#!/usr/bin/python

from __future__ import absolute_import, division, print_function

import sys
import os
import codecs
import collections
import io
import re

# Define an isstr() and isint() that work on both Python2 and Python3.
# See http://stackoverflow.com/questions/11301138
try:
    basestring  # attempt to evaluate basestring

    def isstr(s):
        return isinstance(s, basestring)

    def isint(i):
        return isinstance(i, (int, long))

    LONGTYPE = long
except NameError:

    def isstr(s):
        return isinstance(s, str)

    def isint(i):
        return isinstance(i, int)

    LONGTYPE = int

# Bounds to determine when an "L" suffix should be used during dump().
SMALL_INT_MIN = -2**31
SMALL_INT_MAX = 2**31 - 1

ESCAPE_SEQUENCE_RE = re.compile(r'''
    ( \\x..            # 2-digit hex escapes
    | \\[\\'"abfnrtv]  # Single-character escapes
    )''', re.UNICODE | re.VERBOSE)

SKIP_RE = re.compile(r'\s+|#.*$|//.*$|/\*(.|\n)*?\*/', re.MULTILINE)
UNPRINTABLE_CHARACTER_RE = re.compile(r'[\x00-\x1F\x7F]')


# load() logic
##############

def decode_escapes(s):
    '''Unescape libconfig string literals'''
    def decode_match(match):
        return codecs.decode(match.group(0), 'unicode-escape')

    return ESCAPE_SEQUENCE_RE.sub(decode_match, s)


class AttrDict(collections.OrderedDict):
    '''OrderedDict subclass giving access to string keys via attribute access

    This class derives from collections.OrderedDict. Thus, the original
    order of the config entries in the input stream is maintained.
    '''

    def __getattr__(self, attr):
        # Take care that getattr() raises AttributeError, not KeyError.
        # Required e.g. for hasattr(), deepcopy and OrderedDict.
        try:
            return self.__getitem__(attr)
        except KeyError:
            raise AttributeError("Attribute %r not found" % attr)


class ConfigParseError(RuntimeError):
    '''Exception class raised on errors reading the libconfig input'''
    pass


class ConfigSerializeError(TypeError):
    '''Exception class raised on errors serializing a config object'''
    pass


class Token(object):
    '''Base class for all tokens produced by the libconf tokenizer'''
    def __init__(self, type, text, filename, row, column):
        self.type = type
        self.text = text
        self.filename = filename
        self.row = row
        self.column = column

    def __str__(self):
        return "%r in %r, row %d, column %d" % (
            self.text, self.filename, self.row, self.column)


class FltToken(Token):
    '''Token subclass for floating point values'''
    def __init__(self, *args, **kwargs):
        super(FltToken, self).__init__(*args, **kwargs)
        self.value = float(self.text)


class IntToken(Token):
    '''Token subclass for integral values'''
    def __init__(self, *args, **kwargs):
        super(IntToken, self).__init__(*args, **kwargs)
        self.is_long = self.text.endswith('L')
        self.is_hex = (self.text[1:2].lower() == 'x')
        self.value = int(self.text.rstrip('L'), 0)
        if self.is_long:
            self.value = LibconfInt64(self.value)


class BoolToken(Token):
    '''Token subclass for booleans'''
    def __init__(self, *args, **kwargs):
        super(BoolToken, self).__init__(*args, **kwargs)
        self.value = (self.text[0].lower() == 't')


class StrToken(Token):
    '''Token subclass for strings'''
    def __init__(self, *args, **kwargs):
        super(StrToken, self).__init__(*args, **kwargs)
        self.value = decode_escapes(self.text[1:-1])


def compile_regexes(token_map):
    return [(cls, type, re.compile(regex))
            for cls, type, regex in token_map]


class Tokenizer:
    '''Tokenize an input string

    Typical usage:

        tokens = list(Tokenizer("<memory>").tokenize("""a = 7; b = ();"""))

    The filename argument to the constructor is used only in error messages, no
    data is loaded from the file. The input data is received as argument to the
    tokenize function, which yields tokens or throws a ConfigParseError on
    invalid input.

