OCR/docengine/app/services/template_service.py

from __future__ import annotations

import uuid
from typing import Any

from sqlalchemy.orm import Session

from app.core.logging_config import get_logger
from app.models.document import Document
from app.models.template import (
    DocumentCell,
    DocumentFormat,
    DocumentRegion,
    ImageRegion,
    TableColumn,
    TableFormat,
    TableRow,
    Watermark,
)
from app.repositories.document_repository import DocumentRepository
from app.repositories.template_repository import (
    DocumentCellRepository,
    DocumentRegionRepository,
    ImageRegionRepository,
    TableColumnRepository,
    TableFormatRepository,
    TableRowRepository,
    TemplateRepository,
    WatermarkRepository,
)
from app.services.fingerprint_service import FingerprintService

logger = get_logger(__name__)


class TemplateService:
    """Generate reusable document templates from processed documents."""

    def __init__(self, db: Session) -> None:
        self.db = db
        self.doc_repo = DocumentRepository(db)
        self.template_repo = TemplateRepository(db)
        self.cell_repo = DocumentCellRepository(db)
        self.region_repo = DocumentRegionRepository(db)
        self.table_format_repo = TableFormatRepository(db)
        self.table_column_repo = TableColumnRepository(db)
        self.table_row_repo = TableRowRepository(db)
        self.watermark_repo = WatermarkRepository(db)
        self.image_region_repo = ImageRegionRepository(db)
        self.fingerprint_service = FingerprintService(db)

    def generate_template(
        self,
        document: Document,
        user_id: uuid.UUID | None = None,
    ) -> DocumentFormat:
        """Generate a reusable template from a processed document."""
        if not document.pages:
            raise ValueError(f"Document '{document.id}' has no processed pages")

        # Check if template already exists for this document
        existing = self.template_repo.get_by_source_document(document.id)
        if existing:
            logger.info(
                "template_already_exists",
                document_id=str(document.id),
                template_id=str(existing.id),
            )
            return existing

        first_page = document.pages[0]
        template_name = f"Template_{document.original_filename}_{uuid.uuid4().hex[:8]}"

        # Create template
        template = self.template_repo.create_template(
            name=template_name,
            page_width=first_page.width,
            page_height=first_page.height,
            page_count=document.page_count or len(document.pages),
            description=f"Auto-generated template from {document.original_filename}",
            source_document_id=document.id,
            created_by=user_id,
        )

        # Process each page
        for page in document.pages:
            self._process_page_for_template(template, page)

        # Generate fingerprint
        self.fingerprint_service.generate_fingerprint(template)

        logger.info(
            "template_generated",
            template_id=str(template.id),
            document_id=str(document.id),
            cells=len(template.cells),
            regions=len(template.regions),
        )

        return template

    def _process_page_for_template(
        self,
        template: DocumentFormat,
        page: Any,
    ) -> None:
        """Process a document page and create template components."""
        page_number = page.page_number

        # Create cells from text blocks
        self._create_cells_from_text_blocks(template, page, page_number)

        # Create regions from headers, footers
        self._create_regions(template, page, page_number)

        # Create table formats
        self._create_table_formats(template, page, page_number)

        # Create image regions
        self._create_image_regions(template, page, page_number)

        # Detect watermarks from text blocks
        self._create_watermarks(template, page, page_number)

    def _create_cells_from_text_blocks(
        self,
        template: DocumentFormat,
        page: Any,
        page_number: int,
    ) -> None:
        """Create template cells from extracted text blocks."""
        for seq, block in enumerate(page.text_blocks):
            if block.block_type in ("header", "footer", "watermark"):
                continue

            # Determine if this is a dynamic field
            is_dynamic = self._is_dynamic_field(block.text)
            field_name = self._generate_field_name(block.text, seq) if is_dynamic else None

            self.cell_repo.create_cell(
                format_id=template.id,
                page_number=page_number,
                x=block.x,
                y=block.y,
                width=block.width,
                height=block.height,
                data_type=self._infer_data_type(block.text),
                font_family=block.font_family,
                font_size=block.font_size,
                font_style=block.font_style,
                font_color=block.font_color,
                alignment=self._infer_alignment(block.x, template.page_width),
                static_text=block.text if not is_dynamic else None,
                field_name=field_name,
                sequence=seq,
                is_dynamic=is_dynamic,
            )

    def _create_regions(
        self,
        template: DocumentFormat,
        page: Any,
        page_number: int,
    ) -> None:
        """Create template regions from headers and footers."""
        header_blocks = [b for b in page.text_blocks if b.block_type == "header"]
        footer_blocks = [b for b in page.text_blocks if b.block_type == "footer"]

        if header_blocks:
            # Compute bounding box for all header blocks
            min_x = min(b.x for b in header_blocks)
            min_y = min(b.y for b in header_blocks)
            max_x = max(b.x + b.width for b in header_blocks)
            max_y = max(b.y + b.height for b in header_blocks)

            content = {
                "blocks": [
                    {
                        "text": b.text,
                        "x": b.x,
                        "y": b.y,
                        "width": b.width,
                        "height": b.height,
                        "font_family": b.font_family,
                        "font_size": b.font_size,
                    }
                    for b in header_blocks
                ]
            }

