test_to_csv.py

# -*- coding: utf-8 -*-

from __future__ import print_function

import csv
import pytest

from numpy import nan
import numpy as np

from pandas.compat import (lmap, range, lrange, StringIO, u)
from pandas.io.common import _get_handle
import pandas.core.common as com
from pandas.errors import ParserError
from pandas import (DataFrame, Index, Series, MultiIndex, Timestamp,
                    date_range, read_csv, compat, to_datetime)
import pandas as pd

from pandas.util.testing import (assert_almost_equal,
                                 assert_series_equal,
                                 assert_frame_equal,
                                 ensure_clean,
                                 makeCustomDataframe as mkdf)
import pandas.util.testing as tm

from pandas.tests.frame.common import TestData


MIXED_FLOAT_DTYPES = ['float16', 'float32', 'float64']
MIXED_INT_DTYPES = ['uint8', 'uint16', 'uint32', 'uint64', 'int8', 'int16',
                    'int32', 'int64']


class TestDataFrameToCSV(TestData):

    def read_csv(self, path, **kwargs):
        params = dict(index_col=0, parse_dates=True)
        params.update(**kwargs)

        return pd.read_csv(path, **params)

    def test_from_csv_deprecation(self):
        # see gh-17812
        with ensure_clean('__tmp_from_csv_deprecation__') as path:
            self.tsframe.to_csv(path)

            with tm.assert_produces_warning(FutureWarning):
                depr_recons = DataFrame.from_csv(path)
                assert_frame_equal(self.tsframe, depr_recons)

    def test_to_csv_from_csv1(self):

        with ensure_clean('__tmp_to_csv_from_csv1__') as path:
            self.frame['A'][:5] = nan

            self.frame.to_csv(path)
            self.frame.to_csv(path, columns=['A', 'B'])
            self.frame.to_csv(path, header=False)
            self.frame.to_csv(path, index=False)

            # test roundtrip
            self.tsframe.to_csv(path)
            recons = self.read_csv(path)
            assert_frame_equal(self.tsframe, recons)

            self.tsframe.to_csv(path, index_label='index')
            recons = self.read_csv(path, index_col=None)

            assert(len(recons.columns) == len(self.tsframe.columns) + 1)

            # no index
            self.tsframe.to_csv(path, index=False)
            recons = self.read_csv(path, index_col=None)
            assert_almost_equal(self.tsframe.values, recons.values)

            # corner case
            dm = DataFrame({'s1': Series(lrange(3), lrange(3)),
                            's2': Series(lrange(2), lrange(2))})
            dm.to_csv(path)

            recons = self.read_csv(path)
            assert_frame_equal(dm, recons)

    def test_to_csv_from_csv2(self):

        with ensure_clean('__tmp_to_csv_from_csv2__') as path:

            # duplicate index
            df = DataFrame(np.random.randn(3, 3), index=['a', 'a', 'b'],
                           columns=['x', 'y', 'z'])
            df.to_csv(path)
            result = self.read_csv(path)
            assert_frame_equal(result, df)

            midx = MultiIndex.from_tuples(
                [('A', 1, 2), ('A', 1, 2), ('B', 1, 2)])
            df = DataFrame(np.random.randn(3, 3), index=midx,
                           columns=['x', 'y', 'z'])

            df.to_csv(path)
            result = self.read_csv(path, index_col=[0, 1, 2],
                                   parse_dates=False)
            assert_frame_equal(result, df, check_names=False)

            # column aliases
            col_aliases = Index(['AA', 'X', 'Y', 'Z'])
            self.frame2.to_csv(path, header=col_aliases)

            rs = self.read_csv(path)
            xp = self.frame2.copy()
            xp.columns = col_aliases
            assert_frame_equal(xp, rs)

            pytest.raises(ValueError, self.frame2.to_csv, path,
                          header=['AA', 'X'])

    def test_to_csv_from_csv3(self):

        with ensure_clean('__tmp_to_csv_from_csv3__') as path:
            df1 = DataFrame(np.random.randn(3, 1))
            df2 = DataFrame(np.random.randn(3, 1))

            df1.to_csv(path)
            df2.to_csv(path, mode='a', header=False)
            xp = pd.concat([df1, df2])
            rs = pd.read_csv(path, index_col=0)
            rs.columns = lmap(int, rs.columns)
            xp.columns = lmap(int, xp.columns)
            assert_frame_equal(xp, rs)

    def test_to_csv_from_csv4(self):

        with ensure_clean('__tmp_to_csv_from_csv4__') as path:
            # GH 10833 (TimedeltaIndex formatting)
            dt = pd.Timedelta(seconds=1)
            df = pd.DataFrame({'dt_data': [i * dt for i in range(3)]},
                              index=pd.Index([i * dt for i in range(3)],
                                             name='dt_index'))
            df.to_csv(path)

            result = pd.read_csv(path, index_col='dt_index')
            result.index = pd.to_timedelta(result.index)
            # TODO: remove renaming when GH 10875 is solved
            result.index = result.index.rename('dt_index')
            result['dt_data'] = pd.to_timedelta(result['dt_data'])

            assert_frame_equal(df, result, check_index_type=True)

    def test_to_csv_from_csv5(self):

