#!/usr/bin/env python3
"""Produce a heuristic human-review queue; flags are not translation judgments."""
from __future__ import annotations

import argparse
import csv
import re
from collections import Counter, defaultdict
from pathlib import Path

from verify_output import DEFAULT_BASELINE, DEFAULT_SOURCE, TRANSLATION, audit, read_baseline


def review(rows, baseline):
    duplicates = defaultdict(list)
    for key, text in rows.items():
        duplicates[(key[0], key[1], text.strip())].append(key[2])
    for key, text in sorted(rows.items()):
        flags = []
        original = baseline.get(key, '')
        if not re.search(r'[\u3040-\u30ff\u3400-\u9fff]', text):
            flags.append('no-japanese')
        if re.search(r'[\u0370-\u03ff\u0590-\u05ff\u1f00-\u1fff]', text):
            flags.append('source-script')
        if re.search(r'[A-Za-z]{2,}', text):
            flags.append('latin-text')
        if re.search(r'TODO|TBD|未翻訳|翻訳できません|翻訳は以下|翻訳結果|訳文：', text, re.I):
            flags.append('placeholder-or-commentary')
        if re.search(r'(.{8,}?)\1{2,}', text):
            flags.append('repeated-phrase')
        if len(duplicates[(key[0], key[1], text.strip())]) > 1:
            flags.append('identical-within-chapter')
        if len(text) < 5:
            flags.append('very-short')
        if original and len(original) >= 30 and len(text) < len(original) * 0.35:
            flags.append('short-vs-kougo')
        if original and len(text) > max(100, len(original) * 3):
            flags.append('long-vs-kougo')
        if flags:
            yield {'book': key[0], 'chapter': key[1], 'verse': key[2],
                   'flags': ';'.join(flags), 'text': text, 'kougo_text': original}


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument('--source', type=Path, default=DEFAULT_SOURCE)
    parser.add_argument('--baseline', type=Path, default=DEFAULT_BASELINE)
    parser.add_argument('--report', type=Path, required=True, help='Review CSV destination (outside source)')
    args = parser.parse_args()
    if args.report.resolve().is_relative_to(args.source.resolve()) or args.report.resolve() == args.baseline.resolve():
        parser.error('report must not overwrite source data or baseline')
    try:
        baseline = read_baseline(args.baseline)
        structural, rows = audit(args.source, baseline)
        queue = list(review(rows, baseline))
        args.report.parent.mkdir(parents=True, exist_ok=True)
        with args.report.open('w', encoding='utf-8', newline='') as stream:
            writer = csv.DictWriter(stream, fieldnames=['book', 'chapter', 'verse', 'flags', 'text', 'kougo_text'])
            writer.writeheader()
            writer.writerows(queue)
    except (OSError, ValueError, csv.Error) as error:
        parser.exit(2, f'Cannot review: {error}\n')
    print(f'{TRANSLATION}: {len(queue)} verses flagged for human review')
    for flag, count in sorted(Counter(flag for row in queue for flag in row['flags'].split(';')).items()):
        print(f'  {flag}: {count}')
    print(f'Structural issues: {len(structural["issues"])} (run verify_output.py for details)')
    print(f'Report: {args.report}')
    return 0 if structural['passed'] else 1


if __name__ == '__main__':
    raise SystemExit(main())