    Include directives are not supported, they must be handled at a higher
    level (cf. the TokenStream class).
    '''

    token_map = compile_regexes([
        (FltToken,  'float',     r'([-+]?(\d+)?\.\d*([eE][-+]?\d+)?)|'
                                 r'([-+]?(\d+)(\.\d*)?[eE][-+]?\d+)'),
        (IntToken,  'hex64',     r'0[Xx][0-9A-Fa-f]+(L(L)?)'),
        (IntToken,  'hex',       r'0[Xx][0-9A-Fa-f]+'),
        (IntToken,  'integer64', r'[-+]?[0-9]+L(L)?'),
        (IntToken,  'integer',   r'[-+]?[0-9]+'),
        (BoolToken, 'boolean',   r'(?i)(true|false)\b'),
        (StrToken,  'string',    r'"([^"\\]|\\.)*"'),
        (Token,     'name',      r'[A-Za-z\*][-A-Za-z0-9_\*]*'),
        (Token,     '}',         r'\}'),
        (Token,     '{',         r'\{'),
        (Token,     ')',         r'\)'),
        (Token,     '(',         r'\('),
        (Token,     ']',         r'\]'),
        (Token,     '[',         r'\['),
        (Token,     ',',         r','),
        (Token,     ';',         r';'),
        (Token,     '=',         r'='),
        (Token,     ':',         r':'),
    ])

    def __init__(self, filename):
        self.filename = filename
        self.row = 1
        self.column = 1

    def tokenize(self, string):
        '''Yield tokens from the input string or throw ConfigParseError'''
        pos = 0
        while pos < len(string):
            m = SKIP_RE.match(string, pos=pos)
            if m:
                skip_lines = m.group(0).split('\n')
                if len(skip_lines) > 1:
                    self.row += len(skip_lines) - 1
                    self.column = 1 + len(skip_lines[-1])
                else:
                    self.column += len(skip_lines[0])

                pos = m.end()
                continue

            for cls, type, regex in self.token_map:
                m = regex.match(string, pos=pos)
                if m:
                    yield cls(type, m.group(0),
                              self.filename, self.row, self.column)
                    self.column += len(m.group(0))
                    pos = m.end()
                    break
            else:
                raise ConfigParseError(
                    "Couldn't load config in %r row %d, column %d: %r" %
                    (self.filename, self.row, self.column,
                     string[pos:pos+20]))


class TokenStream:
    '''Offer a parsing-oriented view on tokens

    Provide several methods that are useful to parsers, like ``accept()``,
    ``expect()``, ...

    The ``from_file()`` method is the preferred way to read input files, as
    it handles include directives, which the ``Tokenizer`` class does not do.
    '''

    def __init__(self, tokens):
        self.position = 0
        self.tokens = list(tokens)

    @classmethod
    def from_file(cls, f, filename=None, includedir='', seenfiles=None):
        '''Create a token stream by reading an input file

        Read tokens from `f`. If an include directive ('@include "file.cfg"')
        is found, read its contents as well.

        The `filename` argument is used for error messages and to detect
        circular imports. ``includedir`` sets the lookup directory for included
        files.  ``seenfiles`` is used internally to detect circular includes,
        and should normally not be supplied by users of is function.
        '''

        if filename is None:
            filename = getattr(f, 'name', '<unknown>')
        if seenfiles is None:
            seenfiles = set()

        if filename in seenfiles:
            raise ConfigParseError("Circular include: %r" % (filename,))
        seenfiles = seenfiles | {filename}  # Copy seenfiles, don't alter it.

        tokenizer = Tokenizer(filename=filename)
        lines = []
        tokens = []
        for line in f:
            m = re.match(r'@include "(.*)"$', line.strip())
            if m:
                tokens.extend(tokenizer.tokenize(''.join(lines)))
                lines = [re.sub(r'\S', ' ', line)]

                includefilename = decode_escapes(m.group(1))
                includefilename = os.path.join(includedir, includefilename)
                try:
                    includefile = open(includefilename, "r")
                except IOError:
                    raise ConfigParseError("Could not open include file %r" %
                                           (includefilename,))

                with includefile:
                    includestream = cls.from_file(includefile,
                                                  filename=includefilename,
                                                  includedir=includedir,
                                                  seenfiles=seenfiles)
                tokens.extend(includestream.tokens)

            else:
                lines.append(line)

        tokens.extend(tokenizer.tokenize(''.join(lines)))
        return cls(tokens)

    def peek(self):
        '''Return (but do not consume) the next token

        At the end of input, ``None`` is returned.
        '''

        if self.position >= len(self.tokens):
            return None

        return self.tokens[self.position]

    def accept(self, *args):
        '''Consume and return the next token if it has the correct type

        Multiple token types (as strings, e.g. 'integer64') can be given
        as arguments. If the next token is one of them, consume and return it.