        # tz, 8260
        with ensure_clean('__tmp_to_csv_from_csv5__') as path:

            self.tzframe.to_csv(path)
            result = pd.read_csv(path, index_col=0, parse_dates=['A'])

            converter = lambda c: to_datetime(result[c]).dt.tz_localize(
                'UTC').dt.tz_convert(self.tzframe[c].dt.tz)
            result['B'] = converter('B')
            result['C'] = converter('C')
            assert_frame_equal(result, self.tzframe)

    def test_to_csv_cols_reordering(self):
        # GH3454
        import pandas as pd

        chunksize = 5
        N = int(chunksize * 2.5)

        df = mkdf(N, 3)
        cs = df.columns
        cols = [cs[2], cs[0]]

        with ensure_clean() as path:
            df.to_csv(path, columns=cols, chunksize=chunksize)
            rs_c = pd.read_csv(path, index_col=0)

        assert_frame_equal(df[cols], rs_c, check_names=False)

    def test_to_csv_new_dupe_cols(self):
        import pandas as pd

        def _check_df(df, cols=None):
            with ensure_clean() as path:
                df.to_csv(path, columns=cols, chunksize=chunksize)
                rs_c = pd.read_csv(path, index_col=0)

                # we wrote them in a different order
                # so compare them in that order
                if cols is not None:

                    if df.columns.is_unique:
                        rs_c.columns = cols
                    else:
                        indexer, missing = df.columns.get_indexer_non_unique(
                            cols)
                        rs_c.columns = df.columns.take(indexer)

                    for c in cols:
                        obj_df = df[c]
                        obj_rs = rs_c[c]
                        if isinstance(obj_df, Series):
                            assert_series_equal(obj_df, obj_rs)
                        else:
                            assert_frame_equal(
                                obj_df, obj_rs, check_names=False)

                # wrote in the same order
                else:
                    rs_c.columns = df.columns
                    assert_frame_equal(df, rs_c, check_names=False)

        chunksize = 5
        N = int(chunksize * 2.5)

        # dupe cols
        df = mkdf(N, 3)
        df.columns = ['a', 'a', 'b']
        _check_df(df, None)

        # dupe cols with selection
        cols = ['b', 'a']
        _check_df(df, cols)

    @pytest.mark.slow
    def test_to_csv_dtnat(self):
        # GH3437
        from pandas import NaT

        def make_dtnat_arr(n, nnat=None):
            if nnat is None:
                nnat = int(n * 0.1)  # 10%
            s = list(date_range('2000', freq='5min', periods=n))
            if nnat:
                for i in np.random.randint(0, len(s), nnat):
                    s[i] = NaT
                i = np.random.randint(100)
                s[-i] = NaT
                s[i] = NaT
            return s

        chunksize = 1000
        # N=35000
        s1 = make_dtnat_arr(chunksize + 5)
        s2 = make_dtnat_arr(chunksize + 5, 0)

        # s3=make_dtnjat_arr(chunksize+5,0)
        with ensure_clean('1.csv') as pth:
            df = DataFrame(dict(a=s1, b=s2))
            df.to_csv(pth, chunksize=chunksize)

            recons = self.read_csv(pth)._convert(datetime=True,
                                                 coerce=True)
            assert_frame_equal(df, recons, check_names=False,
                               check_less_precise=True)

    @pytest.mark.slow
    def test_to_csv_moar(self):

        def _do_test(df, r_dtype=None, c_dtype=None,
                     rnlvl=None, cnlvl=None, dupe_col=False):

            kwargs = dict(parse_dates=False)
            if cnlvl:
                if rnlvl is not None:
                    kwargs['index_col'] = lrange(rnlvl)
                kwargs['header'] = lrange(cnlvl)

                with ensure_clean('__tmp_to_csv_moar__') as path:
                    df.to_csv(path, encoding='utf8',
                              chunksize=chunksize)
                    recons = self.read_csv(path, **kwargs)
            else:
                kwargs['header'] = 0

                with ensure_clean('__tmp_to_csv_moar__') as path:
                    df.to_csv(path, encoding='utf8', chunksize=chunksize)
                    recons = self.read_csv(path, **kwargs)

            def _to_uni(x):
                if not isinstance(x, compat.text_type):
                    return x.decode('utf8')
                return x
            if dupe_col:
                # read_Csv disambiguates the columns by
                # labeling them dupe.1,dupe.2, etc'. monkey patch columns
                recons.columns = df.columns
            if rnlvl and not cnlvl:
                delta_lvl = [recons.iloc[
                    :, i].values for i in range(rnlvl - 1)]
                ix = MultiIndex.from_arrays([list(recons.index)] + delta_lvl)
                recons.index = ix
                recons = recons.iloc[:, rnlvl - 1:]

