parsers.py

"""
Module contains tools for processing files into DataFrames or other objects
"""

from __future__ import print_function

from collections import defaultdict
import csv
import datetime
import re
import sys
from textwrap import fill
import warnings

import numpy as np

import pandas._libs.lib as lib
import pandas._libs.ops as libops
import pandas._libs.parsers as parsers
from pandas._libs.tslibs import parsing
import pandas.compat as compat
from pandas.compat import (
    PY3, StringIO, lrange, lzip, map, range, string_types, u, zip)
from pandas.errors import (
    AbstractMethodError, EmptyDataError, ParserError, ParserWarning)
from pandas.util._decorators import Appender

from pandas.core.dtypes.cast import astype_nansafe
from pandas.core.dtypes.common import (
    ensure_object, is_bool_dtype, is_categorical_dtype, is_dtype_equal,
    is_extension_array_dtype, is_float, is_integer, is_integer_dtype,
    is_list_like, is_object_dtype, is_scalar, is_string_dtype, pandas_dtype)
from pandas.core.dtypes.dtypes import CategoricalDtype
from pandas.core.dtypes.missing import isna

from pandas.core import algorithms
from pandas.core.arrays import Categorical
from pandas.core.frame import DataFrame
from pandas.core.index import (
    Index, MultiIndex, RangeIndex, ensure_index_from_sequences)
from pandas.core.series import Series
from pandas.core.tools import datetimes as tools

from pandas.io.common import (
    _NA_VALUES, BaseIterator, UnicodeReader, UTF8Recoder, _get_handle,
    _infer_compression, _validate_header_arg, get_filepath_or_buffer,
    is_file_like)
from pandas.io.date_converters import generic_parser

# BOM character (byte order mark)
# This exists at the beginning of a file to indicate endianness
# of a file (stream). Unfortunately, this marker screws up parsing,
# so we need to remove it if we see it.
_BOM = u('\ufeff')

_doc_read_csv_and_table = r"""
{summary}

Also supports optionally iterating or breaking of the file
into chunks.

Additional help can be found in the online docs for
`IO Tools <http://pandas.pydata.org/pandas-docs/stable/io.html>`_.

Parameters
----------
filepath_or_buffer : str, path object, or file-like object
    Any valid string path is acceptable. The string could be a URL. Valid
    URL schemes include http, ftp, s3, and file. For file URLs, a host is
    expected. A local file could be: file://localhost/path/to/table.csv.

    If you want to pass in a path object, pandas accepts either
    ``pathlib.Path`` or ``py._path.local.LocalPath``.

    By file-like object, we refer to objects with a ``read()`` method, such as
    a file handler (e.g. via builtin ``open`` function) or ``StringIO``.
sep : str, default {_default_sep}
    Delimiter to use. If sep is None, the C engine cannot automatically detect
    the separator, but the Python parsing engine can, meaning the latter will
    be used and automatically detect the separator by Python's builtin sniffer
    tool, ``csv.Sniffer``. In addition, separators longer than 1 character and
    different from ``'\s+'`` will be interpreted as regular expressions and
    will also force the use of the Python parsing engine. Note that regex
    delimiters are prone to ignoring quoted data. Regex example: ``'\r\t'``.
delimiter : str, default ``None``
    Alias for sep.
header : int, list of int, default 'infer'
    Row number(s) to use as the column names, and the start of the
    data.  Default behavior is to infer the column names: if no names
    are passed the behavior is identical to ``header=0`` and column
    names are inferred from the first line of the file, if column
    names are passed explicitly then the behavior is identical to
    ``header=None``. Explicitly pass ``header=0`` to be able to
    replace existing names. The header can be a list of integers that
    specify row locations for a multi-index on the columns
    e.g. [0,1,3]. Intervening rows that are not specified will be
    skipped (e.g. 2 in this example is skipped). Note that this
    parameter ignores commented lines and empty lines if
    ``skip_blank_lines=True``, so ``header=0`` denotes the first line of
    data rather than the first line of the file.
names : array-like, optional
    List of column names to use. If file contains no header row, then you
    should explicitly pass ``header=None``. Duplicates in this list will cause
    a ``UserWarning`` to be issued.
index_col : int/string or sequence of int/string or ``False``, default ``None``
  Column(s) to use as the row labels of the ``DataFrame``, either given as
  string name or column index. If a sequence of int/string is given, a
  MultiIndex is used. Columns used for the index (row names) are dropped from
  the actual columns of the input dataframe. They are accessible via
  ``.index``. (Note: ``index_col=False`` can be used to force pandas to *not*
  use the first column as the index, e.g. when you have a malformed file with
  delimiters at the end of each line.)
usecols : list-like or callable, optional
    Return a subset of the columns. If list-like, all elements must either
    be positional (i.e. integer indices into the document columns) or strings
    that correspond to column names provided either by the user in `names` or
    inferred from the document header row(s). For example, a valid list-like
    `usecols` parameter would be ``[0, 1, 2]`` or ``['foo', 'bar', 'baz']``.
    Element order is ignored, so ``usecols=[0, 1]`` is the same as ``[1, 0]``.
    To instantiate a DataFrame from ``data`` with element order preserved use
    ``pd.read_csv(data, usecols=['foo', 'bar'])[['foo', 'bar']]`` for columns
    in ``['foo', 'bar']`` order or
    ``pd.read_csv(data, usecols=['foo', 'bar'])[['bar', 'foo']]``
    for ``['bar', 'foo']`` order.

    If callable, the callable function will be evaluated against the column
    names, returning names where the callable function evaluates to True. An
    example of a valid callable argument would be ``lambda x: x.upper() in
    ['AAA', 'BBB', 'DDD']``. Using this parameter results in much faster
    parsing time and lower memory usage.
squeeze : bool, default False
    If the parsed data only contains one column then return a Series.
prefix : str, optional
    Prefix to add to column numbers when no header, e.g. 'X' for X0, X1, ...
mangle_dupe_cols : bool, default True
    Duplicate columns will be specified as 'X', 'X.1', ...'X.N', rather than
    'X'...'X'. Passing in False will cause data to be overwritten if there
    are duplicate names in the columns.
dtype : Type name or dict of column -> type, optional
    Data type for data or columns. E.g. {{'a': np.float64, 'b': np.int32,
    'c': 'Int64'}}
    Use `str` or `object` together with suitable `na_values` settings
    to preserve and not interpret dtype.
    If converters are specified, they will be applied INSTEAD
    of dtype conversion.
engine : {{'c', 'python'}}, optional
    Parser engine to use. The C engine is faster while the python engine is
    currently more feature-complete.
converters : dict, optional
    Dict of functions for converting values in certain columns. Keys can either
    be integers or column labels.
true_values : list, optional
    Values to consider as True.
false_values : list, optional
    Values to consider as False.
skipinitialspace : bool, default False
    Skip spaces after delimiter.
skiprows : list-like, int or callable, optional
    Line numbers to skip (0-indexed) or number of lines to skip (int)
    at the start of the file.

    If callable, the callable function will be evaluated against the row
    indices, returning True if the row should be skipped and False otherwise.
    An example of a valid callable argument would be ``lambda x: x in [0, 2]``.
skipfooter : int, default 0
    Number of lines at bottom of file to skip (Unsupported with engine='c').
nrows : int, optional
    Number of rows of file to read. Useful for reading pieces of large files.
na_values : scalar, str, list-like, or dict, optional
    Additional strings to recognize as NA/NaN. If dict passed, specific
    per-column NA values.  By default the following values are interpreted as
    NaN: '""" + fill("', '".join(sorted(_NA_VALUES)),
                     70, subsequent_indent="    ") + """'.
keep_default_na : bool, default True
    Whether or not to include the default NaN values when parsing the data.
    Depending on whether `na_values` is passed in, the behavior is as follows:

    * If `keep_default_na` is True, and `na_values` are specified, `na_values`
      is appended to the default NaN values used for parsing.
    * If `keep_default_na` is True, and `na_values` are not specified, only
      the default NaN values are used for parsing.
    * If `keep_default_na` is False, and `na_values` are specified, only
      the NaN values specified `na_values` are used for parsing.
    * If `keep_default_na` is False, and `na_values` are not specified, no
      strings will be parsed as NaN.

    Note that if `na_filter` is passed in as False, the `keep_default_na` and
    `na_values` parameters will be ignored.
na_filter : bool, default True
    Detect missing value markers (empty strings and the value of na_values). In
    data without any NAs, passing na_filter=False can improve the performance
    of reading a large file.
verbose : bool, default False
    Indicate number of NA values placed in non-numeric columns.
skip_blank_lines : bool, default True
    If True, skip over blank lines rather than interpreting as NaN values.
parse_dates : bool or list of int or names or list of lists or dict, \
default False
    The behavior is as follows:

    * boolean. If True -> try parsing the index.
    * list of int or names. e.g. If [1, 2, 3] -> try parsing columns 1, 2, 3
      each as a separate date column.
    * list of lists. e.g.  If [[1, 3]] -> combine columns 1 and 3 and parse as
      a single date column.
    * dict, e.g. {{'foo' : [1, 3]}} -> parse columns 1, 3 as date and call
      result 'foo'

    If a column or index cannot be represented as an array of datetimes,
    say because of an unparseable value or a mixture of timezones, the column
    or index will be returned unaltered as an object data type. For
    non-standard datetime parsing, use ``pd.to_datetime`` after
    ``pd.read_csv``. To parse an index or column with a mixture of timezones,
    specify ``date_parser`` to be a partially-applied
    :func:`pandas.to_datetime` with ``utc=True``. See
    :ref:`io.csv.mixed_timezones` for more.

    Note: A fast-path exists for iso8601-formatted dates.
infer_datetime_format : bool, default False
    If True and `parse_dates` is enabled, pandas will attempt to infer the
    format of the datetime strings in the columns, and if it can be inferred,
    switch to a faster method of parsing them. In some cases this can increase
    the parsing speed by 5-10x.
keep_date_col : bool, default False
    If True and `parse_dates` specifies combining multiple columns then
    keep the original columns.
date_parser : function, optional
    Function to use for converting a sequence of string columns to an array of
    datetime instances. The default uses ``dateutil.parser.parser`` to do the
    conversion. Pandas will try to call `date_parser` in three different ways,
    advancing to the next if an exception occurs: 1) Pass one or more arrays
    (as defined by `parse_dates`) as arguments; 2) concatenate (row-wise) the
    string values from the columns defined by `parse_dates` into a single array
    and pass that; and 3) call `date_parser` once for each row using one or
    more strings (corresponding to the columns defined by `parse_dates`) as
    arguments.
dayfirst : bool, default False
    DD/MM format dates, international and European format.
iterator : bool, default False
    Return TextFileReader object for iteration or getting chunks with
    ``get_chunk()``.
chunksize : int, optional
    Return TextFileReader object for iteration.
    See the `IO Tools docs
    <http://pandas.pydata.org/pandas-docs/stable/io.html#io-chunking>`_
    for more information on ``iterator`` and ``chunksize``.
compression : {{'infer', 'gzip', 'bz2', 'zip', 'xz', None}}, default 'infer'
    For on-the-fly decompression of on-disk data. If 'infer' and
    `filepath_or_buffer` is path-like, then detect compression from the
    following extensions: '.gz', '.bz2', '.zip', or '.xz' (otherwise no
    decompression). If using 'zip', the ZIP file must contain only one data
    file to be read in. Set to None for no decompression.

    .. versionadded:: 0.18.1 support for 'zip' and 'xz' compression.

thousands : str, optional
    Thousands separator.
decimal : str, default '.'
    Character to recognize as decimal point (e.g. use ',' for European data).
lineterminator : str (length 1), optional
    Character to break file into lines. Only valid with C parser.
quotechar : str (length 1), optional
    The character used to denote the start and end of a quoted item. Quoted
    items can include the delimiter and it will be ignored.
quoting : int or csv.QUOTE_* instance, default 0
    Control field quoting behavior per ``csv.QUOTE_*`` constants. Use one of
    QUOTE_MINIMAL (0), QUOTE_ALL (1), QUOTE_NONNUMERIC (2) or QUOTE_NONE (3).
doublequote : bool, default ``True``
   When quotechar is specified and quoting is not ``QUOTE_NONE``, indicate
   whether or not to interpret two consecutive quotechar elements INSIDE a
   field as a single ``quotechar`` element.
escapechar : str (length 1), optional
    One-character string used to escape other characters.
comment : str, optional
    Indicates remainder of line should not be parsed. If found at the beginning
    of a line, the line will be ignored altogether. This parameter must be a
    single character. Like empty lines (as long as ``skip_blank_lines=True``),
    fully commented lines are ignored by the parameter `header` but not by
    `skiprows`. For example, if ``comment='#'``, parsing
    ``#empty\\na,b,c\\n1,2,3`` with ``header=0`` will result in 'a,b,c' being
    treated as the header.
encoding : str, optional
    Encoding to use for UTF when reading/writing (ex. 'utf-8'). `List of Python
    standard encodings
    <https://docs.python.org/3/library/codecs.html#standard-encodings>`_ .
dialect : str or csv.Dialect, optional
    If provided, this parameter will override values (default or not) for the
    following parameters: `delimiter`, `doublequote`, `escapechar`,
    `skipinitialspace`, `quotechar`, and `quoting`. If it is necessary to
    override values, a ParserWarning will be issued. See csv.Dialect
    documentation for more details.
tupleize_cols : bool, default False
    Leave a list of tuples on columns as is (default is to convert to
    a MultiIndex on the columns).

