# ----------------------------------------------------------------------------
# Copyright (c) 2013--, scikit-bio development team.
#
# Distributed under the terms of the Modified BSD License.
#
# The full license is in the file COPYING.txt, distributed with this software.
# ----------------------------------------------------------------------------

import itertools
import numbers
import textwrap

from abc import ABCMeta, abstractmethod
from skbio._base import ElasticLines


class _MetadataReprBuilder(metaclass=ABCMeta):
    """An ABC for building a repr from an object containing metadata.

    This abstract base class constructs a repr string for an
    object which contains metadata, positional metadata and/or
    interval metadata.

    Parameters
    ----------
    obj : Type varies depending on subclass
        Object to build repr for.
    width : int
        Maximum width of the repr.
    indent : int
        Number of spaces to use for indented lines.

    """
    def __init__(self, obj, width, indent):
        self._obj = obj
        self._width = width
        self._indent = ' ' * indent

    @abstractmethod
    def _process_header(self):
        """Used by `build` Template Method to build header for the repr"""
        raise NotImplementedError

    @abstractmethod
    def _process_data(self):
        """Used by `build` Template Method to build data lines for the repr"""
        raise NotImplementedError

    def build(self):
        """Template method for building the repr"""

        self._lines = ElasticLines()

        self._process_header()
        self._process_metadata()
        self._process_positional_metadata()
        self._process_interval_metadata()
        self._process_stats()
        self._process_data()

        return self._lines.to_str()

    def _process_metadata(self):
        if self._obj.has_metadata():
            self._lines.add_line('Metadata:')
            # Python 3 doesn't allow sorting of mixed types so we can't just
            # use sorted() on the metadata keys. Sort first by type then sort
            # by value within each type.
            for key in self._sorted_keys_grouped_by_type(self._obj.metadata):
                value = self._obj.metadata[key]
                self._lines.add_lines(
                    self._format_metadata_key_value(key, value))

    def _sorted_keys_grouped_by_type(self, dict_):
        """Group keys within a dict by their type and sort within type."""
        type_sorted = sorted(dict_, key=self._type_sort_key)
        type_and_value_sorted = []
        for _, group in itertools.groupby(type_sorted, self._type_sort_key):
            type_and_value_sorted.extend(sorted(group))
        return type_and_value_sorted

    def _type_sort_key(self, key):
        return repr(type(key))

    def _format_metadata_key_value(self, key, value):
        """Format metadata key:value, wrapping across lines if necessary."""
        key_fmt = self._format_key(key)

        supported_type = True
        if isinstance(value, str):
            # extra indent of 1 so that wrapped text lines up:
            #
            #     'foo': 'abc def ghi
            #             jkl mno'
            value_repr = repr(value)
            extra_indent = 1
        elif isinstance(value, bytes):
            # extra indent of 2 so that wrapped text lines up:
            #
            #     'foo': b'abc def ghi
            #              jkl mno'
            value_repr = repr(value)
            extra_indent = 2
        # handles any number, this includes bool
        elif value is None or isinstance(value, numbers.Number):
            value_repr = repr(value)
            extra_indent = 0
        else:
            supported_type = False

        if not supported_type or len(value_repr) > 140:
            value_repr = str(type(value))
            # extra indent of 1 so that wrapped text lines up past the bracket:
            #
            #     'foo': <type
            #             'dict'>
            extra_indent = 1

        return self._wrap_text_with_indent(value_repr, key_fmt, extra_indent)

    def _process_positional_metadata(self):
        if self._obj.has_positional_metadata():
            self._lines.add_line('Positional metadata:')
            for key in self._obj.positional_metadata.columns.values.tolist():
                dtype = self._obj.positional_metadata[key].dtype
                self._lines.add_lines(
                    self._format_positional_metadata_column(key, dtype))

    def _format_positional_metadata_column(self, key, dtype):
        key_fmt = self._format_key(key)
        dtype_fmt = '<dtype: %s>' % str(dtype)
        return self._wrap_text_with_indent(dtype_fmt, key_fmt, 1)

    def _process_interval_metadata(self):
        # TODO: this hasattr check can be removed once all the relevant
        # classes have interval_metadata added to it.
        if (hasattr(self._obj, "has_interval_metadata") and
           self._obj.has_interval_metadata()):
            self._lines.add_line('Interval metadata:')
            n = self._obj.interval_metadata.num_interval_features
            line = self._indent + '%d interval feature' % n
            if n > 1:
                line += 's'
            self._lines.add_line(line)

    def _format_key(self, key):
        """Format metadata key.

        Includes initial indent and trailing colon and space:

            <indent>'foo':<space>

        """
        key_fmt = self._indent + repr(key)
        supported_types = (str, bytes, numbers.Number, type(None))
        if len(key_fmt) > (self._width / 2) or not isinstance(key,
                                                              supported_types):
            key_fmt = self._indent + str(type(key))
        return '%s: ' % key_fmt

    def _wrap_text_with_indent(self, text, initial_text, extra_indent):
        """Wrap text across lines with an initial indentation.

        For example:

            'foo': 'abc def
                    ghi jkl
                    mno pqr'

        <indent>'foo':<space> is `initial_text`. `extra_indent` is 1. Wrapped
        lines are indented such that they line up with the start of the
        previous line of wrapped text.

        """
        return textwrap.wrap(
            text, width=self._width, expand_tabs=False,
            initial_indent=initial_text,
            subsequent_indent=' ' * (len(initial_text) + extra_indent))

    def _process_stats(self):
        self._lines.add_line('Stats:')
        for label, value in self._obj._repr_stats():
            self._lines.add_line('%s%s: %s' % (self._indent, label, value))
        self._lines.add_separator()