            type_map = dict(i='i', f='f', s='O', u='O', dt='O', p='O')
            if r_dtype:
                if r_dtype == 'u':  # unicode
                    r_dtype = 'O'
                    recons.index = np.array(lmap(_to_uni, recons.index),
                                            dtype=r_dtype)
                    df.index = np.array(lmap(_to_uni, df.index), dtype=r_dtype)
                elif r_dtype == 'dt':  # unicode
                    r_dtype = 'O'
                    recons.index = np.array(lmap(Timestamp, recons.index),
                                            dtype=r_dtype)
                    df.index = np.array(
                        lmap(Timestamp, df.index), dtype=r_dtype)
                elif r_dtype == 'p':
                    r_dtype = 'O'
                    recons.index = np.array(
                        list(map(Timestamp, to_datetime(recons.index))),
                        dtype=r_dtype)
                    df.index = np.array(
                        list(map(Timestamp, df.index.to_timestamp())),
                        dtype=r_dtype)
                else:
                    r_dtype = type_map.get(r_dtype)
                    recons.index = np.array(recons.index, dtype=r_dtype)
                    df.index = np.array(df.index, dtype=r_dtype)
            if c_dtype:
                if c_dtype == 'u':
                    c_dtype = 'O'
                    recons.columns = np.array(lmap(_to_uni, recons.columns),
                                              dtype=c_dtype)
                    df.columns = np.array(
                        lmap(_to_uni, df.columns), dtype=c_dtype)
                elif c_dtype == 'dt':
                    c_dtype = 'O'
                    recons.columns = np.array(lmap(Timestamp, recons.columns),
                                              dtype=c_dtype)
                    df.columns = np.array(
                        lmap(Timestamp, df.columns), dtype=c_dtype)
                elif c_dtype == 'p':
                    c_dtype = 'O'
                    recons.columns = np.array(
                        lmap(Timestamp, to_datetime(recons.columns)),
                        dtype=c_dtype)
                    df.columns = np.array(
                        lmap(Timestamp, df.columns.to_timestamp()),
                        dtype=c_dtype)
                else:
                    c_dtype = type_map.get(c_dtype)
                    recons.columns = np.array(recons.columns, dtype=c_dtype)
                    df.columns = np.array(df.columns, dtype=c_dtype)

            assert_frame_equal(df, recons, check_names=False,
                               check_less_precise=True)

        N = 100
        chunksize = 1000

        for ncols in [4]:
            base = int((chunksize // ncols or 1) or 1)
            for nrows in [2, 10, N - 1, N, N + 1, N + 2, 2 * N - 2,
                          2 * N - 1, 2 * N, 2 * N + 1, 2 * N + 2,
                          base - 1, base, base + 1]:
                _do_test(mkdf(nrows, ncols, r_idx_type='dt',
                              c_idx_type='s'), 'dt', 's')

        for ncols in [4]:
            base = int((chunksize // ncols or 1) or 1)
            for nrows in [2, 10, N - 1, N, N + 1, N + 2, 2 * N - 2,
                          2 * N - 1, 2 * N, 2 * N + 1, 2 * N + 2,
                          base - 1, base, base + 1]:
                _do_test(mkdf(nrows, ncols, r_idx_type='dt',
                              c_idx_type='s'), 'dt', 's')
                pass

        for r_idx_type, c_idx_type in [('i', 'i'), ('s', 's'), ('u', 'dt'),
                                       ('p', 'p')]:
            for ncols in [1, 2, 3, 4]:
                base = int((chunksize // ncols or 1) or 1)
                for nrows in [2, 10, N - 1, N, N + 1, N + 2, 2 * N - 2,
                              2 * N - 1, 2 * N, 2 * N + 1, 2 * N + 2,
                              base - 1, base, base + 1]:
                    _do_test(mkdf(nrows, ncols, r_idx_type=r_idx_type,
                                  c_idx_type=c_idx_type),
                             r_idx_type, c_idx_type)

        for ncols in [1, 2, 3, 4]:
            base = int((chunksize // ncols or 1) or 1)
            for nrows in [10, N - 2, N - 1, N, N + 1, N + 2, 2 * N - 2,
                          2 * N - 1, 2 * N, 2 * N + 1, 2 * N + 2,
                          base - 1, base, base + 1]:
                _do_test(mkdf(nrows, ncols))

        for nrows in [10, N - 2, N - 1, N, N + 1, N + 2]:
            df = mkdf(nrows, 3)
            cols = list(df.columns)
            cols[:2] = ["dupe", "dupe"]
            cols[-2:] = ["dupe", "dupe"]
            ix = list(df.index)
            ix[:2] = ["rdupe", "rdupe"]
            ix[-2:] = ["rdupe", "rdupe"]
            df.index = ix
            df.columns = cols
            _do_test(df, dupe_col=True)

        _do_test(DataFrame(index=lrange(10)))
        _do_test(mkdf(chunksize // 2 + 1, 2, r_idx_nlevels=2), rnlvl=2)
        for ncols in [2, 3, 4]:
            base = int(chunksize // ncols)
            for nrows in [10, N - 2, N - 1, N, N + 1, N + 2, 2 * N - 2,
                          2 * N - 1, 2 * N, 2 * N + 1, 2 * N + 2,
                          base - 1, base, base + 1]:
                _do_test(mkdf(nrows, ncols, r_idx_nlevels=2), rnlvl=2)
                _do_test(mkdf(nrows, ncols, c_idx_nlevels=2), cnlvl=2)
                _do_test(mkdf(nrows, ncols, r_idx_nlevels=2, c_idx_nlevels=2),
                         rnlvl=2, cnlvl=2)

    def test_to_csv_from_csv_w_some_infs(self):

        # test roundtrip with inf, -inf, nan, as full columns and mix
        self.frame['G'] = np.nan
        f = lambda x: [np.inf, np.nan][np.random.rand() < .5]
        self.frame['H'] = self.frame.index.map(f)

        with ensure_clean() as path:
            self.frame.to_csv(path)
            recons = self.read_csv(path)

            # TODO to_csv drops column name
            assert_frame_equal(self.frame, recons, check_names=False)
            assert_frame_equal(np.isinf(self.frame),
                               np.isinf(recons), check_names=False)

    def test_to_csv_from_csv_w_all_infs(self):

        # test roundtrip with inf, -inf, nan, as full columns and mix
        self.frame['E'] = np.inf
        self.frame['F'] = -np.inf

        with ensure_clean() as path:
            self.frame.to_csv(path)
            recons = self.read_csv(path)