    .. deprecated:: 0.21.0
       This argument will be removed and will always convert to MultiIndex

error_bad_lines : bool, default True
    Lines with too many fields (e.g. a csv line with too many commas) will by
    default cause an exception to be raised, and no DataFrame will be returned.
    If False, then these "bad lines" will dropped from the DataFrame that is
    returned.
warn_bad_lines : bool, default True
    If error_bad_lines is False, and warn_bad_lines is True, a warning for each
    "bad line" will be output.
delim_whitespace : bool, default False
    Specifies whether or not whitespace (e.g. ``' '`` or ``'\t'``) will be
    used as the sep. Equivalent to setting ``sep='\\s+'``. If this option
    is set to True, nothing should be passed in for the ``delimiter``
    parameter.

    .. versionadded:: 0.18.1 support for the Python parser.

low_memory : bool, default True
    Internally process the file in chunks, resulting in lower memory use
    while parsing, but possibly mixed type inference.  To ensure no mixed
    types either set False, or specify the type with the `dtype` parameter.
    Note that the entire file is read into a single DataFrame regardless,
    use the `chunksize` or `iterator` parameter to return the data in chunks.
    (Only valid with C parser).
memory_map : bool, default False
    If a filepath is provided for `filepath_or_buffer`, map the file object
    directly onto memory and access the data directly from there. Using this
    option can improve performance because there is no longer any I/O overhead.
float_precision : str, optional
    Specifies which converter the C engine should use for floating-point
    values. The options are `None` for the ordinary converter,
    `high` for the high-precision converter, and `round_trip` for the
    round-trip converter.

Returns
-------
DataFrame or TextParser
    A comma-separated values (csv) file is returned as two-dimensional
    data structure with labeled axes.

See Also
--------
to_csv : Write DataFrame to a comma-separated values (csv) file.
read_csv : Read a comma-separated values (csv) file into DataFrame.
read_fwf : Read a table of fixed-width formatted lines into DataFrame.

Examples
--------
>>> pd.{func_name}('data.csv')  # doctest: +SKIP
"""


def _validate_integer(name, val, min_val=0):
    """
    Checks whether the 'name' parameter for parsing is either
    an integer OR float that can SAFELY be cast to an integer
    without losing accuracy. Raises a ValueError if that is
    not the case.

    Parameters
    ----------
    name : string
        Parameter name (used for error reporting)
    val : int or float
        The value to check
    min_val : int
        Minimum allowed value (val < min_val will result in a ValueError)
    """
    msg = "'{name:s}' must be an integer >={min_val:d}".format(name=name,
                                                               min_val=min_val)

    if val is not None:
        if is_float(val):
            if int(val) != val:
                raise ValueError(msg)
            val = int(val)
        elif not (is_integer(val) and val >= min_val):
            raise ValueError(msg)

    return val


def _validate_names(names):
    """
    Check if the `names` parameter contains duplicates.

    If duplicates are found, we issue a warning before returning.

    Parameters
    ----------
    names : array-like or None
        An array containing a list of the names used for the output DataFrame.

    Returns
    -------
    names : array-like or None
        The original `names` parameter.
    """

    if names is not None:
        if len(names) != len(set(names)):
            msg = ("Duplicate names specified. This "
                   "will raise an error in the future.")
            warnings.warn(msg, UserWarning, stacklevel=3)

    return names


def _read(filepath_or_buffer, kwds):
    """Generic reader of line files."""
    encoding = kwds.get('encoding', None)
    if encoding is not None:
        encoding = re.sub('_', '-', encoding).lower()
        kwds['encoding'] = encoding

    compression = kwds.get('compression', 'infer')
    compression = _infer_compression(filepath_or_buffer, compression)
    filepath_or_buffer, _, compression, should_close = get_filepath_or_buffer(
        filepath_or_buffer, encoding, compression)
    kwds['compression'] = compression

    if kwds.get('date_parser', None) is not None:
        if isinstance(kwds['parse_dates'], bool):
            kwds['parse_dates'] = True

    # Extract some of the arguments (pass chunksize on).
    iterator = kwds.get('iterator', False)
    chunksize = _validate_integer('chunksize', kwds.get('chunksize', None), 1)
    nrows = kwds.get('nrows', None)

    # Check for duplicates in names.
    _validate_names(kwds.get("names", None))

    # Create the parser.
    parser = TextFileReader(filepath_or_buffer, **kwds)

    if chunksize or iterator:
        return parser

    try:
        data = parser.read(nrows)
    finally:
        parser.close()

    if should_close:
        try:
            filepath_or_buffer.close()
        except ValueError:
            pass

    return data


_parser_defaults = {
    'delimiter': None,

    'escapechar': None,
    'quotechar': '"',
    'quoting': csv.QUOTE_MINIMAL,
    'doublequote': True,
    'skipinitialspace': False,
    'lineterminator': None,

    'header': 'infer',
    'index_col': None,
    'names': None,
    'prefix': None,
    'skiprows': None,
    'skipfooter': 0,
    'nrows': None,
    'na_values': None,
    'keep_default_na': True,

    'true_values': None,
    'false_values': None,
    'converters': None,
    'dtype': None,

    'thousands': None,
    'comment': None,
    'decimal': b'.',

    # 'engine': 'c',
    'parse_dates': False,
    'keep_date_col': False,
    'dayfirst': False,
    'date_parser': None,
    'usecols': None,

    # 'iterator': False,
    'chunksize': None,
    'verbose': False,
    'encoding': None,
    'squeeze': False,
    'compression': None,
    'mangle_dupe_cols': True,
    'tupleize_cols': False,
    'infer_datetime_format': False,
    'skip_blank_lines': True
}


_c_parser_defaults = {
    'delim_whitespace': False,
    'na_filter': True,
    'low_memory': True,
    'memory_map': False,
    'error_bad_lines': True,
    'warn_bad_lines': True,
    'tupleize_cols': False,
    'float_precision': None
}

_fwf_defaults = {
    'colspecs': 'infer',
    'infer_nrows': 100,
    'widths': None,
}

_c_unsupported = {'skipfooter'}
_python_unsupported = {
    'low_memory',
    'float_precision',
}

_deprecated_defaults = {
    'tupleize_cols': None
}
_deprecated_args = {
    'tupleize_cols',
}


def _make_parser_function(name, default_sep=','):

    # prepare read_table deprecation
    if name == "read_table":
        sep = False
    else:
        sep = default_sep

    def parser_f(filepath_or_buffer,
                 sep=sep,
                 delimiter=None,

                 # Column and Index Locations and Names
                 header='infer',
                 names=None,
                 index_col=None,
                 usecols=None,
                 squeeze=False,
                 prefix=None,
                 mangle_dupe_cols=True,

                 # General Parsing Configuration
                 dtype=None,
                 engine=None,
                 converters=None,
                 true_values=None,
                 false_values=None,
                 skipinitialspace=False,
                 skiprows=None,
                 skipfooter=0,
                 nrows=None,

                 # NA and Missing Data Handling
                 na_values=None,
                 keep_default_na=True,
                 na_filter=True,
                 verbose=False,
                 skip_blank_lines=True,

                 # Datetime Handling
                 parse_dates=False,
                 infer_datetime_format=False,
                 keep_date_col=False,
                 date_parser=None,
                 dayfirst=False,

                 # Iteration
                 iterator=False,
                 chunksize=None,

                 # Quoting, Compression, and File Format
                 compression='infer',
                 thousands=None,
                 decimal=b'.',
                 lineterminator=None,
                 quotechar='"',
                 quoting=csv.QUOTE_MINIMAL,
                 doublequote=True,
                 escapechar=None,
                 comment=None,
                 encoding=None,
                 dialect=None,
                 tupleize_cols=None,

                 # Error Handling
                 error_bad_lines=True,
                 warn_bad_lines=True,

                 # Internal
                 delim_whitespace=False,
                 low_memory=_c_parser_defaults['low_memory'],
                 memory_map=False,
                 float_precision=None):

        # deprecate read_table GH21948
        if name == "read_table":
            if sep is False and delimiter is None:
                warnings.warn("read_table is deprecated, use read_csv "
                              "instead, passing sep='\\t'.",
                              FutureWarning, stacklevel=2)
            else:
                warnings.warn("read_table is deprecated, use read_csv "
                              "instead.",
                              FutureWarning, stacklevel=2)
            if sep is False:
                sep = default_sep

        # gh-23761
        #
        # When a dialect is passed, it overrides any of the overlapping
        # parameters passed in directly. We don't want to warn if the
        # default parameters were passed in (since it probably means
        # that the user didn't pass them in explicitly in the first place).
        #
        # "delimiter" is the annoying corner case because we alias it to
        # "sep" before doing comparison to the dialect values later on.
        # Thus, we need a flag to indicate that we need to "override"
        # the comparison to dialect values by checking if default values
        # for BOTH "delimiter" and "sep" were provided.
        if dialect is not None:
            sep_override = delimiter is None and sep == default_sep
            kwds = dict(sep_override=sep_override)
        else:
            kwds = dict()

        # Alias sep -> delimiter.
        if delimiter is None:
            delimiter = sep

        if delim_whitespace and delimiter != default_sep:
            raise ValueError("Specified a delimiter with both sep and"
                             " delim_whitespace=True; you can only"
                             " specify one.")

        if engine is not None:
            engine_specified = True
        else:
            engine = 'c'
            engine_specified = False

        kwds.update(delimiter=delimiter,
                    engine=engine,
                    dialect=dialect,
                    compression=compression,
                    engine_specified=engine_specified,

                    doublequote=doublequote,
                    escapechar=escapechar,
                    quotechar=quotechar,
                    quoting=quoting,
                    skipinitialspace=skipinitialspace,
                    lineterminator=lineterminator,

                    header=header,
                    index_col=index_col,
                    names=names,
                    prefix=prefix,
                    skiprows=skiprows,
                    skipfooter=skipfooter,
                    na_values=na_values,
                    true_values=true_values,
                    false_values=false_values,
                    keep_default_na=keep_default_na,
                    thousands=thousands,
                    comment=comment,
                    decimal=decimal,

                    parse_dates=parse_dates,
                    keep_date_col=keep_date_col,
                    dayfirst=dayfirst,
                    date_parser=date_parser,

                    nrows=nrows,
                    iterator=iterator,
                    chunksize=chunksize,
                    converters=converters,
                    dtype=dtype,
                    usecols=usecols,
                    verbose=verbose,
                    encoding=encoding,
                    squeeze=squeeze,
                    memory_map=memory_map,
                    float_precision=float_precision,

                    na_filter=na_filter,
                    delim_whitespace=delim_whitespace,
                    warn_bad_lines=warn_bad_lines,
                    error_bad_lines=error_bad_lines,
                    low_memory=low_memory,
                    mangle_dupe_cols=mangle_dupe_cols,
                    tupleize_cols=tupleize_cols,
                    infer_datetime_format=infer_datetime_format,
                    skip_blank_lines=skip_blank_lines)

        return _read(filepath_or_buffer, kwds)

    parser_f.__name__ = name

    return parser_f


read_csv = _make_parser_function('read_csv', default_sep=',')
read_csv = Appender(_doc_read_csv_and_table.format(
                    func_name='read_csv',
                    summary=('Read a comma-separated values (csv) file '
                             'into DataFrame.'),
                    _default_sep="','")
                    )(read_csv)

read_table = _make_parser_function('read_table', default_sep='\t')
read_table = Appender(_doc_read_csv_and_table.format(
                      func_name='read_table',
                      summary="""Read general delimited file into DataFrame.

.. deprecated:: 0.24.0
Use :func:`pandas.read_csv` instead, passing ``sep='\\t'`` if necessary.""",
                      _default_sep=r"'\\t' (tab-stop)")
                      )(read_table)


def read_fwf(filepath_or_buffer, colspecs='infer', widths=None,
             infer_nrows=100, **kwds):

    r"""
    Read a table of fixed-width formatted lines into DataFrame.

    Also supports optionally iterating or breaking of the file
    into chunks.

    Additional help can be found in the `online docs for IO Tools
    <http://pandas.pydata.org/pandas-docs/stable/io.html>`_.

    Parameters
    ----------
    filepath_or_buffer : str, path object, or file-like object
        Any valid string path is acceptable. The string could be a URL. Valid
        URL schemes include http, ftp, s3, and file. For file URLs, a host is
        expected. A local file could be: file://localhost/path/to/table.csv.

        If you want to pass in a path object, pandas accepts either
        ``pathlib.Path`` or ``py._path.local.LocalPath``.

        By file-like object, we refer to objects with a ``read()`` method,
        such as a file handler (e.g. via builtin ``open`` function)
        or ``StringIO``.
    colspecs : list of tuple (int, int) or 'infer'. optional
        A list of tuples giving the extents of the fixed-width
        fields of each line as half-open intervals (i.e.,  [from, to[ ).
        String value 'infer' can be used to instruct the parser to try
        detecting the column specifications from the first 100 rows of
        the data which are not being skipped via skiprows (default='infer').
    widths : list of int, optional
        A list of field widths which can be used instead of 'colspecs' if
        the intervals are contiguous.
    infer_nrows : int, default 100
        The number of rows to consider when letting the parser determine the
        `colspecs`.

