cli.py

import csv
import sys

from pathlib import Path

from datasets import load_dataset
from markdown_it import MarkdownIt
from mdit_plain.renderer import RendererPlain
from minicli import cli, run
from progressist import ProgressBar


csv.field_size_limit(sys.maxsize)
catalog = Path("datasets-filtered.csv")


@cli
def generate():
    """
    Loads catalog, parse markdown description, return concatenated title and description
    """
    print("Loading corpus...")
    corpus = []
    parser = MarkdownIt(renderer_cls=RendererPlain)
    with catalog.open() as f:
        reader = csv.DictReader(f)
        lines = list(reader)
        bar = ProgressBar(total=len(lines))
        for line in bar.iter(lines):
            description = parser.render(line["description"])
            corpus.append({
                "id": line["id"],
                "content": f'{line["title"]} {description}',
                "tags": line["tags"],
            })
    with open("data/datagouvfr-datasets.csv", "w") as f:
        writer = csv.DictWriter(f, fieldnames=["id", "content", "tags"])
        writer.writeheader()
        writer.writerows(corpus)

    return corpus


@cli
def inspect():
    ds = load_dataset("abulte/datagouvfr-datasets", split="train")
    print(list(zip(ds["id"][:10], ds["content"][:10], ds["tags"][:10])))


if __name__ == "__main__":
    run()