#!/usr/bin/env python3
import re
import argparse
import pandas as pd
from pathlib import Path

LINE_RE = re.compile(
    r"""
    ^\s*Debug\s+tet_id=(?P<tet_id>\d+)\s+face_id=(?P<face_id>\d+)\s*\|\s*
    face0=\((?P<face0>[^)]*)\)\s+
    face1=\((?P<face1>[^)]*)\)\s+
    face2=\((?P<face2>[^)]*)\)\s*\|\s*
    tet0=\((?P<tet0>[^)]*)\)\s+
    tet1=\((?P<tet1>[^)]*)\)\s+
    tet2=\((?P<tet2>[^)]*)\)\s+
    tet3=\((?P<tet3>[^)]*)\)\s*$
    """,
    re.VERBOSE
)

def parse_triplet(s: str):
    # s like "9.249584e-03,-5.663737e-19,9.812500e-02"
    parts = [p.strip() for p in s.split(",")]
    if len(parts) != 3:
        raise ValueError(f"Expected 3 components, got {len(parts)} in '{s}'")
    return tuple(float(p) for p in parts)

def parse_log(path: Path):
    rows = []
    with path.open("r", encoding="utf-8", errors="ignore") as f:
        for ln, line in enumerate(f, 1):
            m = LINE_RE.match(line)
            if not m:
                continue  # skip non-matching lines
            try:
                tet_id  = int(m.group("tet_id"))
                face_id = int(m.group("face_id"))
                f0 = parse_triplet(m.group("face0"))
                f1 = parse_triplet(m.group("face1"))
                f2 = parse_triplet(m.group("face2"))
                t0 = parse_triplet(m.group("tet0"))
                t1 = parse_triplet(m.group("tet1"))
                t2 = parse_triplet(m.group("tet2"))
                t3 = parse_triplet(m.group("tet3"))
            except Exception as e:
                raise ValueError(f"Parse error on line {ln}: {e}") from e

            rows.append({
                "tet_id": tet_id, "face_id": face_id,
                "face0_x": f0[0], "face0_y": f0[1], "face0_z": f0[2],
                "face1_x": f1[0], "face1_y": f1[1], "face1_z": f1[2],
                "face2_x": f2[0], "face2_y": f2[1], "face2_z": f2[2],
                "tet0_x": t0[0], "tet0_y": t0[1], "tet0_z": t0[2],
                "tet1_x": t1[0], "tet1_y": t1[1], "tet1_z": t1[2],
                "tet2_x": t2[0], "tet2_y": t2[1], "tet2_z": t2[2],
                "tet3_x": t3[0], "tet3_y": t3[1], "tet3_z": t3[2],
            })
    return pd.DataFrame(rows)

def to_long(df):
    # explode wide columns into long form: one row per node (face0..2, tet0..3)
    long_rows = []
    for _, r in df.iterrows():
        for kind in ["face0","face1","face2","tet0","tet1","tet2","tet3"]:
            long_rows.append({
                "tet_id": r["tet_id"],
                "face_id": r["face_id"],
                "node_kind": kind,
                "x": r[f"{kind}_x"],
                "y": r[f"{kind}_y"],
                "z": r[f"{kind}_z"],
            })
    return pd.DataFrame(long_rows)

def main():
    ap = argparse.ArgumentParser(description="Parse CUDA printf debug lines into CSV")
    ap.add_argument("--log", required=True, help="Path to text file containing the 'Debug tet_id=...' lines")
    ap.add_argument("--out-prefix", default="parsed_debug", help="Output file prefix (no extension)")
    args = ap.parse_args()

    log_path = Path(args.log)
    df = parse_log(log_path)
    if df.empty:
        print("No matching debug lines found.")
        return

    out_wide = Path(f"{args.out_prefix}_wide.csv")
    out_long = Path(f"{args.out_prefix}_long.csv")

    df.to_csv(out_wide, index=False)
    to_long(df).to_csv(out_long, index=False)

    print(f"Parsed {len(df)} tet/face entries")
    print(f"Saved wide CSV to: {out_wide}")
    print(f"Saved long CSV to: {out_long}")

if __name__ == "__main__":
    main()