        .. versionadded:: 0.24.0
    **kwds : optional
        Optional keyword arguments can be passed to ``TextFileReader``.

    Returns
    -------
    DataFrame or TextParser
        A comma-separated values (csv) file is returned as two-dimensional
        data structure with labeled axes.

    See Also
    --------
    to_csv : Write DataFrame to a comma-separated values (csv) file.
    read_csv : Read a comma-separated values (csv) file into DataFrame.

    Examples
    --------
    >>> pd.read_fwf('data.csv')  # doctest: +SKIP
    """

    # Check input arguments.
    if colspecs is None and widths is None:
        raise ValueError("Must specify either colspecs or widths")
    elif colspecs not in (None, 'infer') and widths is not None:
        raise ValueError("You must specify only one of 'widths' and "
                         "'colspecs'")

    # Compute 'colspecs' from 'widths', if specified.
    if widths is not None:
        colspecs, col = [], 0
        for w in widths:
            colspecs.append((col, col + w))
            col += w

    kwds['colspecs'] = colspecs
    kwds['infer_nrows'] = infer_nrows
    kwds['engine'] = 'python-fwf'
    return _read(filepath_or_buffer, kwds)


class TextFileReader(BaseIterator):
    """

    Passed dialect overrides any of the related parser options

    """

    def __init__(self, f, engine=None, **kwds):

        self.f = f

        if engine is not None:
            engine_specified = True
        else:
            engine = 'python'
            engine_specified = False

        self._engine_specified = kwds.get('engine_specified', engine_specified)

        if kwds.get('dialect') is not None:
            dialect = kwds['dialect']
            if dialect in csv.list_dialects():
                dialect = csv.get_dialect(dialect)

            # Any valid dialect should have these attributes.
            # If any are missing, we will raise automatically.
            for param in ('delimiter', 'doublequote', 'escapechar',
                          'skipinitialspace', 'quotechar', 'quoting'):
                try:
                    dialect_val = getattr(dialect, param)
                except AttributeError:
                    raise ValueError("Invalid dialect '{dialect}' provided"
                                     .format(dialect=kwds['dialect']))
                parser_default = _parser_defaults[param]
                provided = kwds.get(param, parser_default)

                # Messages for conflicting values between the dialect
                # instance and the actual parameters provided.
                conflict_msgs = []

                # Don't warn if the default parameter was passed in,
                # even if it conflicts with the dialect (gh-23761).
                if provided != parser_default and provided != dialect_val:
                    msg = ("Conflicting values for '{param}': '{val}' was "
                           "provided, but the dialect specifies '{diaval}'. "
                           "Using the dialect-specified value.".format(
                               param=param, val=provided, diaval=dialect_val))

                    # Annoying corner case for not warning about
                    # conflicts between dialect and delimiter parameter.
                    # Refer to the outer "_read_" function for more info.
                    if not (param == "delimiter" and
                            kwds.pop("sep_override", False)):
                        conflict_msgs.append(msg)

                if conflict_msgs:
                    warnings.warn('\n\n'.join(conflict_msgs), ParserWarning,
                                  stacklevel=2)
                kwds[param] = dialect_val

        if kwds.get("skipfooter"):
            if kwds.get("iterator") or kwds.get("chunksize"):
                raise ValueError("'skipfooter' not supported for 'iteration'")
            if kwds.get("nrows"):
                raise ValueError("'skipfooter' not supported with 'nrows'")

        if kwds.get('header', 'infer') == 'infer':
            kwds['header'] = 0 if kwds.get('names') is None else None

        self.orig_options = kwds

        # miscellanea
        self.engine = engine
        self._engine = None
        self._currow = 0

        options = self._get_options_with_defaults(engine)

        self.chunksize = options.pop('chunksize', None)
        self.nrows = options.pop('nrows', None)
        self.squeeze = options.pop('squeeze', False)

        # might mutate self.engine
        self.engine = self._check_file_or_buffer(f, engine)
        self.options, self.engine = self._clean_options(options, engine)

        if 'has_index_names' in kwds:
            self.options['has_index_names'] = kwds['has_index_names']

        self._make_engine(self.engine)

    def close(self):
        self._engine.close()

    def _get_options_with_defaults(self, engine):
        kwds = self.orig_options

        options = {}

        for argname, default in compat.iteritems(_parser_defaults):
            value = kwds.get(argname, default)

            # see gh-12935
            if argname == 'mangle_dupe_cols' and not value:
                raise ValueError('Setting mangle_dupe_cols=False is '
                                 'not supported yet')
            else:
                options[argname] = value

        for argname, default in compat.iteritems(_c_parser_defaults):
            if argname in kwds:
                value = kwds[argname]

                if engine != 'c' and value != default:
                    if ('python' in engine and
                            argname not in _python_unsupported):
                        pass
                    elif value == _deprecated_defaults.get(argname, default):
                        pass
                    else:
                        raise ValueError(
                            'The %r option is not supported with the'
                            ' %r engine' % (argname, engine))
            else:
                value = _deprecated_defaults.get(argname, default)
            options[argname] = value

        if engine == 'python-fwf':
            for argname, default in compat.iteritems(_fwf_defaults):
                options[argname] = kwds.get(argname, default)

        return options

    def _check_file_or_buffer(self, f, engine):
        # see gh-16530
        if is_file_like(f):
            next_attr = "__next__" if PY3 else "next"

            # The C engine doesn't need the file-like to have the "next" or
            # "__next__" attribute. However, the Python engine explicitly calls
            # "next(...)" when iterating through such an object, meaning it
            # needs to have that attribute ("next" for Python 2.x, "__next__"
            # for Python 3.x)
            if engine != "c" and not hasattr(f, next_attr):
                msg = ("The 'python' engine cannot iterate "
                       "through this file buffer.")
                raise ValueError(msg)

        return engine

    def _clean_options(self, options, engine):
        result = options.copy()

        engine_specified = self._engine_specified
        fallback_reason = None

        sep = options['delimiter']
        delim_whitespace = options['delim_whitespace']

        # C engine not supported yet
        if engine == 'c':
            if options['skipfooter'] > 0:
                fallback_reason = ("the 'c' engine does not support"
                                   " skipfooter")
                engine = 'python'

        encoding = sys.getfilesystemencoding() or 'utf-8'
        if sep is None and not delim_whitespace:
            if engine == 'c':
                fallback_reason = ("the 'c' engine does not support"
                                   " sep=None with delim_whitespace=False")
                engine = 'python'
        elif sep is not None and len(sep) > 1:
            if engine == 'c' and sep == r'\s+':
                result['delim_whitespace'] = True
                del result['delimiter']
            elif engine not in ('python', 'python-fwf'):
                # wait until regex engine integrated
                fallback_reason = ("the 'c' engine does not support"
                                   " regex separators (separators > 1 char and"
                                   r" different from '\s+' are"
                                   " interpreted as regex)")
                engine = 'python'
        elif delim_whitespace:
            if 'python' in engine:
                result['delimiter'] = r'\s+'
        elif sep is not None:
            encodeable = True
            try:
                if len(sep.encode(encoding)) > 1:
                    encodeable = False
            except UnicodeDecodeError:
                encodeable = False
            if not encodeable and engine not in ('python', 'python-fwf'):
                fallback_reason = ("the separator encoded in {encoding}"
                                   " is > 1 char long, and the 'c' engine"
                                   " does not support such separators"
                                   .format(encoding=encoding))
                engine = 'python'

        quotechar = options['quotechar']
        if (quotechar is not None and
                isinstance(quotechar, (str, compat.text_type, bytes))):
            if (len(quotechar) == 1 and ord(quotechar) > 127 and
                    engine not in ('python', 'python-fwf')):
                fallback_reason = ("ord(quotechar) > 127, meaning the "
                                   "quotechar is larger than one byte, "
                                   "and the 'c' engine does not support "
                                   "such quotechars")
                engine = 'python'

        if fallback_reason and engine_specified:
            raise ValueError(fallback_reason)

        if engine == 'c':
            for arg in _c_unsupported:
                del result[arg]

        if 'python' in engine:
            for arg in _python_unsupported:
                if fallback_reason and result[arg] != _c_parser_defaults[arg]:
                    msg = ("Falling back to the 'python' engine because"
                           " {reason}, but this causes {option!r} to be"
                           " ignored as it is not supported by the 'python'"
                           " engine.").format(reason=fallback_reason,
                                              option=arg)
                    raise ValueError(msg)
                del result[arg]

        if fallback_reason:
            warnings.warn(("Falling back to the 'python' engine because"
                           " {0}; you can avoid this warning by specifying"
                           " engine='python'.").format(fallback_reason),
                          ParserWarning, stacklevel=5)

        index_col = options['index_col']
        names = options['names']
        converters = options['converters']
        na_values = options['na_values']
        skiprows = options['skiprows']

        _validate_header_arg(options['header'])

        depr_warning = ''

        for arg in _deprecated_args:
            parser_default = _c_parser_defaults[arg]
            depr_default = _deprecated_defaults[arg]

            msg = ("The '{arg}' argument has been deprecated "
                   "and will be removed in a future version."
                   .format(arg=arg))

            if arg == 'tupleize_cols':
                msg += (' Column tuples will then '
                        'always be converted to MultiIndex.')

            if result.get(arg, depr_default) != depr_default:
                # raise Exception(result.get(arg, depr_default), depr_default)
                depr_warning += msg + '\n\n'
            else:
                result[arg] = parser_default

        if depr_warning != '':
            warnings.warn(depr_warning, FutureWarning, stacklevel=2)

        if index_col is True:
            raise ValueError("The value of index_col couldn't be 'True'")
        if _is_index_col(index_col):
            if not isinstance(index_col, (list, tuple, np.ndarray)):
                index_col = [index_col]
        result['index_col'] = index_col

        names = list(names) if names is not None else names

        # type conversion-related
        if converters is not None:
            if not isinstance(converters, dict):
                raise TypeError('Type converters must be a dict or'
                                ' subclass, input was '
                                'a {0!r}'.format(type(converters).__name__))
        else:
            converters = {}

        # Converting values to NA
        keep_default_na = options['keep_default_na']
        na_values, na_fvalues = _clean_na_values(na_values, keep_default_na)

        # handle skiprows; this is internally handled by the
        # c-engine, so only need for python parsers
        if engine != 'c':
            if is_integer(skiprows):
                skiprows = lrange(skiprows)
            if skiprows is None:
                skiprows = set()
            elif not callable(skiprows):
                skiprows = set(skiprows)

        # put stuff back
        result['names'] = names
        result['converters'] = converters
        result['na_values'] = na_values
        result['na_fvalues'] = na_fvalues
        result['skiprows'] = skiprows

        return result, engine

    def __next__(self):
        try:
            return self.get_chunk()
        except StopIteration:
            self.close()
            raise

    def _make_engine(self, engine='c'):
        if engine == 'c':
            self._engine = CParserWrapper(self.f, **self.options)
        else:
            if engine == 'python':
                klass = PythonParser
            elif engine == 'python-fwf':
                klass = FixedWidthFieldParser
            else:
                raise ValueError('Unknown engine: {engine} (valid options are'
                                 ' "c", "python", or' ' "python-fwf")'.format(
                                     engine=engine))
            self._engine = klass(self.f, **self.options)

    def _failover_to_python(self):
        raise AbstractMethodError(self)

    def read(self, nrows=None):
        nrows = _validate_integer('nrows', nrows)
        ret = self._engine.read(nrows)

        # May alter columns / col_dict
        index, columns, col_dict = self._create_index(ret)

        if index is None:
            if col_dict:
                # Any column is actually fine:
                new_rows = len(compat.next(compat.itervalues(col_dict)))
                index = RangeIndex(self._currow, self._currow + new_rows)
            else:
                new_rows = 0
        else:
            new_rows = len(index)

        df = DataFrame(col_dict, columns=columns, index=index)

        self._currow += new_rows

        if self.squeeze and len(df.columns) == 1:
            return df[df.columns[0]].copy()
        return df

    def _create_index(self, ret):
        index, columns, col_dict = ret
        return index, columns, col_dict

    def get_chunk(self, size=None):
        if size is None:
            size = self.chunksize
        if self.nrows is not None:
            if self._currow >= self.nrows:
                raise StopIteration
            size = min(size, self.nrows - self._currow)
        return self.read(nrows=size)


def _is_index_col(col):
    return col is not None and col is not False


def _is_potential_multi_index(columns):
    """
    Check whether or not the `columns` parameter
    could be converted into a MultiIndex.

    Parameters
    ----------
    columns : array-like
        Object which may or may not be convertible into a MultiIndex

    Returns
    -------
    boolean : Whether or not columns could become a MultiIndex
    """
    return (len(columns) and not isinstance(columns, MultiIndex) and
            all(isinstance(c, tuple) for c in columns))


def _evaluate_usecols(usecols, names):
    """
    Check whether or not the 'usecols' parameter
    is a callable.  If so, enumerates the 'names'
    parameter and returns a set of indices for
    each entry in 'names' that evaluates to True.
    If not a callable, returns 'usecols'.
    """
    if callable(usecols):
        return {i for i, name in enumerate(names) if usecols(name)}
    return usecols


def _validate_usecols_names(usecols, names):
    """
    Validates that all usecols are present in a given
    list of names. If not, raise a ValueError that
    shows what usecols are missing.

