#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""把 inference_sentences_*.json 切成 chunk_{i}.json (每块 chunk_size 句), 供分块 analysis.
用法: split_sentences.py <inference_file> <outdir> <chunk_size>"""
import json, sys
from pathlib import Path

def main(infile, outdir, chunk_size):
    d = json.loads(Path(infile).read_text(encoding="utf-8"))
    sents = d["sentences"]
    outdir = Path(outdir); outdir.mkdir(parents=True, exist_ok=True)
    meta = {k: v for k, v in d.items() if k != "sentences"}
    n = 0
    for i in range(0, len(sents), chunk_size):
        chunk = sents[i:i + chunk_size]
        out = dict(meta); out["sentences"] = chunk; out["num_sentences"] = len(chunk)
        (outdir / f"chunk_{n:03d}.json").write_text(
            json.dumps(out, ensure_ascii=False, separators=(",", ":")), encoding="utf-8")
        n += 1
    print(f"split {len(sents)} sentences -> {n} chunks of {chunk_size} in {outdir}", flush=True)

if __name__ == "__main__":
    main(sys.argv[1], sys.argv[2], int(sys.argv[3]))