pd.read_csv(
    "data.csv.gz",
    delimiter = "^",
    # line numbers to skip (i.e. headers in an excel report)
    skiprows = 2,
    # used to denote the start and end of a quoted item
    quotechar = "|",
    # return a subset of columns
    usecols = ["return_date", "company", "sales"],
    # data type for data or columns
    dtype = { "sales": np.float64 },
    # additional strings to recognize as NA/NaN
    na_values = [".", "?"],
    # convert to datetime, instead of object
    parse_dates = ["return_date"],
    # for on-the-fly decompression of on-disk data
    # options - gzip, bz2, zip, xz
    compression = "gzip",
    # encoding to use for reading
    encoding = "latin1",
    # read in a subset of data
    nrows = 100
)