        If the token type doesn't match, return None.
        '''

        token = self.peek()
        if token is None:
            return None

        for arg in args:
            if token.type == arg:
                self.position += 1
                return token

        return None

    def expect(self, *args):
        '''Consume and return the next token if it has the correct type

        Multiple token types (as strings, e.g. 'integer64') can be given
        as arguments. If the next token is one of them, consume and return it.

        If the token type doesn't match, raise a ConfigParseError.
        '''

        t = self.accept(*args)
        if t is not None:
            return t

        self.error("expected: %r" % (args,))

    def error(self, msg):
        '''Raise a ConfigParseError at the current input position'''
        if self.finished():
            raise ConfigParseError("Unexpected end of input; %s" % (msg,))
        else:
            t = self.peek()
            raise ConfigParseError("Unexpected token %s; %s" % (t, msg))

    def finished(self):
        '''Return ``True`` if the end of the token stream is reached.'''
        return self.position >= len(self.tokens)


class Parser:
    '''Recursive descent parser for libconfig files

    Takes a ``TokenStream`` as input, the ``parse()`` method then returns
    the config file data in a ``json``-module-style format.
    '''

    def __init__(self, tokenstream):
        self.tokens = tokenstream

    def parse(self):
        return self.configuration()

    def configuration(self):
        result = self.setting_list_or_empty()
        if not self.tokens.finished():
            raise ConfigParseError("Expected end of input but found %s" %
                                   (self.tokens.peek(),))

        return result

    def setting_list_or_empty(self):
        result = AttrDict()
        while True:
            s = self.setting()
            if s is None:
                return result

            result[s[0]] = s[1]

    def setting(self):
        name = self.tokens.accept('name')
        if name is None:
            return None

        self.tokens.expect(':', '=')

        value = self.value()
        if value is None:
            self.tokens.error("expected a value")

        self.tokens.accept(';', ',')

        return (name.text, value)

    def value(self):
        acceptable = [self.scalar_value, self.array, self.list, self.group]
        return self._parse_any_of(acceptable)

    def scalar_value(self):
        # This list is ordered so that more common tokens are checked first.
        acceptable = [self.string, self.boolean, self.integer, self.float,
                      self.hex, self.integer64, self.hex64]
        return self._parse_any_of(acceptable)

    def value_list_or_empty(self):
        return tuple(self._comma_separated_list_or_empty(self.value))

    def scalar_value_list_or_empty(self):
        return self._comma_separated_list_or_empty(self.scalar_value)

    def array(self):
        return self._enclosed_block('[', self.scalar_value_list_or_empty, ']')

    def list(self):
        return self._enclosed_block('(', self.value_list_or_empty, ')')

    def group(self):
        return self._enclosed_block('{', self.setting_list_or_empty, '}')

    def boolean(self):
        return self._create_value_node('boolean')

    def integer(self):
        return self._create_value_node('integer')

    def integer64(self):
        return self._create_value_node('integer64')

    def hex(self):
        return self._create_value_node('hex')

    def hex64(self):
        return self._create_value_node('hex64')

    def float(self):
        return self._create_value_node('float')

    def string(self):
        t_first = self.tokens.accept('string')
        if t_first is None:
            return None

        values = [t_first.value]
        while True:
            t = self.tokens.accept('string')
            if t is None:
                break
            values.append(t.value)

        return ''.join(values)

    def _create_value_node(self, tokentype):
        t = self.tokens.accept(tokentype)
        if t is None:
            return None

        return t.value

    def _parse_any_of(self, nonterminals):
        for fun in nonterminals:
            result = fun()
            if result is not None:
                return result

        return None

    def _comma_separated_list_or_empty(self, nonterminal):
        values = []
        while True:
            v = nonterminal()
            if v is None:
                return values
            values.append(v)

            if not self.tokens.accept(','):
                return values

    def _enclosed_block(self, start, nonterminal, end):
        if not self.tokens.accept(start):
            return None
        result = nonterminal()
        self.tokens.expect(end)
        return result


def load(f, filename=None, includedir=''):
    '''Load the contents of ``f`` (a file-like object) to a Python object

    The returned object is a subclass of ``dict`` that exposes string keys as
    attributes as well.

