flibusta.py

import dataclasses
import re
import urllib.parse
import xml.etree.ElementTree
import zipfile
from dataclasses import dataclass
from textwrap import dedent

import datasets


logger = datasets.logging.get_logger(__name__)


FLIBUSTA_URL = "https://flibusta.is"


_CITATION = """\
@ONLINE{flibusta,
    author = "Флибуста",
    title  = "Флибуста",
    url    = "https://flibusta.is"
}
"""


_DESCRIPTION = """\
Convenient access to books in Russian hosted on Flibusta (https://flibusta.is/).

Authors of the dataset do not endorse the usage of Flibusta for illegal
purposes: please read "Licensing Information" before use.
"""


_LICENSE = """\
Books are stored on https://flibusta.is/ and may not be accessible from your
location because of legal reasons.

Please check with your local law if you can use this dataset.

The license Apache 2.0 applies only to the code.
"""


@dataclass
class BookInfo:
    author: str
    id: str
    title: str
    url: str
    url_txt: str


@dataclass
class DownloadedBookInfo:
    book_info: BookInfo
    file_path: str


class FlibustaConfig(datasets.BuilderConfig):
    def __init__(self, *args, books_query, **kwargs):
        super().__init__(*args, version=datasets.utils.Version("0.1.0"), **kwargs)
        self.books_query = books_query


class Flibusta(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIG_CLASS = FlibustaConfig

    def _info(self):
        return datasets.DatasetInfo(
            citation=_CITATION,
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "author": datasets.Value("string"),
                    "id": datasets.Value("string"),
                    "text": datasets.Value("string"),
                    "title": datasets.Value("string"),
                    "url": datasets.Value("string"),
                    "url_txt": datasets.Value("string"),
                }
            ),
            homepage=FLIBUSTA_URL,
            version=datasets.Version("0.1.0"),
        )

    @staticmethod
    def _get_books(dl_manager, query):
        flibusta_search_url = urllib.parse.urljoin(FLIBUSTA_URL, "opds/search")
        continue_traverse = True
        page = 0
        while continue_traverse:
            params_str = urllib.parse.urlencode(
                {
                    "searchType": "books",
                    "searchTerm": query,
                    "pageNumber": page,
                }
            )
            url = f"{flibusta_search_url}?{params_str}"
            logger.info("Extracting book infos from: %s", url)
            search_xml_file_path = dl_manager.download(url)
            search_xml_str = open(search_xml_file_path, encoding="utf8").read()
            search_xml_tree = xml.etree.ElementTree.fromstring(search_xml_str)
            ns = "{http://www.w3.org/2005/Atom}"
            for entry in search_xml_tree.findall(f"./{ns}entry"):
                author = entry.findtext(f"./{ns}author/{ns}name")
                title = entry.findtext(f"./{ns}title")
                url_path_entry = entry.find(f'./{ns}link[@title="Книга на сайте"]')
                url_txt_path_entry = entry.find(
                    f'./{ns}link[@type="application/txt+zip"]'
                )
                if (
                    author
                    and title
                    and url_path_entry is not None
                    and url_txt_path_entry is not None
                ):
                    url_path = url_path_entry.get("href")
                    url = urllib.parse.urljoin(FLIBUSTA_URL, url_path)
                    url_txt_path = url_txt_path_entry.get("href")
                    url_txt = urllib.parse.urljoin(FLIBUSTA_URL, url_txt_path)
                    id_ = url_path.rpartition("/")[2]
                    yield BookInfo(
                        author=author,
                        id=id_,
                        title=title,
                        url=url,
                        url_txt=url_txt,
                    )
            if search_xml_tree.find(f'./{ns}link[@rel="next"]') is not None:
                page += 1
            else:
                continue_traverse = False

    def _split_generators(self, dl_manager):
        book_infos = list(
            self._get_books(dl_manager=dl_manager, query=self.config.books_query)
        )
        book_urls_txt = {book_info.id: book_info.url_txt for book_info in book_infos}
        book_file_paths = dl_manager.download(book_urls_txt)
        downloaded_book_infos = [
            DownloadedBookInfo(
                book_info=book_info, file_path=book_file_paths[book_info.id]
            )
            for book_info in book_infos
        ]
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"downloaded_book_infos": downloaded_book_infos},
            )
        ]

    def _generate_examples(self, downloaded_book_infos):
        for downloaded_book_info in downloaded_book_infos:
            archive = zipfile.ZipFile(open(downloaded_book_info.file_path, "rb"))
            for content_name in archive.namelist():
                if content_name.endswith(".txt"):
                    txt_path = zipfile.Path(archive, at=content_name)
                    text = txt_path.read_text(encoding="utf8")

                    text = re.sub(r"Взято из Флибусты,.*", "", text)
                    text = text.strip("\n")
                    text = dedent(text)

                    example = {
                        "text": text,
                        **dataclasses.asdict(downloaded_book_info.book_info),
                    }
                    yield downloaded_book_info.book_info.id, example
                    break