.. blogpost::
    :title: read_csv and zip files
    :keywords: pandas, read_csv, zip
    :date: 2016-02-06
    :categories: pandas
    :lid: blogpost_read_csv

    `read_csv <http://pandas.pydata.org/pandas-docs/stable/generated/pandas.read_csv.html>`_
    is no longer able to extract a dataframe from a zip file.
    The parameter *format* changed for *compression* but
    the zip format disappeared from the list. I assume
    the reason is that zip files can contains many files.

    *pyquickhelper* now implements the function
    :func:`read_csv <pyquickhelper.pandashelper.readh.read_csv>`
    which can extract all dataframe in a zip file or falls back
    into the regular function if no zip format is detected.
    In that case, it returns a dictionary of dataframes indexed by their name
    in the zip file.

    ::

        from pyquickhelper.pandashelper import read_csv
        dfs = read_csv("url_or_filename.zip", compression="zip")
        print(dfs["dataframe.txt"].head())

    If only one file must be converted as a dataframe,
    the parameter fvalid must be used:

    ::

        from pyquickhelper.pandashelper import read_csv
        dfs = read_csv("url_or_filename.zip", compression="zip",
                       fvalid=lambda name: name == "the_file.txt")
        print(dfs["the_file.txt"].head())

    The others files will be loaded as text. In more details,
    when it is a zip file, the function reads a dataframe
    from a zip file by doing:

    ::

        import io, zipfile, pandas

        def read_zip(local_file, encoding="utf8"):

            with open(local_file, "rb") as local_file:
                content = local_file.read()

            dfs = {}
            with zipfile.ZipFile(io.BytesIO(content)) as myzip:
                infos = myzip.infolist()

                for info in infos:
                    name = info.filename
                    with myzip.open(name, "r") as f:
                        text = f.read()

                    text = text.decode(encoding="utf8")
                    st = io.StringIO(text)
                    df = pandas.read_csv(st, compression=compression, **params)
                    dfs[name] = df

            return dfs