    Example:

        >>> with open('test/example.cfg') as f:
        ...     config = libconf.load(f)
        >>> config['window']['title']
        'libconfig example'
        >>> config.window.title
        'libconfig example'
    '''

    if isinstance(f.read(0), bytes):
        raise TypeError("libconf.load() input file must by unicode")

    tokenstream = TokenStream.from_file(f,
                                        filename=filename,
                                        includedir=includedir)
    return Parser(tokenstream).parse()


def loads(string, filename=None, includedir=''):
    '''Load the contents of ``string`` to a Python object

    The returned object is a subclass of ``dict`` that exposes string keys as
    attributes as well.

    Example:

        >>> config = libconf.loads('window: { title: "libconfig example"; };')
        >>> config['window']['title']
        'libconfig example'
        >>> config.window.title
        'libconfig example'
    '''

    try:
        f = io.StringIO(string)
    except TypeError:
        raise TypeError("libconf.loads() input string must by unicode")

    return load(f, filename=filename, includedir=includedir)


# dump() logic
##############

class LibconfList(tuple):
    pass


class LibconfArray(list):
    pass


class LibconfInt64(LONGTYPE):
    pass


def is_long_int(i):
    '''Return True if argument should be dumped as int64 type

    Either because the argument is an instance of LibconfInt64, or
    because it exceeds the 32bit integer value range.
    '''

    return (isinstance(i, LibconfInt64) or
            not (SMALL_INT_MIN <= i <= SMALL_INT_MAX))


def dump_int(i):
    '''Stringize ``i``, append 'L' suffix if necessary'''
    return str(i) + ('L' if is_long_int(i) else '')


def dump_string(s):
    '''Stringize ``s``, adding double quotes and escaping as necessary

    Backslash escape backslashes, double quotes, ``\f``, ``\n``, ``\r``, and
    ``\t``. Escape all remaining unprintable characters in ``\xFF``-style.
    The returned string will be surrounded by double quotes.
    '''

    s = (s.replace('\\', '\\\\')
          .replace('"', '\\"')
          .replace('\f', r'\f')
          .replace('\n', r'\n')
          .replace('\r', r'\r')
          .replace('\t', r'\t'))
    s = UNPRINTABLE_CHARACTER_RE.sub(
            lambda m: r'\x{:02x}'.format(ord(m.group(0))),
            s)
    return '"' + s + '"'


def get_dump_type(value):
    '''Get the libconfig datatype of a value

    Return values: ``'d'`` (dict), ``'l'`` (list), ``'a'`` (array),
    ``'i'`` (integer), ``'i64'`` (long integer), ``'b'`` (bool),
    ``'f'`` (float), or ``'s'`` (string).

    Produces the proper type for LibconfList, LibconfArray, LibconfInt64
    instances.
    '''

    if isinstance(value, dict):
        return 'd'
    if isinstance(value, tuple):
        return 'l'
    if isinstance(value, list):
        return 'a'

    # Test bool before int since isinstance(True, int) == True.
    if isinstance(value, bool):
        return 'b'
    if isint(value):
        if is_long_int(value):
            return 'i64'
        else:
            return 'i'
    if isinstance(value, float):
        return 'f'
    if isstr(value):
        return 's'

    return None


def get_array_value_dtype(lst):
    '''Return array value type, raise ConfigSerializeError for invalid arrays

    Libconfig arrays must only contain scalar values and all elements must be
    of the same libconfig data type. Raises ConfigSerializeError if these
    invariants are not met.

