HartOMat/backend/app/services/excel_import.py

"""Excel import service — maps parsed rows to Product library."""
from dataclasses import dataclass, field
from pathlib import PurePosixPath
from sqlalchemy.ext.asyncio import AsyncSession

from app.services.product_service import (
    lookup_or_create_product,
    lookup_product,
)


def _stem_lower(name: str | None) -> str:
    """Return the lowercased stem (no extension) of a filename."""
    if not name:
        return ""
    return PurePosixPath(name).stem.lower()


@dataclass
class PreviewResult:
    """Read-only preview: annotates rows without creating anything."""
    rows: list[dict] = field(default_factory=list)
    existing_product_count: int = 0
    new_product_count: int = 0
    no_pim_id_count: int = 0
    has_step_count: int = 0
    no_step_count: int = 0
    duplicate_count: int = 0
    step_conflict_count: int = 0
    cad_name_conflict_count: int = 0
    warnings: list[str] = field(default_factory=list)


@dataclass
class ImportResult:
    rows: list[dict] = field(default_factory=list)
    matched_count: int = 0
    created_count: int = 0
    no_pim_id_count: int = 0
    duplicate_baureihe_count: int = 0
    warnings: list[str] = field(default_factory=list)


async def import_excel_to_products(
    db: AsyncSession,
    parsed_rows: list[dict],
    source_excel: str,
    category_key: str | None = None,
    tenant_id=None,
) -> ImportResult:
    """For each row, look up or create a Product.

    Grouping strategy:
    1. Primary key: produkt_baureihe (lowercased)
    2. Fallback: pim_id (backward compat)

    Annotates each row dict with product_id, product_created.
    """
    result = ImportResult()

    # Track seen produkt_baureihe values to skip duplicates
    seen_baureihe: dict[str, str] = {}  # lower(baureihe) → first product_id

    for row in parsed_rows:
        pim_id = row.get("pim_id")
        produkt_baureihe = row.get("produkt_baureihe")
        row_category = row.get("category_key") or category_key

        # Need at least one identifier
        if not pim_id and not produkt_baureihe:
            row["product_id"] = None
            row["product_created"] = False
            result.no_pim_id_count += 1
            continue

        fields = {
            "name": produkt_baureihe or row.get("gewaehltes_produkt"),
            "category_key": row_category,
            "ebene1": row.get("ebene1"),
            "ebene2": row.get("ebene2"),
            "baureihe": row.get("baureihe"),
            "produkt_baureihe": produkt_baureihe,
            "lagertyp": row.get("lagertyp"),
            "name_cad_modell": row.get("name_cad_modell"),
            "gewuenschte_bildnummer": row.get("gewuenschte_bildnummer"),
            "medias_rendering": row.get("medias_rendering"),
            "components": row.get("components", []),
            "arbeitspaket": row.get("arbeitspaket"),
            "source_excel": source_excel,
        }

        product, was_created = await lookup_or_create_product(db, pim_id, fields, tenant_id=tenant_id)
        row["product_id"] = str(product.id)
        row["product_created"] = was_created
        # Carry forward any STEP file already linked to this product
        row["product_cad_file_id"] = str(product.cad_file_id) if product.cad_file_id else None

        if was_created:
            result.created_count += 1
        else:
            result.matched_count += 1

        # Track duplicate baureihe
        if produkt_baureihe:
            bkey = produkt_baureihe.lower()
            if bkey in seen_baureihe:
                result.duplicate_baureihe_count += 1
            else:
                seen_baureihe[bkey] = str(product.id)

    result.rows = parsed_rows
    # NOTE: caller is responsible for db.commit() — keeps the transaction
    # composable with order + line creation in the finalize endpoint.
    return result


async def preview_excel_rows(
    db: AsyncSession,
    parsed_rows: list[dict],
    category_key: str | None = None,
) -> PreviewResult:
    """Read-only preview: annotates rows with product_exists / product_id / duplicate flags.