            # TODO to_csv drops column name
            assert_frame_equal(self.frame, recons, check_names=False)
            assert_frame_equal(np.isinf(self.frame),
                               np.isinf(recons), check_names=False)

    def test_to_csv_no_index(self):
        # GH 3624, after appending columns, to_csv fails
        with ensure_clean('__tmp_to_csv_no_index__') as path:
            df = DataFrame({'c1': [1, 2, 3], 'c2': [4, 5, 6]})
            df.to_csv(path, index=False)
            result = read_csv(path)
            assert_frame_equal(df, result)
            df['c3'] = Series([7, 8, 9], dtype='int64')
            df.to_csv(path, index=False)
            result = read_csv(path)
            assert_frame_equal(df, result)

    def test_to_csv_with_mix_columns(self):
        # gh-11637: incorrect output when a mix of integer and string column
        # names passed as columns parameter in to_csv

        df = DataFrame({0: ['a', 'b', 'c'],
                        1: ['aa', 'bb', 'cc']})
        df['test'] = 'txt'
        assert df.to_csv() == df.to_csv(columns=[0, 1, 'test'])

    def test_to_csv_headers(self):
        # GH6186, the presence or absence of `index` incorrectly
        # causes to_csv to have different header semantics.
        from_df = DataFrame([[1, 2], [3, 4]], columns=['A', 'B'])
        to_df = DataFrame([[1, 2], [3, 4]], columns=['X', 'Y'])
        with ensure_clean('__tmp_to_csv_headers__') as path:
            from_df.to_csv(path, header=['X', 'Y'])
            recons = self.read_csv(path)

            assert_frame_equal(to_df, recons)

            from_df.to_csv(path, index=False, header=['X', 'Y'])
            recons = self.read_csv(path)

            recons.reset_index(inplace=True)
            assert_frame_equal(to_df, recons)

    def test_to_csv_multiindex(self):

        frame = self.frame
        old_index = frame.index
        arrays = np.arange(len(old_index) * 2).reshape(2, -1)
        new_index = MultiIndex.from_arrays(arrays, names=['first', 'second'])
        frame.index = new_index

        with ensure_clean('__tmp_to_csv_multiindex__') as path:

            frame.to_csv(path, header=False)
            frame.to_csv(path, columns=['A', 'B'])

            # round trip
            frame.to_csv(path)

            df = self.read_csv(path, index_col=[0, 1],
                               parse_dates=False)

            # TODO to_csv drops column name
            assert_frame_equal(frame, df, check_names=False)
            assert frame.index.names == df.index.names

            # needed if setUp becomes a class method
            self.frame.index = old_index

            # try multiindex with dates
            tsframe = self.tsframe
            old_index = tsframe.index
            new_index = [old_index, np.arange(len(old_index))]
            tsframe.index = MultiIndex.from_arrays(new_index)

            tsframe.to_csv(path, index_label=['time', 'foo'])
            recons = self.read_csv(path, index_col=[0, 1])

            # TODO to_csv drops column name
            assert_frame_equal(tsframe, recons, check_names=False)

            # do not load index
            tsframe.to_csv(path)
            recons = self.read_csv(path, index_col=None)
            assert len(recons.columns) == len(tsframe.columns) + 2

            # no index
            tsframe.to_csv(path, index=False)
            recons = self.read_csv(path, index_col=None)
            assert_almost_equal(recons.values, self.tsframe.values)

            # needed if setUp becomes class method
            self.tsframe.index = old_index

        with ensure_clean('__tmp_to_csv_multiindex__') as path:
            # GH3571, GH1651, GH3141

            def _make_frame(names=None):
                if names is True:
                    names = ['first', 'second']
                return DataFrame(np.random.randint(0, 10, size=(3, 3)),
                                 columns=MultiIndex.from_tuples(
                                     [('bah', 'foo'),
                                      ('bah', 'bar'),
                                      ('ban', 'baz')], names=names),
                                 dtype='int64')

            # column & index are multi-index
            df = mkdf(5, 3, r_idx_nlevels=2, c_idx_nlevels=4)
            df.to_csv(path)
            result = read_csv(path, header=[0, 1, 2, 3],
                              index_col=[0, 1])
            assert_frame_equal(df, result)

            # column is mi
            df = mkdf(5, 3, r_idx_nlevels=1, c_idx_nlevels=4)
            df.to_csv(path)
            result = read_csv(
                path, header=[0, 1, 2, 3], index_col=0)
            assert_frame_equal(df, result)

            # dup column names?
            df = mkdf(5, 3, r_idx_nlevels=3, c_idx_nlevels=4)
            df.to_csv(path)
            result = read_csv(path, header=[0, 1, 2, 3],
                              index_col=[0, 1, 2])
            assert_frame_equal(df, result)

            # writing with no index
            df = _make_frame()
            df.to_csv(path, index=False)
            result = read_csv(path, header=[0, 1])
            assert_frame_equal(df, result)

            # we lose the names here
            df = _make_frame(True)
            df.to_csv(path, index=False)
            result = read_csv(path, header=[0, 1])
            assert com._all_none(*result.columns.names)
            result.columns.names = df.columns.names
            assert_frame_equal(df, result)