    Parameters
    ----------
    usecols : iterable of usecols
        The columns to validate are present in names.
    names : iterable of names
        The column names to check against.

    Returns
    -------
    usecols : iterable of usecols
        The `usecols` parameter if the validation succeeds.

    Raises
    ------
    ValueError : Columns were missing. Error message will list them.
    """
    missing = [c for c in usecols if c not in names]
    if len(missing) > 0:
        raise ValueError(
            "Usecols do not match columns, "
            "columns expected but not found: {missing}".format(missing=missing)
        )

    return usecols


def _validate_skipfooter_arg(skipfooter):
    """
    Validate the 'skipfooter' parameter.

    Checks whether 'skipfooter' is a non-negative integer.
    Raises a ValueError if that is not the case.

    Parameters
    ----------
    skipfooter : non-negative integer
        The number of rows to skip at the end of the file.

    Returns
    -------
    validated_skipfooter : non-negative integer
        The original input if the validation succeeds.

    Raises
    ------
    ValueError : 'skipfooter' was not a non-negative integer.
    """

    if not is_integer(skipfooter):
        raise ValueError("skipfooter must be an integer")

    if skipfooter < 0:
        raise ValueError("skipfooter cannot be negative")

    return skipfooter


def _validate_usecols_arg(usecols):
    """
    Validate the 'usecols' parameter.

    Checks whether or not the 'usecols' parameter contains all integers
    (column selection by index), strings (column by name) or is a callable.
    Raises a ValueError if that is not the case.

    Parameters
    ----------
    usecols : list-like, callable, or None
        List of columns to use when parsing or a callable that can be used
        to filter a list of table columns.

    Returns
    -------
    usecols_tuple : tuple
        A tuple of (verified_usecols, usecols_dtype).

        'verified_usecols' is either a set if an array-like is passed in or
        'usecols' if a callable or None is passed in.

        'usecols_dtype` is the inferred dtype of 'usecols' if an array-like
        is passed in or None if a callable or None is passed in.
    """
    msg = ("'usecols' must either be list-like of all strings, all unicode, "
           "all integers or a callable.")
    if usecols is not None:
        if callable(usecols):
            return usecols, None

        if not is_list_like(usecols):
            # see gh-20529
            #
            # Ensure it is iterable container but not string.
            raise ValueError(msg)

        usecols_dtype = lib.infer_dtype(usecols, skipna=False)

        if usecols_dtype not in ("empty", "integer",
                                 "string", "unicode"):
            raise ValueError(msg)

        usecols = set(usecols)

        if usecols_dtype == "unicode":
            # see gh-13253
            #
            # Python 2.x compatibility
            usecols = {col.encode("utf-8") for col in usecols}

        return usecols, usecols_dtype
    return usecols, None


def _validate_parse_dates_arg(parse_dates):
    """
    Check whether or not the 'parse_dates' parameter
    is a non-boolean scalar. Raises a ValueError if
    that is the case.
    """
    msg = ("Only booleans, lists, and "
           "dictionaries are accepted "
           "for the 'parse_dates' parameter")

    if parse_dates is not None:
        if is_scalar(parse_dates):
            if not lib.is_bool(parse_dates):
                raise TypeError(msg)

        elif not isinstance(parse_dates, (list, dict)):
            raise TypeError(msg)

    return parse_dates


class ParserBase(object):

    def __init__(self, kwds):
        self.names = kwds.get('names')
        self.orig_names = None
        self.prefix = kwds.pop('prefix', None)

        self.index_col = kwds.get('index_col', None)
        self.unnamed_cols = set()
        self.index_names = None
        self.col_names = None

        self.parse_dates = _validate_parse_dates_arg(
            kwds.pop('parse_dates', False))
        self.date_parser = kwds.pop('date_parser', None)
        self.dayfirst = kwds.pop('dayfirst', False)
        self.keep_date_col = kwds.pop('keep_date_col', False)

        self.na_values = kwds.get('na_values')
        self.na_fvalues = kwds.get('na_fvalues')
        self.na_filter = kwds.get('na_filter', False)
        self.keep_default_na = kwds.get('keep_default_na', True)

        self.true_values = kwds.get('true_values')
        self.false_values = kwds.get('false_values')
        self.tupleize_cols = kwds.get('tupleize_cols', False)
        self.mangle_dupe_cols = kwds.get('mangle_dupe_cols', True)
        self.infer_datetime_format = kwds.pop('infer_datetime_format', False)

        self._date_conv = _make_date_converter(
            date_parser=self.date_parser,
            dayfirst=self.dayfirst,
            infer_datetime_format=self.infer_datetime_format
        )

        # validate header options for mi
        self.header = kwds.get('header')
        if isinstance(self.header, (list, tuple, np.ndarray)):
            if not all(map(is_integer, self.header)):
                raise ValueError("header must be integer or list of integers")
            if kwds.get('usecols'):
                raise ValueError("cannot specify usecols when "
                                 "specifying a multi-index header")
            if kwds.get('names'):
                raise ValueError("cannot specify names when "
                                 "specifying a multi-index header")

            # validate index_col that only contains integers
            if self.index_col is not None:
                is_sequence = isinstance(self.index_col, (list, tuple,
                                                          np.ndarray))
                if not (is_sequence and
                        all(map(is_integer, self.index_col)) or
                        is_integer(self.index_col)):
                    raise ValueError("index_col must only contain row numbers "
                                     "when specifying a multi-index header")

        # GH 16338
        elif self.header is not None and not is_integer(self.header):
            raise ValueError("header must be integer or list of integers")

        self._name_processed = False

        self._first_chunk = True

        # GH 13932
        # keep references to file handles opened by the parser itself
        self.handles = []

    def close(self):
        for f in self.handles:
            f.close()

    @property
    def _has_complex_date_col(self):
        return (isinstance(self.parse_dates, dict) or
                (isinstance(self.parse_dates, list) and
                 len(self.parse_dates) > 0 and
                 isinstance(self.parse_dates[0], list)))

    def _should_parse_dates(self, i):
        if isinstance(self.parse_dates, bool):
            return self.parse_dates
        else:
            if self.index_names is not None:
                name = self.index_names[i]
            else:
                name = None
            j = self.index_col[i]

            if is_scalar(self.parse_dates):
                return ((j == self.parse_dates) or
                        (name is not None and name == self.parse_dates))
            else:
                return ((j in self.parse_dates) or
                        (name is not None and name in self.parse_dates))

    def _extract_multi_indexer_columns(self, header, index_names, col_names,
                                       passed_names=False):
        """ extract and return the names, index_names, col_names
            header is a list-of-lists returned from the parsers """
        if len(header) < 2:
            return header[0], index_names, col_names, passed_names

        # the names are the tuples of the header that are not the index cols
        # 0 is the name of the index, assuming index_col is a list of column
        # numbers
        ic = self.index_col
        if ic is None:
            ic = []

        if not isinstance(ic, (list, tuple, np.ndarray)):
            ic = [ic]
        sic = set(ic)

        # clean the index_names
        index_names = header.pop(-1)
        index_names, names, index_col = _clean_index_names(index_names,
                                                           self.index_col,
                                                           self.unnamed_cols)

        # extract the columns
        field_count = len(header[0])

        def extract(r):
            return tuple(r[i] for i in range(field_count) if i not in sic)

        columns = lzip(*[extract(r) for r in header])
        names = ic + columns

        # If we find unnamed columns all in a single
        # level, then our header was too long.
        for n in range(len(columns[0])):
            if all(compat.to_str(c[n]) in self.unnamed_cols for c in columns):
                raise ParserError(
                    "Passed header=[%s] are too many rows for this "
                    "multi_index of columns"
                    % ','.join(str(x) for x in self.header)
                )

        # Clean the column names (if we have an index_col).
        if len(ic):
            col_names = [r[0] if (len(r[0]) and
                                  r[0] not in self.unnamed_cols) else None
                         for r in header]
        else:
            col_names = [None] * len(header)

        passed_names = True

        return names, index_names, col_names, passed_names

    def _maybe_dedup_names(self, names):
        # see gh-7160 and gh-9424: this helps to provide
        # immediate alleviation of the duplicate names
        # issue and appears to be satisfactory to users,
        # but ultimately, not needing to butcher the names
        # would be nice!
        if self.mangle_dupe_cols:
            names = list(names)  # so we can index
            counts = defaultdict(int)
            is_potential_mi = _is_potential_multi_index(names)

            for i, col in enumerate(names):
                cur_count = counts[col]

                while cur_count > 0:
                    counts[col] = cur_count + 1

                    if is_potential_mi:
                        col = col[:-1] + ('%s.%d' % (col[-1], cur_count),)
                    else:
                        col = '%s.%d' % (col, cur_count)
                    cur_count = counts[col]

                names[i] = col
                counts[col] = cur_count + 1

        return names

    def _maybe_make_multi_index_columns(self, columns, col_names=None):
        # possibly create a column mi here
        if _is_potential_multi_index(columns):
            columns = MultiIndex.from_tuples(columns, names=col_names)
        return columns

    def _make_index(self, data, alldata, columns, indexnamerow=False):
        if not _is_index_col(self.index_col) or not self.index_col:
            index = None

        elif not self._has_complex_date_col:
            index = self._get_simple_index(alldata, columns)
            index = self._agg_index(index)
        elif self._has_complex_date_col:
            if not self._name_processed:
                (self.index_names, _,
                 self.index_col) = _clean_index_names(list(columns),
                                                      self.index_col,
                                                      self.unnamed_cols)
                self._name_processed = True
            index = self._get_complex_date_index(data, columns)
            index = self._agg_index(index, try_parse_dates=False)

        # add names for the index
        if indexnamerow:
            coffset = len(indexnamerow) - len(columns)
            index = index.set_names(indexnamerow[:coffset])

        # maybe create a mi on the columns
        columns = self._maybe_make_multi_index_columns(columns, self.col_names)

        return index, columns

    _implicit_index = False

    def _get_simple_index(self, data, columns):
        def ix(col):
            if not isinstance(col, compat.string_types):
                return col
            raise ValueError('Index %s invalid' % col)

        to_remove = []
        index = []
        for idx in self.index_col:
            i = ix(idx)
            to_remove.append(i)
            index.append(data[i])

        # remove index items from content and columns, don't pop in
        # loop
        for i in reversed(sorted(to_remove)):
            data.pop(i)
            if not self._implicit_index:
                columns.pop(i)

        return index

    def _get_complex_date_index(self, data, col_names):
        def _get_name(icol):
            if isinstance(icol, compat.string_types):
                return icol

            if col_names is None:
                raise ValueError(('Must supply column order to use %s as '
                                  'index') % str(icol))

            for i, c in enumerate(col_names):
                if i == icol:
                    return c

        to_remove = []
        index = []
        for idx in self.index_col:
            name = _get_name(idx)
            to_remove.append(name)
            index.append(data[name])

        # remove index items from content and columns, don't pop in
        # loop
        for c in reversed(sorted(to_remove)):
            data.pop(c)
            col_names.remove(c)

        return index

    def _agg_index(self, index, try_parse_dates=True):
        arrays = []

        for i, arr in enumerate(index):

            if try_parse_dates and self._should_parse_dates(i):
                arr = self._date_conv(arr)

            if self.na_filter:
                col_na_values = self.na_values
                col_na_fvalues = self.na_fvalues
            else:
                col_na_values = set()
                col_na_fvalues = set()

            if isinstance(self.na_values, dict):
                col_name = self.index_names[i]
                if col_name is not None:
                    col_na_values, col_na_fvalues = _get_na_values(
                        col_name, self.na_values, self.na_fvalues,
                        self.keep_default_na)

            arr, _ = self._infer_types(arr, col_na_values | col_na_fvalues)
            arrays.append(arr)

        names = self.index_names
        index = ensure_index_from_sequences(arrays, names)

        return index

    def _convert_to_ndarrays(self, dct, na_values, na_fvalues, verbose=False,
                             converters=None, dtypes=None):
        result = {}
        for c, values in compat.iteritems(dct):
            conv_f = None if converters is None else converters.get(c, None)
            if isinstance(dtypes, dict):
                cast_type = dtypes.get(c, None)
            else:
                # single dtype or None
                cast_type = dtypes

            if self.na_filter:
                col_na_values, col_na_fvalues = _get_na_values(
                    c, na_values, na_fvalues, self.keep_default_na)
            else:
                col_na_values, col_na_fvalues = set(), set()

            if conv_f is not None:
                # conv_f applied to data before inference
                if cast_type is not None:
                    warnings.warn(("Both a converter and dtype were specified "
                                   "for column {0} - only the converter will "
                                   "be used").format(c), ParserWarning,
                                  stacklevel=7)

                try:
                    values = lib.map_infer(values, conv_f)
                except ValueError:
                    mask = algorithms.isin(
                        values, list(na_values)).view(np.uint8)
                    values = lib.map_infer_mask(values, conv_f, mask)

                cvals, na_count = self._infer_types(
                    values, set(col_na_values) | col_na_fvalues,
                    try_num_bool=False)
            else:
                is_str_or_ea_dtype = (is_string_dtype(cast_type)
                                      or is_extension_array_dtype(cast_type))
                # skip inference if specified dtype is object
                # or casting to an EA
                try_num_bool = not (cast_type and is_str_or_ea_dtype)

