PacketStation/server/packetstation/drawings/fetch.py

"""Copy a released drawing PDF from M-Files into assets.

Downloads only when the M-Files identity (object id and version, file id and version) differs from what the
`drawing_lookups` row last stored, or that row has no asset (or the asset's file is gone). The download goes to
ASSETS_DIR/tmp/<uuid>; the PDF magic and the size limit are checked, then `assets.store_file` hashes it in chunks,
fsyncs and renames it into place, and reuses an existing asset row with the same sha256 and size.

Read-only to M-Files: only `Lookup.download` is called.
"""

from __future__ import annotations

import contextlib
import uuid
from dataclasses import dataclass
from pathlib import Path
from typing import Any

from psycopg.rows import dict_row

from ..api.errors import ApiError
from ..domain import assets
from .lookup import Found, Lookup

PDF_MAGIC = b"%PDF-"


class FetchError(Exception):
    """The drawing could not be copied. The message is safe to log."""


@dataclass(frozen=True, slots=True)
class FetchResult:
    asset_id: uuid.UUID
    sha256: bytes
    pages: int
    downloaded: bool  # False when the stored asset was still current
    name: str | None = None
    reused: bool = False  # downloaded, but the same bytes were already stored


def _current(conn: Any, settings: Any, found: Found, dwg: str) -> FetchResult | None:
    with conn.cursor(row_factory=dict_row) as cur:
        row = cur.execute(
            "SELECT l.obj_id, l.obj_version, l.file_id, l.file_version, l.asset_id, a.sha256, a.pages,"
            " a.storage_key, a.original_name FROM drawing_lookups l LEFT JOIN assets a ON a.id = l.asset_id"
            " WHERE l.dwg = %s",
            (dwg,),
        ).fetchone()
    if row is None or row["asset_id"] is None or row["sha256"] is None:
        return None
    if (row["obj_id"], row["obj_version"], row["file_id"], row["file_version"]) != found.identity:
        return None
    if None in found.identity:
        return None
    try:
        if not assets.path_of(settings, str(row["storage_key"])).is_file():
            return None
    except ApiError:
        return None
    return FetchResult(
        asset_id=row["asset_id"],
        sha256=bytes(row["sha256"]),
        pages=int(row["pages"] or 0),
        downloaded=False,
        name=row["original_name"],
    )


def fetch(conn: Any, settings: Any, lookup: Lookup, found: Found, *, dwg: str) -> FetchResult:
    """The asset for M-Files' released file of `dwg`. Raises FetchError."""
    if not found.ok or found.rev is None:
        raise FetchError(f"{dwg}: no released file in M-Files")
    cur = _current(conn, settings, found, dwg)
    if cur is not None:
        return cur
    limit = int(settings.upload_max_bytes)
    if found.file_size is not None and found.file_size > limit:
        raise FetchError(f"{dwg}: the M-Files file is {found.file_size} bytes, over the {limit} byte limit")
    tmp_dir = Path(settings.assets_dir) / "tmp"
    tmp_dir.mkdir(parents=True, exist_ok=True)
    tmp = tmp_dir / uuid.uuid4().hex
    try:
        try:
            lookup.download(found, tmp)
        except LookupError as e:
            raise FetchError(str(e)) from None
        except OSError as e:
            raise FetchError(f"{dwg}: could not write the download: {type(e).__name__}") from None
        try:
            size = tmp.stat().st_size
            with tmp.open("rb") as f:
                head = f.read(len(PDF_MAGIC))
        except OSError:
            raise FetchError(f"{dwg}: the download left no file") from None
        if size > limit:
            raise FetchError(f"{dwg}: the M-Files file is {size} bytes, over the {limit} byte limit")
        if head != PDF_MAGIC:
            raise FetchError(f"{dwg}: the M-Files file is not a PDF")
        name = found.name or f"{dwg} r{found.rev}.pdf"
        source_ref = {
            "dwg": dwg,
            "obj_id": found.obj_id,
            "obj_version": found.obj_version,
            "file_id": found.file_id,
            "file_version": found.file_version,
            "rev": found.rev,
        }
        try:
            row = assets.store_file(
                conn, settings, tmp, mime=assets.PDF, original_name=name, source="mfiles", source_ref=source_ref
            )
        except ApiError as e:
            raise FetchError(f"{dwg}: {e.detail}") from None
        return FetchResult(
            asset_id=row["id"],
            sha256=bytes(row["sha256"]),
            pages=int(row["pages"] or 0),
            downloaded=True,
            name=row.get("original_name") or name,
            reused=bool(row.get("reused")),
        )
    finally:
        with contextlib.suppress(OSError):
            tmp.unlink()