"""Model-agnostic token-efficiency accounting for benchmark artifacts.

The harness does not assume a tokenizer or provider. It reports a reproducible
upper-bound-like estimate of one token per four UTF-8 bytes so reports can be
compared across runs without adding a model-specific dependency.
"""

from __future__ import annotations

import math
from collections.abc import Iterable, Mapping
from typing import Any

from .metrics import distribution_summary

TOKEN_ESTIMATOR = "ceil(utf8_bytes/4)"


def estimate_text_tokens(text: str) -> int:
    """Return a deterministic, model-agnostic estimate for UTF-8 text."""

    return math.ceil(len(text.encode("utf-8")) / 4)


def _metric_summary(bytes_samples: list[int], findings: list[int]) -> dict[str, Any]:
    tokens = [math.ceil(value / 4) for value in bytes_samples]
    density = [
        1000.0 * count / token
        for count, token in zip(findings, tokens, strict=True)
        if token > 0
    ]
    return {
        "bytes": distribution_summary(bytes_samples),
        "estimated_tokens": distribution_summary(tokens),
        "findings_per_1k_estimated_tokens": distribution_summary(density),
    }


def _nonnegative_int(record: Mapping[str, Any], key: str) -> int:
    value = record.get(key, 0)
    if isinstance(value, bool) or not isinstance(value, (int, float)):
        return 0
    try:
        return max(0, int(value))
    except (OverflowError, ValueError):
        return 0


def artifact_token_efficiency(runs: Iterable[Mapping[str, Any]]) -> dict[str, Any]:
    """Summarize driver and SARIF artifact token costs for completed runs."""

    records = list(runs)
    findings = [_nonnegative_int(record, "finding_count") for record in records]
    return {
        "estimator": TOKEN_ESTIMATOR,
        "driver_output": _metric_summary(
            [_nonnegative_int(record, "output_bytes") for record in records], findings
        ),
        "sarif_report": _metric_summary(
            [_nonnegative_int(record, "report_bytes") for record in records], findings
        ),
    }