                # general type inference and conversion
                cvals, na_count = self._infer_types(
                    values, set(col_na_values) | col_na_fvalues,
                    try_num_bool)

                # type specified in dtype param or cast_type is an EA
                if cast_type and (not is_dtype_equal(cvals, cast_type)
                                  or is_extension_array_dtype(cast_type)):
                    try:
                        if (is_bool_dtype(cast_type) and
                                not is_categorical_dtype(cast_type)
                                and na_count > 0):
                            raise ValueError("Bool column has NA values in "
                                             "column {column}"
                                             .format(column=c))
                    except (AttributeError, TypeError):
                        # invalid input to is_bool_dtype
                        pass
                    cvals = self._cast_types(cvals, cast_type, c)

            result[c] = cvals
            if verbose and na_count:
                print('Filled %d NA values in column %s' % (na_count, str(c)))
        return result

    def _infer_types(self, values, na_values, try_num_bool=True):
        """
        Infer types of values, possibly casting

        Parameters
        ----------
        values : ndarray
        na_values : set
        try_num_bool : bool, default try
           try to cast values to numeric (first preference) or boolean

        Returns:
        --------
        converted : ndarray
        na_count : int
        """
        na_count = 0
        if issubclass(values.dtype.type, (np.number, np.bool_)):
            mask = algorithms.isin(values, list(na_values))
            na_count = mask.sum()
            if na_count > 0:
                if is_integer_dtype(values):
                    values = values.astype(np.float64)
                np.putmask(values, mask, np.nan)
            return values, na_count

        if try_num_bool:
            try:
                result = lib.maybe_convert_numeric(values, na_values, False)
                na_count = isna(result).sum()
            except Exception:
                result = values
                if values.dtype == np.object_:
                    na_count = parsers.sanitize_objects(result,
                                                        na_values, False)
        else:
            result = values
            if values.dtype == np.object_:
                na_count = parsers.sanitize_objects(values, na_values, False)

        if result.dtype == np.object_ and try_num_bool:
            result = libops.maybe_convert_bool(np.asarray(values),
                                               true_values=self.true_values,
                                               false_values=self.false_values)

        return result, na_count

    def _cast_types(self, values, cast_type, column):
        """
        Cast values to specified type

        Parameters
        ----------
        values : ndarray
        cast_type : string or np.dtype
           dtype to cast values to
        column : string
            column name - used only for error reporting

        Returns
        -------
        converted : ndarray
        """

        if is_categorical_dtype(cast_type):
            known_cats = (isinstance(cast_type, CategoricalDtype) and
                          cast_type.categories is not None)

            if not is_object_dtype(values) and not known_cats:
                # XXX this is for consistency with
                # c-parser which parses all categories
                # as strings
                values = astype_nansafe(values, str)

            cats = Index(values).unique().dropna()
            values = Categorical._from_inferred_categories(
                cats, cats.get_indexer(values), cast_type,
                true_values=self.true_values)

        # use the EA's implementation of casting
        elif is_extension_array_dtype(cast_type):
            # ensure cast_type is an actual dtype and not a string
            cast_type = pandas_dtype(cast_type)
            array_type = cast_type.construct_array_type()
            try:
                return array_type._from_sequence_of_strings(values,
                                                            dtype=cast_type)
            except NotImplementedError:
                raise NotImplementedError(
                    "Extension Array: {ea} must implement "
                    "_from_sequence_of_strings in order "
                    "to be used in parser methods".format(ea=array_type))

        else:
            try:
                values = astype_nansafe(values, cast_type,
                                        copy=True, skipna=True)
            except ValueError:
                raise ValueError("Unable to convert column %s to "
                                 "type %s" % (column, cast_type))
        return values

    def _do_date_conversions(self, names, data):
        # returns data, columns

        if self.parse_dates is not None:
            data, names = _process_date_conversion(
                data, self._date_conv, self.parse_dates, self.index_col,
                self.index_names, names, keep_date_col=self.keep_date_col)

        return names, data


class CParserWrapper(ParserBase):
    """

    """

    def __init__(self, src, **kwds):
        self.kwds = kwds
        kwds = kwds.copy()

        ParserBase.__init__(self, kwds)

        if (kwds.get('compression') is None
           and 'utf-16' in (kwds.get('encoding') or '')):
            # if source is utf-16 plain text, convert source to utf-8
            if isinstance(src, compat.string_types):
                src = open(src, 'rb')
                self.handles.append(src)
            src = UTF8Recoder(src, kwds['encoding'])
            kwds['encoding'] = 'utf-8'

        # #2442
        kwds['allow_leading_cols'] = self.index_col is not False

        # GH20529, validate usecol arg before TextReader
        self.usecols, self.usecols_dtype = _validate_usecols_arg(
            kwds['usecols'])
        kwds['usecols'] = self.usecols

        self._reader = parsers.TextReader(src, **kwds)
        self.unnamed_cols = self._reader.unnamed_cols

        passed_names = self.names is None

        if self._reader.header is None:
            self.names = None
        else:
            if len(self._reader.header) > 1:
                # we have a multi index in the columns
                self.names, self.index_names, self.col_names, passed_names = (
                    self._extract_multi_indexer_columns(
                        self._reader.header, self.index_names, self.col_names,
                        passed_names
                    )
                )
            else:
                self.names = list(self._reader.header[0])

        if self.names is None:
            if self.prefix:
                self.names = ['%s%d' % (self.prefix, i)
                              for i in range(self._reader.table_width)]
            else:
                self.names = lrange(self._reader.table_width)

        # gh-9755
        #
        # need to set orig_names here first
        # so that proper indexing can be done
        # with _set_noconvert_columns
        #
        # once names has been filtered, we will
        # then set orig_names again to names
        self.orig_names = self.names[:]

        if self.usecols:
            usecols = _evaluate_usecols(self.usecols, self.orig_names)

            # GH 14671
            if (self.usecols_dtype == 'string' and
                    not set(usecols).issubset(self.orig_names)):
                _validate_usecols_names(usecols, self.orig_names)

            if len(self.names) > len(usecols):
                self.names = [n for i, n in enumerate(self.names)
                              if (i in usecols or n in usecols)]

            if len(self.names) < len(usecols):
                _validate_usecols_names(usecols, self.names)

        self._set_noconvert_columns()

        self.orig_names = self.names

        if not self._has_complex_date_col:
            if (self._reader.leading_cols == 0 and
                    _is_index_col(self.index_col)):

                self._name_processed = True
                (index_names, self.names,
                 self.index_col) = _clean_index_names(self.names,
                                                      self.index_col,
                                                      self.unnamed_cols)

                if self.index_names is None:
                    self.index_names = index_names

            if self._reader.header is None and not passed_names:
                self.index_names = [None] * len(self.index_names)

        self._implicit_index = self._reader.leading_cols > 0

    def close(self):
        for f in self.handles:
            f.close()

        # close additional handles opened by C parser (for compression)
        try:
            self._reader.close()
        except ValueError:
            pass

    def _set_noconvert_columns(self):
        """
        Set the columns that should not undergo dtype conversions.

        Currently, any column that is involved with date parsing will not
        undergo such conversions.
        """
        names = self.orig_names
        if self.usecols_dtype == 'integer':
            # A set of integers will be converted to a list in
            # the correct order every single time.
            usecols = list(self.usecols)
            usecols.sort()
        elif (callable(self.usecols) or
                self.usecols_dtype not in ('empty', None)):
            # The names attribute should have the correct columns
            # in the proper order for indexing with parse_dates.
            usecols = self.names[:]
        else:
            # Usecols is empty.
            usecols = None

        def _set(x):
            if usecols is not None and is_integer(x):
                x = usecols[x]

            if not is_integer(x):
                x = names.index(x)

            self._reader.set_noconvert(x)

        if isinstance(self.parse_dates, list):
            for val in self.parse_dates:
                if isinstance(val, list):
                    for k in val:
                        _set(k)
                else:
                    _set(val)

        elif isinstance(self.parse_dates, dict):
            for val in self.parse_dates.values():
                if isinstance(val, list):
                    for k in val:
                        _set(k)
                else:
                    _set(val)

        elif self.parse_dates:
            if isinstance(self.index_col, list):
                for k in self.index_col:
                    _set(k)
            elif self.index_col is not None:
                _set(self.index_col)

    def set_error_bad_lines(self, status):
        self._reader.set_error_bad_lines(int(status))

    def read(self, nrows=None):
        try:
            data = self._reader.read(nrows)
        except StopIteration:
            if self._first_chunk:
                self._first_chunk = False
                names = self._maybe_dedup_names(self.orig_names)
                index, columns, col_dict = _get_empty_meta(
                    names, self.index_col, self.index_names,
                    dtype=self.kwds.get('dtype'))
                columns = self._maybe_make_multi_index_columns(
                    columns, self.col_names)

                if self.usecols is not None:
                    columns = self._filter_usecols(columns)

                col_dict = dict(filter(lambda item: item[0] in columns,
                                       col_dict.items()))

                return index, columns, col_dict

            else:
                raise

        # Done with first read, next time raise StopIteration
        self._first_chunk = False

        names = self.names

        if self._reader.leading_cols:
            if self._has_complex_date_col:
                raise NotImplementedError('file structure not yet supported')

            # implicit index, no index names
            arrays = []

            for i in range(self._reader.leading_cols):
                if self.index_col is None:
                    values = data.pop(i)
                else:
                    values = data.pop(self.index_col[i])

                values = self._maybe_parse_dates(values, i,
                                                 try_parse_dates=True)
                arrays.append(values)

            index = ensure_index_from_sequences(arrays)

            if self.usecols is not None:
                names = self._filter_usecols(names)

            names = self._maybe_dedup_names(names)

            # rename dict keys
            data = sorted(data.items())
            data = {k: v for k, (i, v) in zip(names, data)}

            names, data = self._do_date_conversions(names, data)

        else:
            # rename dict keys
            data = sorted(data.items())

            # ugh, mutation
            names = list(self.orig_names)
            names = self._maybe_dedup_names(names)

            if self.usecols is not None:
                names = self._filter_usecols(names)

            # columns as list
            alldata = [x[1] for x in data]

            data = {k: v for k, (i, v) in zip(names, data)}

            names, data = self._do_date_conversions(names, data)
            index, names = self._make_index(data, alldata, names)

        # maybe create a mi on the columns
        names = self._maybe_make_multi_index_columns(names, self.col_names)

        return index, names, data

    def _filter_usecols(self, names):
        # hackish
        usecols = _evaluate_usecols(self.usecols, names)
        if usecols is not None and len(names) != len(usecols):
            names = [name for i, name in enumerate(names)
                     if i in usecols or name in usecols]
        return names

    def _get_index_names(self):
        names = list(self._reader.header[0])
        idx_names = None

        if self._reader.leading_cols == 0 and self.index_col is not None:
            (idx_names, names,
             self.index_col) = _clean_index_names(names, self.index_col,
                                                  self.unnamed_cols)

        return names, idx_names

    def _maybe_parse_dates(self, values, index, try_parse_dates=True):
        if try_parse_dates and self._should_parse_dates(index):
            values = self._date_conv(values)
        return values


def TextParser(*args, **kwds):
    """
    Converts lists of lists/tuples into DataFrames with proper type inference
    and optional (e.g. string to datetime) conversion. Also enables iterating
    lazily over chunks of large files

    Parameters
    ----------
    data : file-like object or list
    delimiter : separator character to use
    dialect : str or csv.Dialect instance, optional
        Ignored if delimiter is longer than 1 character
    names : sequence, default
    header : int, default 0
        Row to use to parse column labels. Defaults to the first row. Prior
        rows will be discarded
    index_col : int or list, optional
        Column or columns to use as the (possibly hierarchical) index
    has_index_names: bool, default False
        True if the cols defined in index_col have an index name and are
        not in the header.
    na_values : scalar, str, list-like, or dict, optional
        Additional strings to recognize as NA/NaN.
    keep_default_na : bool, default True
    thousands : str, optional
        Thousands separator
    comment : str, optional
        Comment out remainder of line
    parse_dates : bool, default False
    keep_date_col : bool, default False
    date_parser : function, optional
    skiprows : list of integers
        Row numbers to skip
    skipfooter : int
        Number of line at bottom of file to skip
    converters : dict, optional
        Dict of functions for converting values in certain columns. Keys can
        either be integers or column labels, values are functions that take one
        input argument, the cell (not column) content, and return the
        transformed content.
    encoding : str, optional
        Encoding to use for UTF when reading/writing (ex. 'utf-8')
    squeeze : bool, default False
        returns Series if only one column.
    infer_datetime_format: bool, default False
        If True and `parse_dates` is True for a column, try to infer the
        datetime format based on the first datetime string. If the format
        can be inferred, there often will be a large parsing speed-up.
    float_precision : str, optional
        Specifies which converter the C engine should use for floating-point
        values. The options are None for the ordinary converter,
        'high' for the high-precision converter, and 'round_trip' for the
        round-trip converter.
    """
    kwds['engine'] = 'python'
    return TextFileReader(*args, **kwds)


def count_empty_vals(vals):
    return sum(1 for v in vals if v == '' or v is None)


class PythonParser(ParserBase):

    def __init__(self, f, **kwds):
        """
        Workhorse function for processing nested list into DataFrame