    Uses lookup_product (read-only) to check what already exists in the DB.
    New-vs-existing is determined per unique produkt_baureihe (or pim_id fallback).
    Duplicate rows (same produkt_baureihe seen more than once in this batch) are
    annotated with is_duplicate=True and duplicate_of_row=<first_row_index>.
    """
    result = PreviewResult()
    # Track unique identifiers we've already resolved in this batch
    # key = lower(baureihe) or pim_id  →  (product_exists, product_id_str | None, has_step, first_row_index, name_cad_modell_stem)
    seen: dict[str, tuple[bool, str | None, bool, int, str]] = {}

    for row in parsed_rows:
        pim_id = row.get("pim_id")
        produkt_baureihe = row.get("produkt_baureihe")
        row_index = row.get("row_index", 0)
        row["category_key"] = row.get("category_key") or category_key

        # Default conflict fields
        row["step_conflict"] = False
        row["step_conflict_existing_name"] = None
        row["step_conflict_excel_name"] = None
        row["cad_name_conflict"] = False
        row["cad_name_conflict_other_name"] = None
        row["cad_name_conflict_row"] = None

        # Must have at least one identifier
        if not pim_id and not produkt_baureihe:
            row["product_exists"] = False
            row["product_id"] = None
            row["has_step"] = False
            row["is_duplicate"] = False
            result.no_pim_id_count += 1
            continue

        # Build a cache key
        cache_key = (produkt_baureihe or "").lower() or pim_id or ""

        excel_cad_name = row.get("name_cad_modell")
        excel_cad_stem = _stem_lower(excel_cad_name)

        if cache_key in seen:
            exists, pid, has_step, first_row, first_cad_stem = seen[cache_key]
            row["product_exists"] = exists
            row["product_id"] = pid
            row["has_step"] = has_step
            row["is_duplicate"] = True
            row["duplicate_of_row"] = first_row

            # Intra-Excel conflict: same product key, different name_cad_modell
            if excel_cad_stem and first_cad_stem and excel_cad_stem != first_cad_stem:
                row["cad_name_conflict"] = True
                row["cad_name_conflict_other_name"] = first_cad_stem
                row["cad_name_conflict_row"] = first_row
                result.cad_name_conflict_count += 1

            result.duplicate_count += 1
            continue

        product = await lookup_product(db, pim_id, produkt_baureihe)
        row["is_duplicate"] = False
        if product is not None:
            has_step = product.cad_file_id is not None
            row["product_exists"] = True
            row["product_id"] = str(product.id)
            row["has_step"] = has_step
            seen[cache_key] = (True, str(product.id), has_step, row_index, excel_cad_stem)
            result.existing_product_count += 1
            if has_step:
                result.has_step_count += 1
            else:
                result.no_step_count += 1

            # STEP conflict: product already has a different STEP file
            if has_step and excel_cad_stem and product.cad_file:
                existing_stem = _stem_lower(product.cad_file.original_name)
                if existing_stem and existing_stem != excel_cad_stem:
                    row["step_conflict"] = True
                    row["step_conflict_existing_name"] = existing_stem
                    row["step_conflict_excel_name"] = excel_cad_stem
                    result.step_conflict_count += 1
        else:
            row["product_exists"] = False
            row["product_id"] = None
            row["has_step"] = False
            seen[cache_key] = (False, None, False, row_index, excel_cad_stem)
            result.new_product_count += 1
            result.no_step_count += 1

    result.rows = parsed_rows
    if result.duplicate_count > 0:
        result.warnings.append(
            f"{result.duplicate_count} duplicate Produkt-Baureihe row(s) detected — "
            "these are pre-unchecked. Only one row per product will be imported."
        )
    if result.step_conflict_count > 0:
        result.warnings.append(
            f"{result.step_conflict_count} product(s) already have a different STEP file linked — "
            "importing will not replace the existing STEP file automatically."
        )
    if result.cad_name_conflict_count > 0:
        result.warnings.append(
            f"{result.cad_name_conflict_count} row(s) reference the same product with a different CAD model name."
        )
    return result