bundle.py

import os
import itertools
from lxml import etree
from documentstore_migracao.config import BASE_PATH


def find_assets_nodes(xmltree: etree.ElementTree) -> iter:
    """Busca por todos os assets e retorna um iterator com
    de todos os nós etree.Element encontrados"""
    paths = [
        ".//ext-link[@xlink:href]",
        ".//graphic[@xlink:href]",
        ".//inline-graphic[@xlink:href]",
        ".//inline-supplementary-material[@xlink:href]",
        ".//media[@xlink:href]",
        ".//supplementary-material[@xlink:href]",
    ]
    iterators = [
        xmltree.iterfind(path, namespaces={"xlink": "http://www.w3.org/1999/xlink"})
        for path in paths
    ]

    return itertools.chain(*iterators)


def get_assets_list(assets_nodes: iter) -> list:
    """Retorna uma lista contendo dicionários de ativos digitais"""
    link_selector = "{http://www.w3.org/1999/xlink}href"
    assets = []

    for node in assets_nodes:
        asset_name = node.get(link_selector).split("/")[-1]
        asset_name = os.path.splitext(asset_name)[0]
        assets.append({"id": asset_name, "url": node.get(link_selector)})
    # import pdb; pdb.set_trace()
    return assets


path = os.path.join(
    BASE_PATH, "xml/sps_packages/S0066-782X2004001200001/1678-4170-abc-82-s06-1-14.xml"
)

xml = etree.parse(path)
assets_iterator = find_assets_nodes(xml)

from documentstore_migracao.utils.xylose_converter import parse_date
from uuid import uuid4


def get_document_bundle_manifest(document: etree.ElementTree, document_url: str) -> dict:
    """Cria um manifesto no formato do Kernel a partir de um
    documento xml"""

    try:
        _id = document.find(".//article-id[@pub-id-type='scielo-id']").text
    except AttributeError:
        # raise ValueError("Document requires an scielo-id") from None
        _id = str(uuid4())

    # TODO: Qual é a data adequada pra usar como "creation_date"?
    date = document.find(".//pub-date[@pub-type='epub-ppub']")

    if date is None:
        raise ValueError("A creation date is required") from None

    _creation_date = parse_date(
        "%s-%s" % (date.find(".//year").text, date.find(".//month").text)
    )

    _version = {"data": document_url, "assets": {}, "timestamp": _creation_date}
    _document = {"id": _id, "versions": [_version]}

    asset_nodes = find_assets_nodes(document)
    asset_list = get_assets_list(asset_nodes)

    for asset in asset_list:
        _version["assets"][asset.get("id")] = [[_creation_date, asset.get("url")]]

    return _document

get_document_bundle_manifest(xml, "http://scielo-bucket.scielo.br/document.xml")

fluxo.txt

 +-----------------------------+      +------------------------------+
 |                             |      |                              |
 |                             |      |                              |
 |                             |      |        Converte XMLS         |
 |        Extrai XMLS          +------>        Extraídos             |
 |                             |      |                              |
 |                             |      |                              |
 |                             |      |                              |
 +-----------------------------+      +---------------+--------------+
                                                      |
                                                      |
                                                      |
+------------------------------+      +---------------v--------------+
|                              |      |                              |
|                              |      |                              |
|      Gera pacotes SPS        |      |                              |
|      dos XMLS válidos        <------+  Valida os XMLS convertidos  |
|                              |      |                              |
|                              |      |                              |
|                              |      |                              |
+---------------+--------------+      +------------------------------+
                |
                |
                |
+---------------v--------------+      +------------------------------+
|                              |      |                              |
|                              |      |                              |
|         Aplica o XML         |      |    Envia pacotes SPS para    |
|         Constructor          +------>    o Object Storage          |
|                              |      |                              |
|                              |      |                              |
|                              |      |                              |
+------------------------------+      +--------------+---------------+
                                                     |
                                                     v
                                      +--------------+----------------+
                                      |                               |
                                      |                               |
                                      |    Gera o manifesto do XML    |
                                      |    e registra no Banco        |
                                      |    de dados                   |
                                      |                               |
                                      |                               |
                                      +-------------------------------+

manifest.json

{
    "id": "e8ec3b28-825d-4b60-9711-7dbc106ecfbf",
    "versions": [
        {
            "data": "http://scielo-bucket.scielo.br/document.xml",
            "assets": {
                "rad": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "http://www.vtmed.org/vascular/rad.htm"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-glogo": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-glogo.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img01": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img01.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img02": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img02.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img03": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img03.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img04": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img04.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img05": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img05.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img06": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img06.gif"
                    ]
                ],
                "1678-4170-abc-82-s06-1-14-ga01img07": [
                    [
                        "2004-04-01T00:00:00.000000Z",
                        "1678-4170-abc-82-s06-1-14-ga01img07.gif"
                    ]
                ]
            },
            "timestamp": "2004-04-01T00:00:00.000000Z"
        }
    ]
}