        Should be replaced by np.genfromtxt eventually?
        """
        ParserBase.__init__(self, kwds)

        self.data = None
        self.buf = []
        self.pos = 0
        self.line_pos = 0

        self.encoding = kwds['encoding']
        self.compression = kwds['compression']
        self.memory_map = kwds['memory_map']
        self.skiprows = kwds['skiprows']

        if callable(self.skiprows):
            self.skipfunc = self.skiprows
        else:
            self.skipfunc = lambda x: x in self.skiprows

        self.skipfooter = _validate_skipfooter_arg(kwds['skipfooter'])
        self.delimiter = kwds['delimiter']

        self.quotechar = kwds['quotechar']
        if isinstance(self.quotechar, compat.text_type):
            self.quotechar = str(self.quotechar)

        self.escapechar = kwds['escapechar']
        self.doublequote = kwds['doublequote']
        self.skipinitialspace = kwds['skipinitialspace']
        self.lineterminator = kwds['lineterminator']
        self.quoting = kwds['quoting']
        self.usecols, _ = _validate_usecols_arg(kwds['usecols'])
        self.skip_blank_lines = kwds['skip_blank_lines']

        self.warn_bad_lines = kwds['warn_bad_lines']
        self.error_bad_lines = kwds['error_bad_lines']

        self.names_passed = kwds['names'] or None

        self.has_index_names = False
        if 'has_index_names' in kwds:
            self.has_index_names = kwds['has_index_names']

        self.verbose = kwds['verbose']
        self.converters = kwds['converters']

        self.dtype = kwds['dtype']
        self.thousands = kwds['thousands']
        self.decimal = kwds['decimal']

        self.comment = kwds['comment']
        self._comment_lines = []

        mode = 'r' if PY3 else 'rb'
        f, handles = _get_handle(f, mode, encoding=self.encoding,
                                 compression=self.compression,
                                 memory_map=self.memory_map)
        self.handles.extend(handles)

        # Set self.data to something that can read lines.
        if hasattr(f, 'readline'):
            self._make_reader(f)
        else:
            self.data = f

        # Get columns in two steps: infer from data, then
        # infer column indices from self.usecols if it is specified.
        self._col_indices = None
        (self.columns, self.num_original_columns,
         self.unnamed_cols) = self._infer_columns()

        # Now self.columns has the set of columns that we will process.
        # The original set is stored in self.original_columns.
        if len(self.columns) > 1:
            # we are processing a multi index column
            self.columns, self.index_names, self.col_names, _ = (
                self._extract_multi_indexer_columns(
                    self.columns, self.index_names, self.col_names
                )
            )
            # Update list of original names to include all indices.
            self.num_original_columns = len(self.columns)
        else:
            self.columns = self.columns[0]

        # get popped off for index
        self.orig_names = list(self.columns)

        # needs to be cleaned/refactored
        # multiple date column thing turning into a real spaghetti factory

        if not self._has_complex_date_col:
            (index_names, self.orig_names, self.columns) = (
                self._get_index_name(self.columns))
            self._name_processed = True
            if self.index_names is None:
                self.index_names = index_names

        if self.parse_dates:
            self._no_thousands_columns = self._set_no_thousands_columns()
        else:
            self._no_thousands_columns = None

        if len(self.decimal) != 1:
            raise ValueError('Only length-1 decimal markers supported')

        if self.thousands is None:
            self.nonnum = re.compile('[^-^0-9^%s]+' % self.decimal)
        else:
            self.nonnum = re.compile('[^-^0-9^%s^%s]+' % (self.thousands,
                                                          self.decimal))

    def _set_no_thousands_columns(self):
        # Create a set of column ids that are not to be stripped of thousands
        # operators.
        noconvert_columns = set()

        def _set(x):
            if is_integer(x):
                noconvert_columns.add(x)
            else:
                noconvert_columns.add(self.columns.index(x))

        if isinstance(self.parse_dates, list):
            for val in self.parse_dates:
                if isinstance(val, list):
                    for k in val:
                        _set(k)
                else:
                    _set(val)

        elif isinstance(self.parse_dates, dict):
            for val in self.parse_dates.values():
                if isinstance(val, list):
                    for k in val:
                        _set(k)
                else:
                    _set(val)

        elif self.parse_dates:
            if isinstance(self.index_col, list):
                for k in self.index_col:
                    _set(k)
            elif self.index_col is not None:
                _set(self.index_col)

        return noconvert_columns

    def _make_reader(self, f):
        sep = self.delimiter

        if sep is None or len(sep) == 1:
            if self.lineterminator:
                raise ValueError('Custom line terminators not supported in '
                                 'python parser (yet)')

            class MyDialect(csv.Dialect):
                delimiter = self.delimiter
                quotechar = self.quotechar
                escapechar = self.escapechar
                doublequote = self.doublequote
                skipinitialspace = self.skipinitialspace
                quoting = self.quoting
                lineterminator = '\n'

            dia = MyDialect

            sniff_sep = True

            if sep is not None:
                sniff_sep = False
                dia.delimiter = sep
            # attempt to sniff the delimiter
            if sniff_sep:
                line = f.readline()
                while self.skipfunc(self.pos):
                    self.pos += 1
                    line = f.readline()

                line = self._check_comments([line])[0]

                self.pos += 1
                self.line_pos += 1
                sniffed = csv.Sniffer().sniff(line)
                dia.delimiter = sniffed.delimiter
                if self.encoding is not None:
                    self.buf.extend(list(
                        UnicodeReader(StringIO(line),
                                      dialect=dia,
                                      encoding=self.encoding)))
                else:
                    self.buf.extend(list(csv.reader(StringIO(line),
                                                    dialect=dia)))

            if self.encoding is not None:
                reader = UnicodeReader(f, dialect=dia,
                                       encoding=self.encoding,
                                       strict=True)
            else:
                reader = csv.reader(f, dialect=dia,
                                    strict=True)

        else:
            def _read():
                line = f.readline()

                if compat.PY2 and self.encoding:
                    line = line.decode(self.encoding)

                pat = re.compile(sep)
                yield pat.split(line.strip())
                for line in f:
                    yield pat.split(line.strip())
            reader = _read()

        self.data = reader

    def read(self, rows=None):
        try:
            content = self._get_lines(rows)
        except StopIteration:
            if self._first_chunk:
                content = []
            else:
                raise

        # done with first read, next time raise StopIteration
        self._first_chunk = False

        columns = list(self.orig_names)
        if not len(content):  # pragma: no cover
            # DataFrame with the right metadata, even though it's length 0
            names = self._maybe_dedup_names(self.orig_names)
            index, columns, col_dict = _get_empty_meta(
                names, self.index_col, self.index_names, self.dtype)
            columns = self._maybe_make_multi_index_columns(
                columns, self.col_names)
            return index, columns, col_dict

        # handle new style for names in index
        count_empty_content_vals = count_empty_vals(content[0])
        indexnamerow = None
        if self.has_index_names and count_empty_content_vals == len(columns):
            indexnamerow = content[0]
            content = content[1:]

        alldata = self._rows_to_cols(content)
        data = self._exclude_implicit_index(alldata)

        columns = self._maybe_dedup_names(self.columns)
        columns, data = self._do_date_conversions(columns, data)

        data = self._convert_data(data)
        index, columns = self._make_index(data, alldata, columns, indexnamerow)

        return index, columns, data

    def _exclude_implicit_index(self, alldata):
        names = self._maybe_dedup_names(self.orig_names)

        if self._implicit_index:
            excl_indices = self.index_col

            data = {}
            offset = 0
            for i, col in enumerate(names):
                while i + offset in excl_indices:
                    offset += 1
                data[col] = alldata[i + offset]
        else:
            data = {k: v for k, v in zip(names, alldata)}

        return data

    # legacy
    def get_chunk(self, size=None):
        if size is None:
            size = self.chunksize
        return self.read(rows=size)

    def _convert_data(self, data):
        # apply converters
        def _clean_mapping(mapping):
            "converts col numbers to names"
            clean = {}
            for col, v in compat.iteritems(mapping):
                if isinstance(col, int) and col not in self.orig_names:
                    col = self.orig_names[col]
                clean[col] = v
            return clean

        clean_conv = _clean_mapping(self.converters)
        if not isinstance(self.dtype, dict):
            # handles single dtype applied to all columns
            clean_dtypes = self.dtype
        else:
            clean_dtypes = _clean_mapping(self.dtype)

        # Apply NA values.
        clean_na_values = {}
        clean_na_fvalues = {}

        if isinstance(self.na_values, dict):
            for col in self.na_values:
                na_value = self.na_values[col]
                na_fvalue = self.na_fvalues[col]

                if isinstance(col, int) and col not in self.orig_names:
                    col = self.orig_names[col]

                clean_na_values[col] = na_value
                clean_na_fvalues[col] = na_fvalue
        else:
            clean_na_values = self.na_values
            clean_na_fvalues = self.na_fvalues

        return self._convert_to_ndarrays(data, clean_na_values,
                                         clean_na_fvalues, self.verbose,
                                         clean_conv, clean_dtypes)

    def _infer_columns(self):
        names = self.names
        num_original_columns = 0
        clear_buffer = True
        unnamed_cols = set()

        if self.header is not None:
            header = self.header

            if isinstance(header, (list, tuple, np.ndarray)):
                have_mi_columns = len(header) > 1
                # we have a mi columns, so read an extra line
                if have_mi_columns:
                    header = list(header) + [header[-1] + 1]
            else:
                have_mi_columns = False
                header = [header]

            columns = []
            for level, hr in enumerate(header):
                try:
                    line = self._buffered_line()

                    while self.line_pos <= hr:
                        line = self._next_line()

                except StopIteration:
                    if self.line_pos < hr:
                        raise ValueError(
                            'Passed header=%s but only %d lines in file'
                            % (hr, self.line_pos + 1))

                    # We have an empty file, so check
                    # if columns are provided. That will
                    # serve as the 'line' for parsing
                    if have_mi_columns and hr > 0:
                        if clear_buffer:
                            self._clear_buffer()
                        columns.append([None] * len(columns[-1]))
                        return columns, num_original_columns, unnamed_cols

                    if not self.names:
                        raise EmptyDataError(
                            "No columns to parse from file")

                    line = self.names[:]

                this_columns = []
                this_unnamed_cols = []

                for i, c in enumerate(line):
                    if c == '':
                        if have_mi_columns:
                            col_name = ("Unnamed: {i}_level_{level}"
                                        .format(i=i, level=level))
                        else:
                            col_name = "Unnamed: {i}".format(i=i)

                        this_unnamed_cols.append(i)
                        this_columns.append(col_name)
                    else:
                        this_columns.append(c)

                if not have_mi_columns and self.mangle_dupe_cols:
                    counts = defaultdict(int)

                    for i, col in enumerate(this_columns):
                        cur_count = counts[col]

                        while cur_count > 0:
                            counts[col] = cur_count + 1
                            col = "%s.%d" % (col, cur_count)
                            cur_count = counts[col]

                        this_columns[i] = col
                        counts[col] = cur_count + 1
                elif have_mi_columns:

                    # if we have grabbed an extra line, but its not in our
                    # format so save in the buffer, and create an blank extra
                    # line for the rest of the parsing code
                    if hr == header[-1]:
                        lc = len(this_columns)
                        ic = (len(self.index_col)
                              if self.index_col is not None else 0)
                        unnamed_count = len(this_unnamed_cols)