            # tupleize_cols=True and index=False
            df = _make_frame(True)
            with tm.assert_produces_warning(FutureWarning):
                df.to_csv(path, tupleize_cols=True, index=False)

            with tm.assert_produces_warning(FutureWarning,
                                            check_stacklevel=False):
                result = read_csv(path, header=0,
                                  tupleize_cols=True,
                                  index_col=None)
            result.columns = df.columns
            assert_frame_equal(df, result)

            # whatsnew example
            df = _make_frame()
            df.to_csv(path)
            result = read_csv(path, header=[0, 1],
                              index_col=[0])
            assert_frame_equal(df, result)

            df = _make_frame(True)
            df.to_csv(path)
            result = read_csv(path, header=[0, 1],
                              index_col=[0])
            assert_frame_equal(df, result)

            # column & index are multi-index (compatibility)
            df = mkdf(5, 3, r_idx_nlevels=2, c_idx_nlevels=4)
            with tm.assert_produces_warning(FutureWarning):
                df.to_csv(path, tupleize_cols=True)

            with tm.assert_produces_warning(FutureWarning,
                                            check_stacklevel=False):
                result = read_csv(path, header=0, index_col=[0, 1],
                                  tupleize_cols=True)
            result.columns = df.columns
            assert_frame_equal(df, result)

            # invalid options
            df = _make_frame(True)
            df.to_csv(path)

            for i in [6, 7]:
                msg = 'len of {i}, but only 5 lines in file'.format(i=i)
                with tm.assert_raises_regex(ParserError, msg):
                    read_csv(path, header=lrange(i), index_col=0)

            # write with cols
            with tm.assert_raises_regex(TypeError, 'cannot specify cols '
                                        'with a MultiIndex'):
                df.to_csv(path, columns=['foo', 'bar'])

        with ensure_clean('__tmp_to_csv_multiindex__') as path:
            # empty
            tsframe[:0].to_csv(path)
            recons = self.read_csv(path)

            exp = tsframe[:0]
            exp.index = []

            tm.assert_index_equal(recons.columns, exp.columns)
            assert len(recons) == 0

    def test_to_csv_float32_nanrep(self):
        df = DataFrame(np.random.randn(1, 4).astype(np.float32))
        df[1] = np.nan

        with ensure_clean('__tmp_to_csv_float32_nanrep__.csv') as path:
            df.to_csv(path, na_rep=999)

            with open(path) as f:
                lines = f.readlines()
                assert lines[1].split(',')[2] == '999'

    def test_to_csv_withcommas(self):

        # Commas inside fields should be correctly escaped when saving as CSV.
        df = DataFrame({'A': [1, 2, 3], 'B': ['5,6', '7,8', '9,0']})

        with ensure_clean('__tmp_to_csv_withcommas__.csv') as path:
            df.to_csv(path)
            df2 = self.read_csv(path)
            assert_frame_equal(df2, df)

    def test_to_csv_mixed(self):

        def create_cols(name):
            return ["%s%03d" % (name, i) for i in range(5)]

        df_float = DataFrame(np.random.randn(
            100, 5), dtype='float64', columns=create_cols('float'))
        df_int = DataFrame(np.random.randn(100, 5),
                           dtype='int64', columns=create_cols('int'))
        df_bool = DataFrame(True, index=df_float.index,
                            columns=create_cols('bool'))
        df_object = DataFrame('foo', index=df_float.index,
                              columns=create_cols('object'))
        df_dt = DataFrame(Timestamp('20010101'),
                          index=df_float.index, columns=create_cols('date'))

        # add in some nans
        df_float.loc[30:50, 1:3] = np.nan

        # ## this is a bug in read_csv right now ####
        # df_dt.loc[30:50,1:3] = np.nan

        df = pd.concat([df_float, df_int, df_bool, df_object, df_dt], axis=1)

        # dtype
        dtypes = dict()
        for n, dtype in [('float', np.float64), ('int', np.int64),
                         ('bool', np.bool), ('object', np.object)]:
            for c in create_cols(n):
                dtypes[c] = dtype

        with ensure_clean() as filename:
            df.to_csv(filename)
            rs = read_csv(filename, index_col=0, dtype=dtypes,
                          parse_dates=create_cols('date'))
            assert_frame_equal(rs, df)

    def test_to_csv_dups_cols(self):

        df = DataFrame(np.random.randn(1000, 30), columns=lrange(
            15) + lrange(15), dtype='float64')

        with ensure_clean() as filename:
            df.to_csv(filename)  # single dtype, fine
            result = read_csv(filename, index_col=0)
            result.columns = df.columns
            assert_frame_equal(result, df)

        df_float = DataFrame(np.random.randn(1000, 3), dtype='float64')
        df_int = DataFrame(np.random.randn(1000, 3), dtype='int64')
        df_bool = DataFrame(True, index=df_float.index, columns=lrange(3))
        df_object = DataFrame('foo', index=df_float.index, columns=lrange(3))
        df_dt = DataFrame(Timestamp('20010101'),
                          index=df_float.index, columns=lrange(3))
        df = pd.concat([df_float, df_int, df_bool, df_object,
                        df_dt], axis=1, ignore_index=True)

        cols = []
        for i in range(5):
            cols.extend([0, 1, 2])
        df.columns = cols

        with ensure_clean() as filename:
            df.to_csv(filename)
            result = read_csv(filename, index_col=0)

            # date cols
            for i in ['0.4', '1.4', '2.4']:
                result[i] = to_datetime(result[i])

            result.columns = df.columns
            assert_frame_equal(result, df)

