src/ppt/deck_generator.py

from __future__ import annotations

import json
import re
import tempfile
from pathlib import Path

from pptx import Presentation
from pptx.dml.color import RGBColor
from pptx.util import Pt

from src.ppt.placeholder_mapper import choose_layout_index
from src.utils.validators import validate_slides, validate_template


def _extract_json_array_block(raw_text: str) -> str:
    text = raw_text.strip()

    # Remove fenced markdown wrappers if present.
    text = re.sub(r"^```(?:json)?", "", text, flags=re.IGNORECASE).strip()
    text = re.sub(r"```$", "", text).strip()

    if text.startswith("[") and text.endswith("]"):
        return text

    match = re.search(r"\[[\s\S]*\]", text)
    if match:
        return match.group(0)

    raise ValueError("No JSON array block found in model output.")


def parse_slides_payload(slides_payload: str | list[dict]) -> list[dict]:
    if isinstance(slides_payload, list):
        parsed = slides_payload
    elif isinstance(slides_payload, str):
        block = _extract_json_array_block(slides_payload)
        try:
            parsed = json.loads(block)
        except json.JSONDecodeError as exc:
            raise ValueError(f"Malformed slide JSON: {exc}") from exc
    else:
        raise ValueError("slides_payload must be a JSON string or list of dict objects.")

    normalized = []
    for item in parsed:
        if not isinstance(item, dict):
            continue
        title = str(item.get("title", "Untitled Slide")).strip() or "Untitled Slide"
        bullets = item.get("bullets", item.get("content", []))
        if isinstance(bullets, str):
            bullets = [bullets]
        if not isinstance(bullets, list):
            bullets = []
        bullets = [str(b).strip() for b in bullets if str(b).strip()]
        layout_type = str(item.get("layout_type", "content_slide"))

        normalized.append(
            {
                "title": title,
                "bullets": bullets,
                "layout_type": layout_type,
            }
        )

    validate_slides(normalized)
    return normalized


def _apply_run_style(run, style_profile: dict):
    font_style = (style_profile or {}).get("font", {})
    if not font_style:
        return

    if font_style.get("name"):
        run.font.name = font_style["name"]
    if font_style.get("size_pt"):
        run.font.size = Pt(font_style["size_pt"])
    if font_style.get("bold") is not None:
        run.font.bold = bool(font_style["bold"])
    if font_style.get("italic") is not None:
        run.font.italic = bool(font_style["italic"])

    color = font_style.get("color_rgb") or (style_profile or {}).get("dominant_text_color")
    if color and isinstance(color, (list, tuple)) and len(color) == 3:
        run.font.color.rgb = RGBColor(int(color[0]), int(color[1]), int(color[2]))


def _remove_all_slides(prs: Presentation):
    # Remove slide relationship + slide ID to avoid duplicate part warnings on save.
    for idx in range(len(prs.slides) - 1, -1, -1):
        slide_id = prs.slides._sldIdLst[idx]  # pylint: disable=protected-access
        rel_id = slide_id.rId
        prs.part.drop_rel(rel_id)
        del prs.slides._sldIdLst[idx]  # pylint: disable=protected-access


def generate_presentation(
    slides_payload: str | list[dict],
    template_path: str | Path,
    output_path: str | Path | None = None,
    style_profile: dict | None = None,
) -> Presentation:
    validated_template = validate_template(template_path)
    slides = parse_slides_payload(slides_payload)

    prs = Presentation(str(validated_template))
    _remove_all_slides(prs)

    for slide_data in slides:
        layout_idx = choose_layout_index(
            slide_data.get("layout_type", "content_slide"),
            style_profile or {},
            len(prs.slide_layouts),
        )
        slide = prs.slides.add_slide(prs.slide_layouts[layout_idx])

        title_shape = slide.shapes.title
        if title_shape and title_shape.text_frame:
            title_shape.text_frame.clear()
            run = title_shape.text_frame.paragraphs[0].add_run()
            run.text = slide_data["title"]
            _apply_run_style(run, style_profile or {})

        body_shape = None
        for shape in slide.placeholders:
            if shape is not title_shape and getattr(shape, "has_text_frame", False):
                body_shape = shape
                break

        if body_shape and body_shape.text_frame:
            body_shape.text_frame.clear()
            for idx, bullet in enumerate(slide_data["bullets"]):
                paragraph = (
                    body_shape.text_frame.paragraphs[0]
                    if idx == 0
                    else body_shape.text_frame.add_paragraph()
                )
                run = paragraph.add_run()
                run.text = bullet
                _apply_run_style(run, style_profile or {})

    if output_path:
        prs.save(str(output_path))

    return prs


def compile_presentation(
    slides_payload: str | list[dict],
    template_path: str | Path,
    style_profile: dict | None = None,
    output_path: str | Path | None = None,
) -> str:
    destination = Path(output_path) if output_path else Path(tempfile.NamedTemporaryFile(suffix=".pptx", delete=False).name)
    generate_presentation(
        slides_payload=slides_payload,
        template_path=template_path,
        output_path=destination,
        style_profile=style_profile,
    )
    return str(destination)


def create_deck(template_data: dict, slide_data: str | list[dict], template_path: str | Path, output_path: str | Path | None = None):
    """Backward-compatible wrapper used by service layer."""
    return compile_presentation(
        slides_payload=slide_data,
        template_path=template_path,
        style_profile=template_data,
        output_path=output_path,
    )