                        if lc != unnamed_count and lc - ic > unnamed_count:
                            clear_buffer = False
                            this_columns = [None] * lc
                            self.buf = [self.buf[-1]]

                columns.append(this_columns)
                unnamed_cols.update({this_columns[i]
                                     for i in this_unnamed_cols})

                if len(columns) == 1:
                    num_original_columns = len(this_columns)

            if clear_buffer:
                self._clear_buffer()

            if names is not None:
                if ((self.usecols is not None and
                     len(names) != len(self.usecols)) or
                    (self.usecols is None and
                     len(names) != len(columns[0]))):
                    raise ValueError('Number of passed names did not match '
                                     'number of header fields in the file')
                if len(columns) > 1:
                    raise TypeError('Cannot pass names with multi-index '
                                    'columns')

                if self.usecols is not None:
                    # Set _use_cols. We don't store columns because they are
                    # overwritten.
                    self._handle_usecols(columns, names)
                else:
                    self._col_indices = None
                    num_original_columns = len(names)
                columns = [names]
            else:
                columns = self._handle_usecols(columns, columns[0])
        else:
            try:
                line = self._buffered_line()

            except StopIteration:
                if not names:
                    raise EmptyDataError(
                        "No columns to parse from file")

                line = names[:]

            ncols = len(line)
            num_original_columns = ncols

            if not names:
                if self.prefix:
                    columns = [['%s%d' % (self.prefix, i)
                                for i in range(ncols)]]
                else:
                    columns = [lrange(ncols)]
                columns = self._handle_usecols(columns, columns[0])
            else:
                if self.usecols is None or len(names) >= num_original_columns:
                    columns = self._handle_usecols([names], names)
                    num_original_columns = len(names)
                else:
                    if (not callable(self.usecols) and
                            len(names) != len(self.usecols)):
                        raise ValueError(
                            'Number of passed names did not match number of '
                            'header fields in the file'
                        )
                    # Ignore output but set used columns.
                    self._handle_usecols([names], names)
                    columns = [names]
                    num_original_columns = ncols

        return columns, num_original_columns, unnamed_cols

    def _handle_usecols(self, columns, usecols_key):
        """
        Sets self._col_indices

        usecols_key is used if there are string usecols.
        """
        if self.usecols is not None:
            if callable(self.usecols):
                col_indices = _evaluate_usecols(self.usecols, usecols_key)
            elif any(isinstance(u, string_types) for u in self.usecols):
                if len(columns) > 1:
                    raise ValueError("If using multiple headers, usecols must "
                                     "be integers.")
                col_indices = []

                for col in self.usecols:
                    if isinstance(col, string_types):
                        try:
                            col_indices.append(usecols_key.index(col))
                        except ValueError:
                            _validate_usecols_names(self.usecols, usecols_key)
                    else:
                        col_indices.append(col)
            else:
                col_indices = self.usecols

            columns = [[n for i, n in enumerate(column) if i in col_indices]
                       for column in columns]
            self._col_indices = col_indices
        return columns

    def _buffered_line(self):
        """
        Return a line from buffer, filling buffer if required.
        """
        if len(self.buf) > 0:
            return self.buf[0]
        else:
            return self._next_line()

    def _check_for_bom(self, first_row):
        """
        Checks whether the file begins with the BOM character.
        If it does, remove it. In addition, if there is quoting
        in the field subsequent to the BOM, remove it as well
        because it technically takes place at the beginning of
        the name, not the middle of it.
        """
        # first_row will be a list, so we need to check
        # that that list is not empty before proceeding.
        if not first_row:
            return first_row

        # The first element of this row is the one that could have the
        # BOM that we want to remove. Check that the first element is a
        # string before proceeding.
        if not isinstance(first_row[0], compat.string_types):
            return first_row

        # Check that the string is not empty, as that would
        # obviously not have a BOM at the start of it.
        if not first_row[0]:
            return first_row

        # Since the string is non-empty, check that it does
        # in fact begin with a BOM.
        first_elt = first_row[0][0]

        # This is to avoid warnings we get in Python 2.x if
        # we find ourselves comparing with non-Unicode
        if compat.PY2 and not isinstance(first_elt, unicode):  # noqa
            try:
                first_elt = u(first_elt)
            except UnicodeDecodeError:
                return first_row

        if first_elt != _BOM:
            return first_row

        first_row = first_row[0]

        if len(first_row) > 1 and first_row[1] == self.quotechar:
            start = 2
            quote = first_row[1]
            end = first_row[2:].index(quote) + 2

            # Extract the data between the quotation marks
            new_row = first_row[start:end]

            # Extract any remaining data after the second
            # quotation mark.
            if len(first_row) > end + 1:
                new_row += first_row[end + 1:]
            return [new_row]
        elif len(first_row) > 1:
            return [first_row[1:]]
        else:
            # First row is just the BOM, so we
            # return an empty string.
            return [""]

    def _is_line_empty(self, line):
        """
        Check if a line is empty or not.

        Parameters
        ----------
        line : str, array-like
            The line of data to check.

        Returns
        -------
        boolean : Whether or not the line is empty.
        """
        return not line or all(not x for x in line)

    def _next_line(self):
        if isinstance(self.data, list):
            while self.skipfunc(self.pos):
                self.pos += 1

            while True:
                try:
                    line = self._check_comments([self.data[self.pos]])[0]
                    self.pos += 1
                    # either uncommented or blank to begin with
                    if (not self.skip_blank_lines and
                            (self._is_line_empty(
                                self.data[self.pos - 1]) or line)):
                        break
                    elif self.skip_blank_lines:
                        ret = self._remove_empty_lines([line])
                        if ret:
                            line = ret[0]
                            break
                except IndexError:
                    raise StopIteration
        else:
            while self.skipfunc(self.pos):
                self.pos += 1
                next(self.data)

            while True:
                orig_line = self._next_iter_line(row_num=self.pos + 1)
                self.pos += 1

                if orig_line is not None:
                    line = self._check_comments([orig_line])[0]

                    if self.skip_blank_lines:
                        ret = self._remove_empty_lines([line])

                        if ret:
                            line = ret[0]
                            break
                    elif self._is_line_empty(orig_line) or line:
                        break

        # This was the first line of the file,
        # which could contain the BOM at the
        # beginning of it.
        if self.pos == 1:
            line = self._check_for_bom(line)

        self.line_pos += 1
        self.buf.append(line)
        return line

    def _alert_malformed(self, msg, row_num):
        """
        Alert a user about a malformed row.

        If `self.error_bad_lines` is True, the alert will be `ParserError`.
        If `self.warn_bad_lines` is True, the alert will be printed out.

        Parameters
        ----------
        msg : The error message to display.
        row_num : The row number where the parsing error occurred.
                  Because this row number is displayed, we 1-index,
                  even though we 0-index internally.
        """

        if self.error_bad_lines:
            raise ParserError(msg)
        elif self.warn_bad_lines:
            base = 'Skipping line {row_num}: '.format(row_num=row_num)
            sys.stderr.write(base + msg + '\n')

    def _next_iter_line(self, row_num):
        """
        Wrapper around iterating through `self.data` (CSV source).

        When a CSV error is raised, we check for specific
        error messages that allow us to customize the
        error message displayed to the user.

        Parameters
        ----------
        row_num : The row number of the line being parsed.
        """

        try:
            return next(self.data)
        except csv.Error as e:
            if self.warn_bad_lines or self.error_bad_lines:
                msg = str(e)

                if 'NULL byte' in msg:
                    msg = ('NULL byte detected. This byte '
                           'cannot be processed in Python\'s '
                           'native csv library at the moment, '
                           'so please pass in engine=\'c\' instead')

                if self.skipfooter > 0:
                    reason = ('Error could possibly be due to '
                              'parsing errors in the skipped footer rows '
                              '(the skipfooter keyword is only applied '
                              'after Python\'s csv library has parsed '
                              'all rows).')
                    msg += '. ' + reason

                self._alert_malformed(msg, row_num)
            return None

    def _check_comments(self, lines):
        if self.comment is None:
            return lines
        ret = []
        for l in lines:
            rl = []
            for x in l:
                if (not isinstance(x, compat.string_types) or
                        self.comment not in x):
                    rl.append(x)
                else:
                    x = x[:x.find(self.comment)]
                    if len(x) > 0:
                        rl.append(x)
                    break
            ret.append(rl)
        return ret

    def _remove_empty_lines(self, lines):
        """
        Iterate through the lines and remove any that are
        either empty or contain only one whitespace value

        Parameters
        ----------
        lines : array-like
            The array of lines that we are to filter.

        Returns
        -------
        filtered_lines : array-like
            The same array of lines with the "empty" ones removed.
        """

        ret = []
        for l in lines:
            # Remove empty lines and lines with only one whitespace value
            if (len(l) > 1 or len(l) == 1 and
                    (not isinstance(l[0], compat.string_types) or
                     l[0].strip())):
                ret.append(l)
        return ret

    def _check_thousands(self, lines):
        if self.thousands is None:
            return lines

        return self._search_replace_num_columns(lines=lines,
                                                search=self.thousands,
                                                replace='')

    def _search_replace_num_columns(self, lines, search, replace):
        ret = []
        for l in lines:
            rl = []
            for i, x in enumerate(l):
                if (not isinstance(x, compat.string_types) or
                    search not in x or
                    (self._no_thousands_columns and
                     i in self._no_thousands_columns) or
                        self.nonnum.search(x.strip())):
                    rl.append(x)
                else:
                    rl.append(x.replace(search, replace))
            ret.append(rl)
        return ret

    def _check_decimal(self, lines):
        if self.decimal == _parser_defaults['decimal']:
            return lines

        return self._search_replace_num_columns(lines=lines,
                                                search=self.decimal,
                                                replace='.')

    def _clear_buffer(self):
        self.buf = []

    _implicit_index = False

    def _get_index_name(self, columns):
        """
        Try several cases to get lines:

        0) There are headers on row 0 and row 1 and their
        total summed lengths equals the length of the next line.
        Treat row 0 as columns and row 1 as indices
        1) Look for implicit index: there are more columns
        on row 1 than row 0. If this is true, assume that row
        1 lists index columns and row 0 lists normal columns.
        2) Get index from the columns if it was listed.
        """
        orig_names = list(columns)
        columns = list(columns)

        try:
            line = self._next_line()
        except StopIteration:
            line = None

        try:
            next_line = self._next_line()
        except StopIteration:
            next_line = None

        # implicitly index_col=0 b/c 1 fewer column names
        implicit_first_cols = 0
        if line is not None:
            # leave it 0, #2442
            # Case 1
            if self.index_col is not False:
                implicit_first_cols = len(line) - self.num_original_columns

            # Case 0
            if next_line is not None:
                if len(next_line) == len(line) + self.num_original_columns:
                    # column and index names on diff rows
                    self.index_col = lrange(len(line))
                    self.buf = self.buf[1:]

                    for c in reversed(line):
                        columns.insert(0, c)

                    # Update list of original names to include all indices.
                    orig_names = list(columns)
                    self.num_original_columns = len(columns)
                    return line, orig_names, columns

        if implicit_first_cols > 0:
            # Case 1
            self._implicit_index = True
            if self.index_col is None:
                self.index_col = lrange(implicit_first_cols)

            index_name = None

        else:
            # Case 2
            (index_name, columns_,
             self.index_col) = _clean_index_names(columns, self.index_col,
                                                  self.unnamed_cols)

        return index_name, orig_names, columns

    def _rows_to_cols(self, content):
        col_len = self.num_original_columns

        if self._implicit_index:
            col_len += len(self.index_col)

        max_len = max(len(row) for row in content)

        # Check that there are no rows with too many
        # elements in their row (rows with too few
        # elements are padded with NaN).
        if (max_len > col_len and
                self.index_col is not False and
                self.usecols is None):

            footers = self.skipfooter if self.skipfooter else 0
            bad_lines = []

            iter_content = enumerate(content)
            content_len = len(content)
            content = []

            for (i, l) in iter_content:
                actual_len = len(l)

                if actual_len > col_len:
                    if self.error_bad_lines or self.warn_bad_lines:
                        row_num = self.pos - (content_len - i + footers)
                        bad_lines.append((row_num, actual_len))

                        if self.error_bad_lines:
                            break
                else:
                    content.append(l)

            for row_num, actual_len in bad_lines:
                msg = ('Expected %d fields in line %d, saw %d' %
                       (col_len, row_num + 1, actual_len))
                if (self.delimiter and
                        len(self.delimiter) > 1 and
                        self.quoting != csv.QUOTE_NONE):
                    # see gh-13374
                    reason = ('Error could possibly be due to quotes being '
                              'ignored when a multi-char delimiter is used.')
                    msg += '. ' + reason

                self._alert_malformed(msg, row_num + 1)

        # see gh-13320
        zipped_content = list(lib.to_object_array(
            content, min_width=col_len).T)

        if self.usecols:
            if self._implicit_index:
                zipped_content = [
                    a for i, a in enumerate(zipped_content)
                    if (i < len(self.index_col) or
                        i - len(self.index_col) in self._col_indices)]
            else:
                zipped_content = [a for i, a in enumerate(zipped_content)
                                  if i in self._col_indices]
        return zipped_content

    def _get_lines(self, rows=None):
        lines = self.buf
        new_rows = None

        # already fetched some number
        if rows is not None:
            # we already have the lines in the buffer
            if len(self.buf) >= rows:
                new_rows, self.buf = self.buf[:rows], self.buf[rows:]

            # need some lines
            else:
                rows -= len(self.buf)

        if new_rows is None:
            if isinstance(self.data, list):
                if self.pos > len(self.data):
                    raise StopIteration
                if rows is None:
                    new_rows = self.data[self.pos:]
                    new_pos = len(self.data)
                else:
                    new_rows = self.data[self.pos:self.pos + rows]
                    new_pos = self.pos + rows

                # Check for stop rows. n.b.: self.skiprows is a set.
                if self.skiprows:
                    new_rows = [row for i, row in enumerate(new_rows)
                                if not self.skipfunc(i + self.pos)]

                lines.extend(new_rows)
                self.pos = new_pos

            else:
                new_rows = []
                try:
                    if rows is not None:
                        for _ in range(rows):
                            new_rows.append(next(self.data))
                        lines.extend(new_rows)
                    else:
                        rows = 0

                        while True:
                            new_row = self._next_iter_line(
                                row_num=self.pos + rows + 1)
                            rows += 1