        # GH3457
        from pandas.util.testing import makeCustomDataframe as mkdf

        N = 10
        df = mkdf(N, 3)
        df.columns = ['a', 'a', 'b']

        with ensure_clean() as filename:
            df.to_csv(filename)

            # read_csv will rename the dups columns
            result = read_csv(filename, index_col=0)
            result = result.rename(columns={'a.1': 'a'})
            assert_frame_equal(result, df)

    def test_to_csv_chunking(self):

        aa = DataFrame({'A': lrange(100000)})
        aa['B'] = aa.A + 1.0
        aa['C'] = aa.A + 2.0
        aa['D'] = aa.A + 3.0

        for chunksize in [10000, 50000, 100000]:
            with ensure_clean() as filename:
                aa.to_csv(filename, chunksize=chunksize)
                rs = read_csv(filename, index_col=0)
                assert_frame_equal(rs, aa)

    @pytest.mark.slow
    def test_to_csv_wide_frame_formatting(self):
        # Issue #8621
        df = DataFrame(np.random.randn(1, 100010), columns=None, index=None)
        with ensure_clean() as filename:
            df.to_csv(filename, header=False, index=False)
            rs = read_csv(filename, header=None)
            assert_frame_equal(rs, df)

    def test_to_csv_bug(self):
        f1 = StringIO('a,1.0\nb,2.0')
        df = self.read_csv(f1, header=None)
        newdf = DataFrame({'t': df[df.columns[0]]})

        with ensure_clean() as path:
            newdf.to_csv(path)

            recons = read_csv(path, index_col=0)
            # don't check_names as t != 1
            assert_frame_equal(recons, newdf, check_names=False)

    def test_to_csv_unicode(self):

        df = DataFrame({u('c/\u03c3'): [1, 2, 3]})
        with ensure_clean() as path:

            df.to_csv(path, encoding='UTF-8')
            df2 = read_csv(path, index_col=0, encoding='UTF-8')
            assert_frame_equal(df, df2)

            df.to_csv(path, encoding='UTF-8', index=False)
            df2 = read_csv(path, index_col=None, encoding='UTF-8')
            assert_frame_equal(df, df2)

    def test_to_csv_unicode_index_col(self):
        buf = StringIO('')
        df = DataFrame(
            [[u("\u05d0"), "d2", "d3", "d4"], ["a1", "a2", "a3", "a4"]],
            columns=[u("\u05d0"),
                     u("\u05d1"), u("\u05d2"), u("\u05d3")],
            index=[u("\u05d0"), u("\u05d1")])

        df.to_csv(buf, encoding='UTF-8')
        buf.seek(0)

        df2 = read_csv(buf, index_col=0, encoding='UTF-8')
        assert_frame_equal(df, df2)

    def test_to_csv_stringio(self):
        buf = StringIO()
        self.frame.to_csv(buf)
        buf.seek(0)
        recons = read_csv(buf, index_col=0)
        # TODO to_csv drops column name
        assert_frame_equal(recons, self.frame, check_names=False)

    def test_to_csv_float_format(self):

        df = DataFrame([[0.123456, 0.234567, 0.567567],
                        [12.32112, 123123.2, 321321.2]],
                       index=['A', 'B'], columns=['X', 'Y', 'Z'])

        with ensure_clean() as filename:

            df.to_csv(filename, float_format='%.2f')

            rs = read_csv(filename, index_col=0)
            xp = DataFrame([[0.12, 0.23, 0.57],
                            [12.32, 123123.20, 321321.20]],
                           index=['A', 'B'], columns=['X', 'Y', 'Z'])
            assert_frame_equal(rs, xp)

    def test_to_csv_unicodewriter_quoting(self):
        df = DataFrame({'A': [1, 2, 3], 'B': ['foo', 'bar', 'baz']})

        buf = StringIO()
        df.to_csv(buf, index=False, quoting=csv.QUOTE_NONNUMERIC,
                  encoding='utf-8')

        result = buf.getvalue()
        expected = ('"A","B"\n'
                    '1,"foo"\n'
                    '2,"bar"\n'
                    '3,"baz"\n')

        assert result == expected

    def test_to_csv_quote_none(self):
        # GH4328
        df = DataFrame({'A': ['hello', '{"hello"}']})
        for encoding in (None, 'utf-8'):
            buf = StringIO()
            df.to_csv(buf, quoting=csv.QUOTE_NONE,
                      encoding=encoding, index=False)
            result = buf.getvalue()
            expected = 'A\nhello\n{"hello"}\n'
            assert result == expected

    def test_to_csv_index_no_leading_comma(self):
        df = DataFrame({'A': [1, 2, 3], 'B': [4, 5, 6]},
                       index=['one', 'two', 'three'])

        buf = StringIO()
        df.to_csv(buf, index_label=False)
        expected = ('A,B\n'
                    'one,1,4\n'
                    'two,2,5\n'
                    'three,3,6\n')
        assert buf.getvalue() == expected

    def test_to_csv_line_terminators(self):
        df = DataFrame({'A': [1, 2, 3], 'B': [4, 5, 6]},
                       index=['one', 'two', 'three'])

        buf = StringIO()
        df.to_csv(buf, line_terminator='\r\n')
        expected = (',A,B\r\n'
                    'one,1,4\r\n'
                    'two,2,5\r\n'
                    'three,3,6\r\n')
        assert buf.getvalue() == expected

        buf = StringIO()
        df.to_csv(buf)  # The default line terminator remains \n
        expected = (',A,B\n'
                    'one,1,4\n'
                    'two,2,5\n'
                    'three,3,6\n')
        assert buf.getvalue() == expected

    def test_to_csv_from_csv_categorical(self):