    Returns the value type of the array. If an array contains both int and
    long int data types, the return datatype will be ``'i64'``.
    '''

    array_value_type = None
    for value in lst:
        dtype = get_dump_type(value)
        if dtype not in {'b', 'i', 'i64', 'f', 's'}:
            raise ConfigSerializeError(
                "Invalid datatype in array (may only contain scalars):"
                "%r of type %s" % (value, type(value)))

        if array_value_type is None:
            array_value_type = dtype
            continue

        if array_value_type == dtype:
            continue

        if array_value_type == 'i' and dtype == 'i64':
            array_value_type = 'i64'
            continue

        if array_value_type == 'i64' and dtype == 'i':
            continue

        raise ConfigSerializeError(
            "Mixed types in array (all elements must have same type):"
            "%r of type %s" % (value, type(value)))

    return array_value_type


def dump_value(key, value, f, indent=0):
    '''Save a value of any libconfig type

    This function serializes takes ``key`` and ``value`` and serializes them
    into ``f``. If ``key`` is ``None``, a list-style output is produced.
    Otherwise, output has ``key = value`` format.
    '''

    spaces = ' ' * indent

    if key is None:
        key_prefix = ''
        key_prefix_nl = ''
    else:
        key_prefix = key + ' = '
        key_prefix_nl = key + ' =\n' + spaces

    dtype = get_dump_type(value)
    if dtype == 'd':
        f.write(u'{}{}{{\n'.format(spaces, key_prefix_nl))
        dump_dict(value, f, indent + 4)
        f.write(u'{}}}'.format(spaces))
    elif dtype == 'l':
        f.write(u'{}{}(\n'.format(spaces, key_prefix_nl))
        dump_collection(value, f, indent + 4)
        f.write(u'\n{})'.format(spaces))
    elif dtype == 'a':
        f.write(u'{}{}[\n'.format(spaces, key_prefix_nl))
        value_dtype = get_array_value_dtype(value)

        # If int array contains one or more Int64, promote all values to i64.
        if value_dtype == 'i64':
            value = [LibconfInt64(v) for v in value]
        dump_collection(value, f, indent + 4)
        f.write(u'\n{}]'.format(spaces))
    elif dtype == 's':
        f.write(u'{}{}{}'.format(spaces, key_prefix, dump_string(value)))
    elif dtype == 'i' or dtype == 'i64':
        f.write(u'{}{}{}'.format(spaces, key_prefix, dump_int(value)))
    elif dtype == 'f' or dtype == 'b':
        f.write(u'{}{}{}'.format(spaces, key_prefix, value))
    else:
        raise ConfigSerializeError("Can not serialize object %r of type %s" %
                                   (value, type(value)))


def dump_collection(cfg, f, indent=0):
    '''Save a collection of attributes'''

    for i, value in enumerate(cfg):
        dump_value(None, value, f, indent)
        if i < len(cfg) - 1:
            f.write(u',\n')


def dump_dict(cfg, f, indent=0):
    '''Save a dictionary of attributes'''

    for key in cfg:
        if not isstr(key):
            raise ConfigSerializeError("Dict keys must be strings: %r" %
                                       (key,))
        dump_value(key, cfg[key], f, indent)
        f.write(u';\n')


def dumps(cfg):
    '''Serialize ``cfg`` into a libconfig-formatted ``str``

    ``cfg`` must be a ``dict`` with ``str`` keys and libconf-supported values
    (numbers, strings, booleans, possibly nested dicts, lists, and tuples).

    Returns the formatted string.
    '''

    str_file = io.StringIO()
    dump(cfg, str_file)
    return str_file.getvalue()


def dump(cfg, f):
    '''Serialize ``cfg`` as a libconfig-formatted stream into ``f``

    ``cfg`` must be a ``dict`` with ``str`` keys and libconf-supported values
    (numbers, strings, booleans, possibly nested dicts, lists, and tuples).

    ``f`` must be a ``file``-like object with a ``write()`` method.
    '''

    if not isinstance(cfg, dict):
        raise ConfigSerializeError(
                'dump() requires a dict as input, not %r of type %r' %
                (cfg, type(cfg)))

    dump_dict(cfg, f, 0)


# main(): small example of how to use libconf
#############################################

def main():
    '''Open the libconfig file specified by sys.argv[1] and pretty-print it'''
    global output
    if len(sys.argv[1:]) == 1:
        with io.open(sys.argv[1], 'r', encoding='utf-8') as f:
            output = load(f)
    else:
        output = load(sys.stdin)

    dump(output, sys.stdout)


if __name__ == '__main__':
    main()