Function _read_csv

dask/dataframe/io/csv.py:296–364

(
    block,
    part,
    columns,
    *,
    reader,
    header,
    dtypes,
    head,
    colname,
    full_columns,
    enforce,
    kwargs,
    blocksize,
)

def _read_csv(
    block,
    part,
    columns,
    *,
    reader,
    header,
    dtypes,
    head,
    colname,
    full_columns,
    enforce,
    kwargs,
    blocksize,
):
    # Part will be a 3-element tuple
    path, is_first, is_last = part

    # Construct `path_info`
    if path is not None:
        path_info = (
            colname,
            path,
            sorted(head[colname].cat.categories),
        )
    else:
        path_info = None

    # Deal with arguments that are special
    # for the first block of each file
    write_header = False
    rest_kwargs = kwargs.copy()
    if not is_first:
        if rest_kwargs.get("names", None) is None:
            write_header = True
        rest_kwargs.pop("skiprows", None)
        if rest_kwargs.get("header", 0) is not None:
            rest_kwargs.pop("header", None)
    if not is_last:
        rest_kwargs.pop("skipfooter", None)

    # Deal with column projection
    _columns = full_columns
    project_after_read = False
    if columns is not None and columns != full_columns:
        if kwargs:
            # To be safe, if any kwargs are defined, avoid
            # changing `usecols` here. Instead, we can just
            # select columns after the read
            project_after_read = True
        else:
            _columns = columns
            rest_kwargs["usecols"] = _columns

    # Call `pandas_read_text`
    df = pandas_read_text(
        reader,
        block,