        # CSV with categoricals should result in the same output as when one
        # would add a "normal" Series/DataFrame.
        s = Series(pd.Categorical(['a', 'b', 'b', 'a', 'a', 'c', 'c', 'c']))
        s2 = Series(['a', 'b', 'b', 'a', 'a', 'c', 'c', 'c'])
        res = StringIO()
        s.to_csv(res)
        exp = StringIO()
        s2.to_csv(exp)
        assert res.getvalue() == exp.getvalue()

        df = DataFrame({"s": s})
        df2 = DataFrame({"s": s2})
        res = StringIO()
        df.to_csv(res)
        exp = StringIO()
        df2.to_csv(exp)
        assert res.getvalue() == exp.getvalue()

    def test_to_csv_path_is_none(self):
        # GH 8215
        # Make sure we return string for consistency with
        # Series.to_csv()
        csv_str = self.frame.to_csv(path_or_buf=None)
        assert isinstance(csv_str, str)
        recons = pd.read_csv(StringIO(csv_str), index_col=0)
        assert_frame_equal(self.frame, recons)

    @pytest.mark.parametrize('df,encoding', [
        (DataFrame([[0.123456, 0.234567, 0.567567],
                    [12.32112, 123123.2, 321321.2]],
                   index=['A', 'B'], columns=['X', 'Y', 'Z']), None),
        # GH 21241, 21118
        (DataFrame([['abc', 'def', 'ghi']], columns=['X', 'Y', 'Z']), 'ascii'),
        (DataFrame(5 * [[123, u"你好", u"世界"]],
                   columns=['X', 'Y', 'Z']), 'gb2312'),
        (DataFrame(5 * [[123, u"Γειά σου", u"Κόσμε"]],
                   columns=['X', 'Y', 'Z']), 'cp737')
    ])
    def test_to_csv_compression(self, df, encoding, compression):

        with ensure_clean() as filename:

            df.to_csv(filename, compression=compression, encoding=encoding)
            # test the round trip - to_csv -> read_csv
            result = read_csv(filename, compression=compression,
                              index_col=0, encoding=encoding)
            assert_frame_equal(df, result)

            # test the round trip using file handle - to_csv -> read_csv
            f, _handles = _get_handle(filename, 'w', compression=compression,
                                      encoding=encoding)
            with f:
                df.to_csv(f, encoding=encoding)
            result = pd.read_csv(filename, compression=compression,
                                 encoding=encoding, index_col=0, squeeze=True)
            assert_frame_equal(df, result)

            # explicitly make sure file is compressed
            with tm.decompress_file(filename, compression) as fh:
                text = fh.read().decode(encoding or 'utf8')
                for col in df.columns:
                    assert col in text

            with tm.decompress_file(filename, compression) as fh:
                assert_frame_equal(df, read_csv(fh,
                                                index_col=0,
                                                encoding=encoding))

    def test_to_csv_date_format(self):
        with ensure_clean('__tmp_to_csv_date_format__') as path:
            dt_index = self.tsframe.index
            datetime_frame = DataFrame(
                {'A': dt_index, 'B': dt_index.shift(1)}, index=dt_index)
            datetime_frame.to_csv(path, date_format='%Y%m%d')

            # Check that the data was put in the specified format
            test = read_csv(path, index_col=0)

            datetime_frame_int = datetime_frame.applymap(
                lambda x: int(x.strftime('%Y%m%d')))
            datetime_frame_int.index = datetime_frame_int.index.map(
                lambda x: int(x.strftime('%Y%m%d')))

            assert_frame_equal(test, datetime_frame_int)

            datetime_frame.to_csv(path, date_format='%Y-%m-%d')

            # Check that the data was put in the specified format
            test = read_csv(path, index_col=0)
            datetime_frame_str = datetime_frame.applymap(
                lambda x: x.strftime('%Y-%m-%d'))
            datetime_frame_str.index = datetime_frame_str.index.map(
                lambda x: x.strftime('%Y-%m-%d'))

            assert_frame_equal(test, datetime_frame_str)

            # Check that columns get converted
            datetime_frame_columns = datetime_frame.T
            datetime_frame_columns.to_csv(path, date_format='%Y%m%d')

            test = read_csv(path, index_col=0)

            datetime_frame_columns = datetime_frame_columns.applymap(
                lambda x: int(x.strftime('%Y%m%d')))
            # Columns don't get converted to ints by read_csv
            datetime_frame_columns.columns = (
                datetime_frame_columns.columns
                .map(lambda x: x.strftime('%Y%m%d')))

            assert_frame_equal(test, datetime_frame_columns)

            # test NaTs
            nat_index = to_datetime(
                ['NaT'] * 10 + ['2000-01-01', '1/1/2000', '1-1-2000'])
            nat_frame = DataFrame({'A': nat_index}, index=nat_index)
            nat_frame.to_csv(path, date_format='%Y-%m-%d')

            test = read_csv(path, parse_dates=[0, 1], index_col=0)

            assert_frame_equal(test, nat_frame)

    def test_to_csv_with_dst_transitions(self):

        with ensure_clean('csv_date_format_with_dst') as path:
            # make sure we are not failing on transitions
            times = pd.date_range("2013-10-26 23:00", "2013-10-27 01:00",
                                  tz="Europe/London",
                                  freq="H",
                                  ambiguous='infer')

            for i in [times, times + pd.Timedelta('10s')]:
                time_range = np.array(range(len(i)), dtype='int64')
                df = DataFrame({'A': time_range}, index=i)
                df.to_csv(path, index=True)