            self.region_repo.create_region(
                format_id=template.id,
                page_number=page_number,
                region_type="header",
                x=min_x,
                y=min_y,
                width=max_x - min_x,
                height=max_y - min_y,
                content=content,
                sequence=0,
            )

        if footer_blocks:
            min_x = min(b.x for b in footer_blocks)
            min_y = min(b.y for b in footer_blocks)
            max_x = max(b.x + b.width for b in footer_blocks)
            max_y = max(b.y + b.height for b in footer_blocks)

            content = {
                "blocks": [
                    {
                        "text": b.text,
                        "x": b.x,
                        "y": b.y,
                        "width": b.width,
                        "height": b.height,
                        "font_family": b.font_family,
                        "font_size": b.font_size,
                    }
                    for b in footer_blocks
                ]
            }

            self.region_repo.create_region(
                format_id=template.id,
                page_number=page_number,
                region_type="footer",
                x=min_x,
                y=min_y,
                width=max_x - min_x,
                height=max_y - min_y,
                content=content,
                sequence=1,
            )

    def _create_table_formats(
        self,
        template: DocumentFormat,
        page: Any,
        page_number: int,
    ) -> None:
        """Create table format definitions from detected tables."""
        for table in page.tables:
            table_format = self.table_format_repo.create_table_format(
                format_id=template.id,
                page_number=page_number,
                x=table.x,
                y=table.y,
                width=table.width,
                height=table.height,
                rows=table.rows,
                columns=table.columns,
            )

            # Create columns
            col_width = table.width / max(table.columns, 1)
            for col_idx in range(table.columns):
                self.table_column_repo.create_column(
                    table_format_id=table_format.id,
                    column_index=col_idx,
                    width=col_width,
                    data_type="text",
                    alignment="left",
                )

            # Create rows
            row_height = table.height / max(table.rows, 1)
            for row_idx in range(table.rows):
                self.table_row_repo.create_row(
                    table_format_id=table_format.id,
                    row_index=row_idx,
                    height=row_height,
                    is_header=(row_idx == 0),
                )

    def _create_image_regions(
        self,
        template: DocumentFormat,
        page: Any,
        page_number: int,
    ) -> None:
        """Create image region definitions from detected images."""
        for img in page.images:
            self.image_region_repo.create_image_region(
                format_id=template.id,
                page_number=page_number,
                x=img.x,
                y=img.y,
                width=img.width,
                height=img.height,
                image_path=img.image_path,
                image_type=img.image_type,
                is_static=True,
            )

    def _create_watermarks(
        self,
        template: DocumentFormat,
        page: Any,
        page_number: int,
    ) -> None:
        """Create watermark definitions from detected watermark text blocks."""
        watermark_blocks = [b for b in page.text_blocks if b.block_type == "watermark"]
        for block in watermark_blocks:
            self.watermark_repo.create_watermark(
                format_id=template.id,
                page_number=page_number,
                text=block.text,
                x=block.x,
                y=block.y,
                width=block.width,
                height=block.height,
                opacity=0.3,
                rotation=0.0,
                font_family=block.font_family,
                font_size=block.font_size,
                font_color=block.font_color or "#CCCCCC",
            )

    def _is_dynamic_field(self, text: str) -> bool:
        """Determine if a text block represents a dynamic (variable) field."""
        if not text:
            return False

        # Common patterns indicating dynamic content
        dynamic_patterns = [
            "{{", "}}", "${", "##",
            "__________", "___", "...........",
        ]
        for pattern in dynamic_patterns:
            if pattern in text:
                return True

        # Short single-word values that might be labels are static
        # Longer values with numbers/dates tend to be dynamic
        import re
        # Date patterns
        if re.search(r"\d{1,2}[/\-\.]\d{1,2}[/\-\.]\d{2,4}", text):
            return True
        # Currency patterns
        if re.search(r"[$€£¥]\s*[\d,]+\.?\d*", text):
            return True
        # Phone patterns
        if re.search(r"\+?\d[\d\s\-()]{7,}", text):
            return True

        return False

    def _generate_field_name(self, text: str, sequence: int) -> str:
        """Generate a field name from text content."""
        import re
        # Clean text
        clean = re.sub(r"[^a-zA-Z0-9\s]", "", text)
        clean = clean.strip().lower()
        words = clean.split()[:3]
        if words:
            return "_".join(words)
        return f"field_{sequence}"

    def _infer_data_type(self, text: str) -> str:
        """Infer the data type from text content."""
        import re

        if not text:
            return "text"

        stripped = text.strip()

        # Number
        if re.match(r"^-?[\d,]+\.?\d*$", stripped.replace(",", "")):
            return "number"

        # Date
        if re.search(r"\d{1,2}[/\-\.]\d{1,2}[/\-\.]\d{2,4}", stripped):
            return "date"

        # Currency
        if re.search(r"^[$€£¥]\s*[\d,]+\.?\d*$", stripped):
            return "currency"

        # Email
        if re.search(r"[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}", stripped):
            return "email"

        return "text"

    def _infer_alignment(self, x: float, page_width: float) -> str:
        """Infer text alignment based on horizontal position."""
        if page_width <= 0:
            return "left"

        relative_x = x / page_width

        if relative_x < 0.15:
            return "left"
        elif relative_x > 0.6:
            return "right"
        elif 0.35 < relative_x < 0.65:
            return "center"

        return "left"