#!/usr/bin/env python3
# -----------------------------------------------------------------------------
#  validate_catch22.py                          MMQ - Muhammad Minhas Qamar
#
#  Reference cross-check for the MQL5 CCatch22 engine.
#
#  The MQL5 validation script (Catch22Validate.mq5) writes the *exact* window
#  it evaluated to  MQL5\Files\Catch22\input_<tag>.csv  (one value per line).
#  This tool loads that identical vector, runs the canonical pycatch22
#  reference implementation on it, and prints the 22 features in the same
#  canonical order the MQL5 engine uses. Paste the two side by side (or use
#  --mql5 to auto-diff) to confirm the port.
#
#  Usage:
#     python validate_catch22.py                       # uses input_fixture.csv
#     python validate_catch22.py --tag returns
#     python validate_catch22.py --csv path\to\vector.csv
#     python validate_catch22.py --mql5 mql5_output.txt # diff against MQL5 log
#
#  Requires:  pip install pycatch22 numpy
# -----------------------------------------------------------------------------
import argparse
import os
import re
import sys

import numpy as np
import pycatch22

# Canonical order used by ENUM_CATCH22 in Catch22.mqh. pycatch22 returns its
# own order via catch22_all(); we remap into this list so column i here equals
# column i in the MQL5 output.
CANONICAL = [
    "DN_HistogramMode_5",
    "DN_HistogramMode_10",
    "CO_f1ecac",
    "CO_FirstMin_ac",
    "CO_HistogramAMI_even_2_5",
    "CO_trev_1_num",
    "MD_hrv_classic_pnn40",
    "SB_BinaryStats_mean_longstretch1",
    "SB_TransitionMatrix_3ac_sumdiagcov",
    "PD_PeriodicityWang_th0_01",
    "CO_Embed2_Dist_tau_d_expfit_meandiff",
    "IN_AutoMutualInfoStats_40_gaussian_fmmi",
    "FC_LocalSimple_mean1_tauresrat",
    "DN_OutlierInclude_p_001_mdrmd",
    "DN_OutlierInclude_n_001_mdrmd",
    "SP_Summaries_welch_rect_area_5_1",
    "SB_BinaryStats_diff_longstretch0",
    "SB_MotifThree_quantile_hh",
    "SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1",
    "SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1",
    "SP_Summaries_welch_rect_centroid",
    "FC_LocalSimple_mean3_stderr",
]

# Default location of the MQL5 sandbox Files directory on this machine.
FILES_DIR = os.path.join(
    os.environ.get("APPDATA", ""),
    r"MetaQuotes\Terminal\D0E8209F77C8CF37AD8BF550E51FF075\MQL5\Files\Catch22",
)


def load_vector(path):
    """Load a one-value-per-line CSV written by the MQL5 script."""
    vals = []
    with open(path, "r") as fh:
        for line in fh:
            line = line.strip().rstrip(",")
            if line:
                vals.append(float(line))
    return np.asarray(vals, dtype=float)


def reference_features(x):
    """Run pycatch22 and return an ordered dict in CANONICAL order."""
    res = pycatch22.catch22_all(x, catch24=False)
    by_name = dict(zip(res["names"], res["values"]))
    out = []
    for name in CANONICAL:
        out.append(by_name.get(name, float("nan")))
    return out


def parse_mql5_log(path):
    """Extract 'NN  feature_name = value' lines from a pasted MQL5 log."""
    got = {}
    pat = re.compile(r"\d+\s+([A-Za-z0-9_]+)\s*=\s*(-?\d+\.?\d*(?:[eE][-+]?\d+)?)")
    with open(path, "r") as fh:
        for line in fh:
            m = pat.search(line)
            if m:
                got[m.group(1)] = float(m.group(2))
    return got


def main():
    ap = argparse.ArgumentParser(description="pycatch22 reference cross-check")
    ap.add_argument("--tag", default="fixture", help="input_<tag>.csv in Files\\Catch22")
    ap.add_argument("--csv", default=None, help="explicit path to the input vector")
    ap.add_argument("--mql5", default=None, help="MQL5 log file to auto-diff against")
    args = ap.parse_args()

    csv_path = args.csv or os.path.join(FILES_DIR, f"input_{args.tag}.csv")
    if not os.path.isfile(csv_path):
        sys.exit(f"input vector not found: {csv_path}")

    x = load_vector(csv_path)
    print(f"loaded {len(x)} values from {csv_path}\n")

    ref = reference_features(x)
    mql5 = parse_mql5_log(args.mql5) if args.mql5 else None

    if mql5:
        hdr = f"{'#':>2}  {'feature':<46} {'pycatch22':>14} {'mql5':>14} {'abs.diff':>12}"
    else:
        hdr = f"{'#':>2}  {'feature':<46} {'pycatch22':>14}"
    print(hdr)
    print("-" * len(hdr))

    max_rel = 0.0
    for i, name in enumerate(CANONICAL):
        rv = ref[i]
        if mql5 is not None:
            mv = mql5.get(name, float("nan"))
            diff = abs(rv - mv)
            denom = max(1e-9, abs(rv))
            max_rel = max(max_rel, diff / denom)
            flag = "" if diff / denom < 0.02 else "  <-- CHECK"
            print(f"{i:>2}  {name:<46} {rv:>14.8f} {mv:>14.8f} {diff:>12.6f}{flag}")
        else:
            print(f"{i:>2}  {name:<46} {rv:>14.8f}")

    if mql5 is not None:
        print(f"\nmax relative deviation: {max_rel:.4%}")


if __name__ == "__main__":
    main()