europeana_newspapers.py

# coding=utf-8
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""TODO"""


import pyarrow as pa
import pyarrow.parquet as pq
from datasets import Value, Sequence
import datasets
from datasets.config import PYARROW_VERSION
from datasets.utils.logging import get_logger

logger = get_logger(__name__)

if PYARROW_VERSION.major <= 6:
    msg = f"pyarrow version >= 7.0.0 required for this loading script, you have {PYARROW_VERSION}"
    logger.warning(msg)
    raise RuntimeError(msg)


_DESCRIPTION = "TODO"

_HOMEPAGE = "TODO"

_LANG_CONFIGS = {"fi", "sv"}

_DATA = {
    "sv": {"1900": "sv_1900.parquet", "1910": "sv_1910.parquet"},
    "fi": {"1900": "fi_1900.parquet", "1910": "fi_1910.parquet"},
}


class EuropeanaNewspapersConfig(datasets.BuilderConfig):
    """BuilderConfig for the Europeana Newspapers dataset."""

    def __init__(
        self, *args, languages=None, min_decade=None, max_decade=None, **kwargs
    ):
        """BuilderConfig for the GitHub Code dataset.

        Args:
            languages (:obj:`List[str]`): List of languages to load.
            **kwargs: keyword arguments forwarded to super.
        """

        super().__init__(
            *args,
            name="+".join(languages),
            **kwargs,
        )
        for lang in languages:
            if not lang in _LANG_CONFIGS:
                raise ValueError(
                    f"{lang} not a valid language key for this dataset, valid keys are {_LANG_CONFIGS}"
                )
        self.languages = languages
        self.min_decade = min_decade
        self.max_decade = max_decade


class EuropeanaNewspapers(datasets.GeneratorBasedBuilder):
    """TODO."""

    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIG_CLASS = EuropeanaNewspapersConfig
    BUILDER_CONFIGS = [
        EuropeanaNewspapersConfig(languages=[lang]) for lang in _LANG_CONFIGS
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "text": Value(dtype="string"),
                    "mean_ocr": Value(dtype="float64"),
                    "std_ocr": Value(dtype="float64"),
                    "bounding_boxes": Sequence(
                        feature=Sequence(
                            feature=Value(dtype="float64", id=None),
                            length=-1,
                        ),
                    ),
                    "title": Value(dtype="string"),
                    "date": Value(dtype="string"),
                    "language": Sequence(
                        feature=Value(dtype="string", id=None),
                    ),
                    "item_iiif_url": Value(
                        dtype="string",
                    ),
                    # "multi_language": Value(dtype="bool"),
                    "issue_uri": Value(dtype="string"),
                    "id": Value(dtype="string"),
                }
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            license="Multiple: see the 'license' field of each sample.",
        )

    def _split_generators(self, dl_manager):
        # parquet_files = list(Path(".").rglob("*.parquet"))
        languages = self.config.languages
        min_decade = self.config.min_decade
        max_decade = self.config.max_decade
        data_files = []
        for language in languages:
            for decade, file in _DATA[language].items():
                decade = int(decade)
                if max_decade is None and min_decade is None:
                    data_files.append(file)
                if (
                    max_decade is not None
                    and min_decade is not None
                    and min_decade <= decade <= max_decade
                ):
                    data_files.append(file)
                if (
                    min_decade is not None
                    and max_decade is None
                    and decade >= min_decade
                ):
                    data_files.append(file)
                if (
                    min_decade is None
                    and max_decade is not None
                    and decade <= max_decade
                ):
                    data_files.append(file)

        files = dl_manager.download(data_files)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "files": files,
                },
            ),
        ]

    def _generate_examples(self, files):
        key = 0
        for file in files:
            with open(file, "rb") as f:
                parquet_file = pq.ParquetFile(f)
                for record_batch in parquet_file.iter_batches(batch_size=10_000):
                    pa_table = pa.Table.from_batches([record_batch])
                    rows = pa_table.to_pylist()
                    for row in rows:
                        row.pop("multi_language")
                        yield key, row
                        key += 1