                            if new_row is not None:
                                new_rows.append(new_row)

                except StopIteration:
                    if self.skiprows:
                        new_rows = [row for i, row in enumerate(new_rows)
                                    if not self.skipfunc(i + self.pos)]
                    lines.extend(new_rows)
                    if len(lines) == 0:
                        raise
                self.pos += len(new_rows)

            self.buf = []
        else:
            lines = new_rows

        if self.skipfooter:
            lines = lines[:-self.skipfooter]

        lines = self._check_comments(lines)
        if self.skip_blank_lines:
            lines = self._remove_empty_lines(lines)
        lines = self._check_thousands(lines)
        return self._check_decimal(lines)


def _make_date_converter(date_parser=None, dayfirst=False,
                         infer_datetime_format=False):
    def converter(*date_cols):
        if date_parser is None:
            strs = _concat_date_cols(date_cols)

            try:
                return tools.to_datetime(
                    ensure_object(strs),
                    utc=None,
                    box=False,
                    dayfirst=dayfirst,
                    errors='ignore',
                    infer_datetime_format=infer_datetime_format
                )
            except ValueError:
                return tools.to_datetime(
                    parsing.try_parse_dates(strs, dayfirst=dayfirst))
        else:
            try:
                result = tools.to_datetime(
                    date_parser(*date_cols), errors='ignore')
                if isinstance(result, datetime.datetime):
                    raise Exception('scalar parser')
                return result
            except Exception:
                try:
                    return tools.to_datetime(
                        parsing.try_parse_dates(_concat_date_cols(date_cols),
                                                parser=date_parser,
                                                dayfirst=dayfirst),
                        errors='ignore')
                except Exception:
                    return generic_parser(date_parser, *date_cols)

    return converter


def _process_date_conversion(data_dict, converter, parse_spec,
                             index_col, index_names, columns,
                             keep_date_col=False):
    def _isindex(colspec):
        return ((isinstance(index_col, list) and
                 colspec in index_col) or
                (isinstance(index_names, list) and
                 colspec in index_names))

    new_cols = []
    new_data = {}

    orig_names = columns
    columns = list(columns)

    date_cols = set()

    if parse_spec is None or isinstance(parse_spec, bool):
        return data_dict, columns

    if isinstance(parse_spec, list):
        # list of column lists
        for colspec in parse_spec:
            if is_scalar(colspec):
                if isinstance(colspec, int) and colspec not in data_dict:
                    colspec = orig_names[colspec]
                if _isindex(colspec):
                    continue
                data_dict[colspec] = converter(data_dict[colspec])
            else:
                new_name, col, old_names = _try_convert_dates(
                    converter, colspec, data_dict, orig_names)
                if new_name in data_dict:
                    raise ValueError('New date column already in dict %s' %
                                     new_name)
                new_data[new_name] = col
                new_cols.append(new_name)
                date_cols.update(old_names)

    elif isinstance(parse_spec, dict):
        # dict of new name to column list
        for new_name, colspec in compat.iteritems(parse_spec):
            if new_name in data_dict:
                raise ValueError('Date column %s already in dict' %
                                 new_name)

            _, col, old_names = _try_convert_dates(converter, colspec,
                                                   data_dict, orig_names)

            new_data[new_name] = col
            new_cols.append(new_name)
            date_cols.update(old_names)

    data_dict.update(new_data)
    new_cols.extend(columns)

    if not keep_date_col:
        for c in list(date_cols):
            data_dict.pop(c)
            new_cols.remove(c)

    return data_dict, new_cols


def _try_convert_dates(parser, colspec, data_dict, columns):
    colset = set(columns)
    colnames = []

    for c in colspec:
        if c in colset:
            colnames.append(c)
        elif isinstance(c, int) and c not in columns:
            colnames.append(columns[c])
        else:
            colnames.append(c)

    new_name = '_'.join(str(x) for x in colnames)
    to_parse = [data_dict[c] for c in colnames if c in data_dict]

    new_col = parser(*to_parse)
    return new_name, new_col, colnames


def _clean_na_values(na_values, keep_default_na=True):

    if na_values is None:
        if keep_default_na:
            na_values = _NA_VALUES
        else:
            na_values = set()
        na_fvalues = set()
    elif isinstance(na_values, dict):
        old_na_values = na_values.copy()
        na_values = {}  # Prevent aliasing.

        # Convert the values in the na_values dictionary
        # into array-likes for further use. This is also
        # where we append the default NaN values, provided
        # that `keep_default_na=True`.
        for k, v in compat.iteritems(old_na_values):
            if not is_list_like(v):
                v = [v]

            if keep_default_na:
                v = set(v) | _NA_VALUES

            na_values[k] = v
        na_fvalues = {k: _floatify_na_values(v) for k, v in na_values.items()}
    else:
        if not is_list_like(na_values):
            na_values = [na_values]
        na_values = _stringify_na_values(na_values)
        if keep_default_na:
            na_values = na_values | _NA_VALUES

        na_fvalues = _floatify_na_values(na_values)

    return na_values, na_fvalues


def _clean_index_names(columns, index_col, unnamed_cols):
    if not _is_index_col(index_col):
        return None, columns, index_col

    columns = list(columns)

    cp_cols = list(columns)
    index_names = []

    # don't mutate
    index_col = list(index_col)

    for i, c in enumerate(index_col):
        if isinstance(c, compat.string_types):
            index_names.append(c)
            for j, name in enumerate(cp_cols):
                if name == c:
                    index_col[i] = j
                    columns.remove(name)
                    break
        else:
            name = cp_cols[c]
            columns.remove(name)
            index_names.append(name)

    # Only clean index names that were placeholders.
    for i, name in enumerate(index_names):
        if isinstance(name, compat.string_types) and name in unnamed_cols:
            index_names[i] = None

    return index_names, columns, index_col


def _get_empty_meta(columns, index_col, index_names, dtype=None):
    columns = list(columns)

    # Convert `dtype` to a defaultdict of some kind.
    # This will enable us to write `dtype[col_name]`
    # without worrying about KeyError issues later on.
    if not isinstance(dtype, dict):
        # if dtype == None, default will be np.object.
        default_dtype = dtype or np.object
        dtype = defaultdict(lambda: default_dtype)
    else:
        # Save a copy of the dictionary.
        _dtype = dtype.copy()
        dtype = defaultdict(lambda: np.object)

        # Convert column indexes to column names.
        for k, v in compat.iteritems(_dtype):
            col = columns[k] if is_integer(k) else k
            dtype[col] = v

    # Even though we have no data, the "index" of the empty DataFrame
    # could for example still be an empty MultiIndex. Thus, we need to
    # check whether we have any index columns specified, via either:
    #
    # 1) index_col (column indices)
    # 2) index_names (column names)
    #
    # Both must be non-null to ensure a successful construction. Otherwise,
    # we have to create a generic emtpy Index.
    if (index_col is None or index_col is False) or index_names is None:
        index = Index([])
    else:
        data = [Series([], dtype=dtype[name]) for name in index_names]
        index = ensure_index_from_sequences(data, names=index_names)
        index_col.sort()

        for i, n in enumerate(index_col):
            columns.pop(n - i)

    col_dict = {col_name: Series([], dtype=dtype[col_name])
                for col_name in columns}

    return index, columns, col_dict


def _floatify_na_values(na_values):
    # create float versions of the na_values
    result = set()
    for v in na_values:
        try:
            v = float(v)
            if not np.isnan(v):
                result.add(v)
        except (TypeError, ValueError, OverflowError):
            pass
    return result


def _stringify_na_values(na_values):
    """ return a stringified and numeric for these values """
    result = []
    for x in na_values:
        result.append(str(x))
        result.append(x)
        try:
            v = float(x)

            # we are like 999 here
            if v == int(v):
                v = int(v)
                result.append("%s.0" % v)
                result.append(str(v))

            result.append(v)
        except (TypeError, ValueError, OverflowError):
            pass
        try:
            result.append(int(x))
        except (TypeError, ValueError, OverflowError):
            pass
    return set(result)


def _get_na_values(col, na_values, na_fvalues, keep_default_na):
    """
    Get the NaN values for a given column.

    Parameters
    ----------
    col : str
        The name of the column.
    na_values : array-like, dict
        The object listing the NaN values as strings.
    na_fvalues : array-like, dict
        The object listing the NaN values as floats.
    keep_default_na : bool
        If `na_values` is a dict, and the column is not mapped in the
        dictionary, whether to return the default NaN values or the empty set.

    Returns
    -------
    nan_tuple : A length-two tuple composed of

        1) na_values : the string NaN values for that column.
        2) na_fvalues : the float NaN values for that column.
    """

    if isinstance(na_values, dict):
        if col in na_values:
            return na_values[col], na_fvalues[col]
        else:
            if keep_default_na:
                return _NA_VALUES, set()

            return set(), set()
    else:
        return na_values, na_fvalues


def _get_col_names(colspec, columns):
    colset = set(columns)
    colnames = []
    for c in colspec:
        if c in colset:
            colnames.append(c)
        elif isinstance(c, int):
            colnames.append(columns[c])
    return colnames


def _concat_date_cols(date_cols):
    if len(date_cols) == 1:
        if compat.PY3:
            return np.array([compat.text_type(x) for x in date_cols[0]],
                            dtype=object)
        else:
            return np.array([
                str(x) if not isinstance(x, compat.string_types) else x
                for x in date_cols[0]
            ], dtype=object)

    rs = np.array([' '.join(compat.text_type(y) for y in x)
                   for x in zip(*date_cols)], dtype=object)
    return rs


class FixedWidthReader(BaseIterator):
    """
    A reader of fixed-width lines.
    """

    def __init__(self, f, colspecs, delimiter, comment, skiprows=None,
                 infer_nrows=100):
        self.f = f
        self.buffer = None
        self.delimiter = '\r\n' + delimiter if delimiter else '\n\r\t '
        self.comment = comment
        if colspecs == 'infer':
            self.colspecs = self.detect_colspecs(infer_nrows=infer_nrows,
                                                 skiprows=skiprows)
        else:
            self.colspecs = colspecs

        if not isinstance(self.colspecs, (tuple, list)):
            raise TypeError("column specifications must be a list or tuple, "
                            "input was a %r" % type(colspecs).__name__)

        for colspec in self.colspecs:
            if not (isinstance(colspec, (tuple, list)) and
                    len(colspec) == 2 and
                    isinstance(colspec[0], (int, np.integer, type(None))) and
                    isinstance(colspec[1], (int, np.integer, type(None)))):
                raise TypeError('Each column specification must be '
                                '2 element tuple or list of integers')

    def get_rows(self, infer_nrows, skiprows=None):
        """
        Read rows from self.f, skipping as specified.

        We distinguish buffer_rows (the first <= infer_nrows
        lines) from the rows returned to detect_colspecs
        because it's simpler to leave the other locations
        with skiprows logic alone than to modify them to
        deal with the fact we skipped some rows here as
        well.

        Parameters
        ----------
        infer_nrows : int
            Number of rows to read from self.f, not counting
            rows that are skipped.
        skiprows: set, optional
            Indices of rows to skip.

        Returns
        -------
        detect_rows : list of str
            A list containing the rows to read.

        """
        if skiprows is None:
            skiprows = set()
        buffer_rows = []
        detect_rows = []
        for i, row in enumerate(self.f):
            if i not in skiprows:
                detect_rows.append(row)
            buffer_rows.append(row)
            if len(detect_rows) >= infer_nrows:
                break
        self.buffer = iter(buffer_rows)
        return detect_rows

    def detect_colspecs(self, infer_nrows=100, skiprows=None):
        # Regex escape the delimiters
        delimiters = ''.join(r'\%s' % x for x in self.delimiter)
        pattern = re.compile('([^%s]+)' % delimiters)
        rows = self.get_rows(infer_nrows, skiprows)
        if not rows:
            raise EmptyDataError("No rows from which to infer column width")
        max_len = max(map(len, rows))
        mask = np.zeros(max_len + 1, dtype=int)
        if self.comment is not None:
            rows = [row.partition(self.comment)[0] for row in rows]
        for row in rows:
            for m in pattern.finditer(row):
                mask[m.start():m.end()] = 1
        shifted = np.roll(mask, 1)
        shifted[0] = 0
        edges = np.where((mask ^ shifted) == 1)[0]
        edge_pairs = list(zip(edges[::2], edges[1::2]))
        return edge_pairs

    def __next__(self):
        if self.buffer is not None:
            try:
                line = next(self.buffer)
            except StopIteration:
                self.buffer = None
                line = next(self.f)
        else:
            line = next(self.f)
        # Note: 'colspecs' is a sequence of half-open intervals.
        return [line[fromm:to].strip(self.delimiter)
                for (fromm, to) in self.colspecs]


class FixedWidthFieldParser(PythonParser):
    """
    Specialization that Converts fixed-width fields into DataFrames.
    See PythonParser for details.
    """

    def __init__(self, f, **kwds):
        # Support iterators, convert to a list.
        self.colspecs = kwds.pop('colspecs')
        self.infer_nrows = kwds.pop('infer_nrows')
        PythonParser.__init__(self, f, **kwds)

    def _make_reader(self, f):
        self.data = FixedWidthReader(f, self.colspecs, self.delimiter,
                                     self.comment, self.skiprows,
                                     self.infer_nrows)