

    # --- Task_Color_Var1 ---
    path1 = os.path.join(METADATA_DIR, "Task_Color_Var1_metadata.csv")
    if os.path.exists(path1):
        for row in read_csv(path1):
            id_val = row["id"].strip()
            entries.append({
                "id": f"id_{id_val}",
                "prompt": row.get("prompt", "").strip(),
                # "split": row.get("split", "test"),
                "color_1": row.get("color_1", ""),
                "hex_val_1": row.get("hex_val_1", ""),
                "task": 1,
                "ground_truth": f"Task_Color_Var1/id_{id_val}.png",
            })
    else:
        print(f"Warning: {path1} not found, skipping Task_Color_Var1")

    # --- Task_Color_Var2 ---
    path2 = os.path.join(METADATA_DIR, "Task_Color_Var2_metadata.csv")
    if os.path.exists(path2):
        for row in read_csv(path2):
            id_val = row["id"].strip()
            entries.append({
                "id": f"id_{id_val}",
                "prompt": row.get("prompt", "").strip(),
                # "split": row.get("split", "test"),
                "color_1": row.get("color_1", ""),
                "color_2": row.get("color_2", ""),
                "hex_val_1": row.get("hex_val_1", ""),
                "hex_val_2": row.get("hex_val_2", ""),
                "direction": row.get("direction", ""),
                "task": 2,
                "ground_truth": f"Task_Color_Var2/id_{id_val}.png",
            })
    else:
        print(f"Warning: {path2} not found, skipping Task_Color_Var2")

    # --- Task_Geometric ---
    path3 = os.path.join(METADATA_DIR, "Task_Geometric_metadata.csv")
    if os.path.exists(path3):
        for row in read_csv(path3):
            id_val = row["id"].strip()
            entries.append({
                "id": f"id_{id_val}",
                "prompt": row.get("prompt", "").strip(),
                "shape": row.get("shape", ""),
                "position": row.get("position", ""),
                "size_ratio": row.get("size_ratio", ""),
                "center_x": row.get("center_x", ""),
                "center_y": row.get("center_y", ""),
                "task": 3,
                "ground_truth": f"Task_Geometric/id_{id_val}.png",
            })
    else:
        print(f"Warning: {path3} not found, skipping Task_Geometric")

    # --- Task_Image_Mask ---
    for mask_type in ["inpainting", "outpainting", "random"]:
        path4 = os.path.join(METADATA_DIR, f"Task_Image_Mask_{mask_type}_metadata.csv")
        if not os.path.exists(path4):
            print(f"Warning: {path4} not found, skipping Task_Image_Mask {mask_type}")
            continue
        for row in read_csv(path4):
            id_val = row["id"].strip()
            img1 = row.get("image1_path", "").strip().replace('\\', '/')
            img2 = row.get("image2_path", "").strip().replace('\\', '/')
            entries.append({
                "id": f"id_{id_val}",
                "prompt": row.get("prompt", "").strip(),
                "mask_type": mask_type,
                "image_id": row.get("image_id", ""),
                "image1_path": img1,
                "image2_path": img2,
                "task": 4,
                "ground_truth": f"Task_Image_Mask/{mask_type}/id_{id_val}.png",
            })

    os.makedirs(os.path.dirname(OUTPUT_JSONL), exist_ok=True)
    with open(OUTPUT_JSONL, "w", encoding="utf-8") as f:
        for entry in entries:
            f.write(json.dumps(entry, ensure_ascii=False) + "\n")

    print(f"Total {len(entries)} entries → {OUTPUT_JSONL}")


if __name__ == "__main__":
    build_metadata()