"""
Output Formatter Module
=========================
Export datasets in multiple JSONL formats:
- OpenAI Chat JSONL
- Completion JSONL
- Classification JSONL
- Custom schema JSONL
"""

import json
from dataclasses import dataclass, field
from typing import List, Dict, Any, Optional
from pathlib import Path
import pandas as pd


@dataclass
class OutputFormatConfig:
    """Configuration for output formatting."""
    format_type: str = "openai_chat"  # "openai_chat", "completion", "classification", "custom"
    custom_schema: Dict[str, str] = field(default_factory=dict)
    # custom_schema maps output_key -> source_column, e.g. {"text": "instruction", "label": "category"}


def format_openai_chat(
    df: pd.DataFrame,
    system_prompt: str,
    instruction_col: str,
    output_col: str,
    input_col: Optional[str] = None,
) -> List[Dict[str, Any]]:
    """
    Format as OpenAI Chat JSONL.
    Each entry: {"messages": [{"role": "system", ...}, {"role": "user", ...}, {"role": "assistant", ...}]}
    """
    data = []
    for _, row in df.iterrows():
        messages = []
        if system_prompt:
            messages.append({"role": "system", "content": system_prompt})

        user_content = str(row[instruction_col])
        if input_col and input_col in df.columns:
            context = str(row.get(input_col, ''))
            if context and context != 'nan':
                user_content += f"\n\nContext: {context}"

        messages.append({"role": "user", "content": user_content})
        messages.append({"role": "assistant", "content": str(row[output_col])})

        data.append({"messages": messages})
    return data


def format_completion(
    df: pd.DataFrame,
    instruction_col: str,
    output_col: str,
) -> List[Dict[str, Any]]:
    """
    Format as Completion JSONL.
    Each entry: {"prompt": "...", "completion": "..."}
    """
    data = []
    for _, row in df.iterrows():
        data.append({
            "prompt": str(row[instruction_col]),
            "completion": str(row[output_col]),
        })
    return data


def format_classification(
    df: pd.DataFrame,
    text_col: str,
    label_col: str,
) -> List[Dict[str, Any]]:
    """
    Format as Classification JSONL.
    Each entry: {"text": "...", "label": "..."}
    """
    data = []
    for _, row in df.iterrows():
        data.append({
            "text": str(row[text_col]),
            "label": str(row[label_col]),
        })
    return data


def format_custom(
    df: pd.DataFrame,
    schema: Dict[str, str],
) -> List[Dict[str, Any]]:
    """
    Format using a custom schema.
    schema: dict mapping output_key -> source_column name
    """
    data = []
    for _, row in df.iterrows():
        entry = {}
        for out_key, src_col in schema.items():
            if src_col in df.columns:
                entry[out_key] = str(row[src_col])
            else:
                entry[out_key] = ""
        data.append(entry)
    return data


def export_jsonl(data: List[Dict[str, Any]], path: str) -> str:
    """Write a list of dicts as JSONL to a file."""
    output_path = Path(path)
    output_path.parent.mkdir(parents=True, exist_ok=True)

    with open(output_path, 'w', encoding='utf-8') as f:
        for entry in data:
            f.write(json.dumps(entry, ensure_ascii=False) + '\n')

    return str(output_path)


def generate_preview(data: List[Dict[str, Any]], n: int = 3) -> str:
    """Return a pretty-printed JSON string of the first n entries."""
    return json.dumps(data[:n], indent=2, ensure_ascii=False)


def format_dataset(
    df: pd.DataFrame,
    config: OutputFormatConfig,
    system_prompt: str = "",
    instruction_col: str = "",
    output_col: str = "",
    input_col: Optional[str] = None,
    label_col: Optional[str] = None,
) -> List[Dict[str, Any]]:
    """Format the dataset according to the configured format type."""
    if config.format_type == "openai_chat":
        return format_openai_chat(df, system_prompt, instruction_col, output_col, input_col)
    elif config.format_type == "completion":
        return format_completion(df, instruction_col, output_col)
    elif config.format_type == "classification":
        text_col = instruction_col or (list(df.columns)[0] if len(df.columns) > 0 else "")
        lbl_col = label_col or output_col
        return format_classification(df, text_col, lbl_col)
    elif config.format_type == "custom":
        return format_custom(df, config.custom_schema)
    else:
        return format_openai_chat(df, system_prompt, instruction_col, output_col, input_col)