                # we have to reconvert the index as we
                # don't parse the tz's
                result = read_csv(path, index_col=0)
                result.index = to_datetime(result.index).tz_localize(
                    'UTC').tz_convert('Europe/London')
                assert_frame_equal(result, df)

        # GH11619
        idx = pd.date_range('2015-01-01', '2015-12-31',
                            freq='H', tz='Europe/Paris')
        df = DataFrame({'values': 1, 'idx': idx},
                       index=idx)
        with ensure_clean('csv_date_format_with_dst') as path:
            df.to_csv(path, index=True)
            result = read_csv(path, index_col=0)
            result.index = to_datetime(result.index).tz_localize(
                'UTC').tz_convert('Europe/Paris')
            result['idx'] = to_datetime(result['idx']).astype(
                'datetime64[ns, Europe/Paris]')
            assert_frame_equal(result, df)

        # assert working
        df.astype(str)

        with ensure_clean('csv_date_format_with_dst') as path:
            df.to_pickle(path)
            result = pd.read_pickle(path)
            assert_frame_equal(result, df)

    def test_to_csv_quoting(self):
        df = DataFrame({
            'c_bool': [True, False],
            'c_float': [1.0, 3.2],
            'c_int': [42, np.nan],
            'c_string': ['a', 'b,c'],
        })

        expected = """\
,c_bool,c_float,c_int,c_string
0,True,1.0,42.0,a
1,False,3.2,,"b,c"
"""
        result = df.to_csv()
        assert result == expected

        result = df.to_csv(quoting=None)
        assert result == expected

        result = df.to_csv(quoting=csv.QUOTE_MINIMAL)
        assert result == expected

        expected = """\
"","c_bool","c_float","c_int","c_string"
"0","True","1.0","42.0","a"
"1","False","3.2","","b,c"
"""
        result = df.to_csv(quoting=csv.QUOTE_ALL)
        assert result == expected

        # see gh-12922, gh-13259: make sure changes to
        # the formatters do not break this behaviour
        expected = """\
"","c_bool","c_float","c_int","c_string"
0,True,1.0,42.0,"a"
1,False,3.2,"","b,c"
"""
        result = df.to_csv(quoting=csv.QUOTE_NONNUMERIC)
        assert result == expected

        msg = "need to escape, but no escapechar set"
        tm.assert_raises_regex(csv.Error, msg, df.to_csv,
                               quoting=csv.QUOTE_NONE)
        tm.assert_raises_regex(csv.Error, msg, df.to_csv,
                               quoting=csv.QUOTE_NONE,
                               escapechar=None)

        expected = """\
,c_bool,c_float,c_int,c_string
0,True,1.0,42.0,a
1,False,3.2,,b!,c
"""
        result = df.to_csv(quoting=csv.QUOTE_NONE,
                           escapechar='!')
        assert result == expected

        expected = """\
,c_bool,c_ffloat,c_int,c_string
0,True,1.0,42.0,a
1,False,3.2,,bf,c
"""
        result = df.to_csv(quoting=csv.QUOTE_NONE,
                           escapechar='f')
        assert result == expected

        # see gh-3503: quoting Windows line terminators
        # presents with encoding?
        text = 'a,b,c\n1,"test \r\n",3\n'
        df = pd.read_csv(StringIO(text))
        buf = StringIO()
        df.to_csv(buf, encoding='utf-8', index=False)
        assert buf.getvalue() == text

        # xref gh-7791: make sure the quoting parameter is passed through
        # with multi-indexes
        df = pd.DataFrame({'a': [1, 2], 'b': [3, 4], 'c': [5, 6]})
        df = df.set_index(['a', 'b'])
        expected = '"a","b","c"\n"1","3","5"\n"2","4","6"\n'
        assert df.to_csv(quoting=csv.QUOTE_ALL) == expected

    def test_period_index_date_overflow(self):
        # see gh-15982

        dates = ["1990-01-01", "2000-01-01", "3005-01-01"]
        index = pd.PeriodIndex(dates, freq="D")

        df = pd.DataFrame([4, 5, 6], index=index)
        result = df.to_csv()

        expected = ',0\n1990-01-01,4\n2000-01-01,5\n3005-01-01,6\n'
        assert result == expected

        date_format = "%m-%d-%Y"
        result = df.to_csv(date_format=date_format)

        expected = ',0\n01-01-1990,4\n01-01-2000,5\n01-01-3005,6\n'
        assert result == expected

        # Overflow with pd.NaT
        dates = ["1990-01-01", pd.NaT, "3005-01-01"]
        index = pd.PeriodIndex(dates, freq="D")

        df = pd.DataFrame([4, 5, 6], index=index)
        result = df.to_csv()

        expected = ',0\n1990-01-01,4\n,5\n3005-01-01,6\n'
        assert result == expected

    def test_multi_index_header(self):
        # see gh-5539
        columns = pd.MultiIndex.from_tuples([("a", 1), ("a", 2),
                                             ("b", 1), ("b", 2)])
        df = pd.DataFrame([[1, 2, 3, 4], [5, 6, 7, 8]])
        df.columns = columns

        header = ["a", "b", "c", "d"]
        result = df.to_csv(header=header)

        expected = ",a,b,c,d\n0,1,2,3,4\n1,5,6,7,8\n"
        assert result == expected