path: root/gsettings/gsettings-schema-convert

#!/usr/bin/env python
# vim: set ts=4 sw=4 et: coding=UTF-8
#
# Copyright (c) 2010, Novell, Inc.
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public
# License along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
# USA.
#
# Authors: Vincent Untz <vuntz@gnome.org>

# TODO: add alias support for choices
#       choices: 'this-is-an-alias' = 'real', 'other', 'real'
# TODO: we don't support migrating a pair from a gconf schema. It has yet to be
#       seen in real-world usage, though.

import os
import sys

import optparse

try:
    from lxml import etree as ET
except ImportError:
    try:
        from xml.etree import cElementTree as ET
    except ImportError:
        import cElementTree as ET


GSETTINGS_SIMPLE_SCHEMA_INDENT = '  '
TYPES_FOR_CHOICES = [ 's' ]
TYPES_FOR_RANGE = [ 'y', 'n', 'q', 'i', 'u', 'x', 't', 'h', 'd' ]


######################################


def is_schema_id_valid(id):
    # FIXME: there's currently no restriction on what an id should contain,
    # but there might be some later on
    return True


def is_key_name_valid(name):
    # FIXME: we could check that name is valid ([-a-z0-9], no leading/trailing
    # -, no leading digit, 32 char max). Note that we don't want to validate
    # the key when converting from gconf, though, since gconf keys use
    # underscores.
    return True


def are_choices_valid(choices):
    # FIXME: we could check that all values have the same type with GVariant
    return True


def is_range_valid(minmax):
    # FIXME: we'll be able to easily check min < max once we can convert the
    # values with GVariant
    return True


######################################


class GSettingsSchemaConvertException(Exception):
    pass


######################################


class GSettingsSchemaRoot:

    def __init__(self):
        self.gettext_domain = None
        self.schemas = []

    def get_simple_string(self):
        need_empty_line = False
        result = ''

        for schema in self.schemas:
            if need_empty_line:
                result += '\n'
            result += schema.get_simple_string()
            if result:
                need_empty_line = True

        # Only put the gettext domain if we have some content
        if result and self.gettext_domain:
            result = 'gettext-domain: %s\n\n%s' % (self.gettext_domain, result)

        return result

    def get_xml_node(self):
        schemalist_node = ET.Element('schemalist')
        if self.gettext_domain:
            schemalist_node.set('gettext-domain', self.gettext_domain)
        for schema in self.schemas:
            for schema_node in schema.get_xml_nodes():
                schemalist_node.append(schema_node)
        return schemalist_node


######################################


class GSettingsSchema:

    def __init__(self):
        self.id = None
        self.path = None
        # only set when this schema is a child
        self.name = None

        self.gettext_domain = None
        self.children = []
        self.keys = []

    def get_simple_string(self, current_indent = '', parent_path = ''):
        if not self.children and not self.keys:
            return ''

        content = self._get_simple_string_for_content(current_indent)
        if not content:
            return ''

        if self.name:
            id = 'child %s' % self.name
            force_empty_line = False
        else:
            id = 'schema %s' % self.id
            force_empty_line = True

        result = ''
        result += '%s%s:\n' % (current_indent, id)
        result += self._get_simple_string_for_attributes(current_indent, parent_path, force_empty_line)
        result += content

        return result

    def _get_simple_string_for_attributes(self, current_indent, parent_path, force_empty_line):
        need_empty_line = force_empty_line
        result = ''

        if self.gettext_domain:
            result += '%sgettext-domain: %s\n' % (current_indent + GSETTINGS_SIMPLE_SCHEMA_INDENT, self.gettext_domain)
            need_empty_line = True
        if self.path and (not parent_path or (self.path != '%s%s/' % (parent_path, self.name))):
            result += '%spath: %s\n' % (current_indent + GSETTINGS_SIMPLE_SCHEMA_INDENT, self.path)
            need_empty_line = True
        if need_empty_line:
            result += '\n'

        return result

    def _get_simple_string_for_content(self, current_indent):
        need_empty_line = False
        result = ''

        for key in self.keys:
            result += key.get_simple_string(current_indent + GSETTINGS_SIMPLE_SCHEMA_INDENT)
            need_empty_line = True

        for child in self.children:
            if need_empty_line:
                result += '\n'
            result += child.get_simple_string(current_indent + GSETTINGS_SIMPLE_SCHEMA_INDENT, self.path)
            if result:
                need_empty_line = True

        return result

    def get_xml_nodes(self):
        if not self.children and not self.keys:
            return []

        (node, children_nodes) = self._get_xml_nodes_for_content()
        if node is None:
            return []

        node.set('id', self.id)
        if self.path:
            node.set('path', self.path)

        nodes = [ node ]
        nodes.extend(children_nodes)

        return nodes

    def _get_xml_nodes_for_content(self):
        if not self.keys and not self.children:
            return (None, None)

        children_nodes = []

        schema_node = ET.Element('schema')
        if self.gettext_domain:
            schema_node.set('gettext-domain', self.gettext_domain)

        for key in self.keys:
            key_node = key.get_xml_node()
            schema_node.append(key_node)
        for child in self.children:
            child_nodes = child.get_xml_nodes()
            children_nodes.extend(child_nodes)
            child_node = ET.SubElement(schema_node, 'child')
            if not child.name:
                raise GSettingsSchemaConvertException('Internal error: child being processed with no schema id.')
            child_node.set('name', child.name)
            child_node.set('schema', '%s' % child.id)

        return (schema_node, children_nodes)


######################################


class GSettingsSchemaKey:

    def __init__(self):
        self.name = None
        self.type = None
        self.default = None
        self.typed_default = None
        self.l10n = None
        self.l10n_context = None
        self.summary = None
        self.description = None
        self.choices = None
        self.range = None

    def fill(self, name, type, default, typed_default, l10n, l10n_context, summary, description, choices, range):
        self.name = name
        self.type = type
        self.default = default
        self.typed_default = typed_default
        self.l10n = l10n
        self.l10n_context = l10n_context
        self.summary = summary
        self.description = description
        self.choices = choices
        self.range = range

    def _has_range_choices(self):
        return self.choices is not None and self.type in TYPES_FOR_CHOICES

    def _has_range_minmax(self):
        return self.range is not None and len(self.range) == 2 and self.type in TYPES_FOR_RANGE

    def get_simple_string(self, current_indent):
        # FIXME: kill this when we'll have python bindings for GVariant. Right
        # now, every simple format schema we'll generate has to have an
        # explicit type since we can't guess the type later on when converting
        # to XML.
        self.typed_default = '@%s %s' % (self.type, self.default)

        result = ''
        result += '%skey %s = %s\n' % (current_indent, self.name, self.typed_default or self.default)
        current_indent += GSETTINGS_SIMPLE_SCHEMA_INDENT
        if self.l10n:
            l10n = self.l10n
            if self.l10n_context:
                l10n += ' %s' % self.l10n_context
            result += '%sl10n: %s\n' % (current_indent, l10n)
        if self.summary:
            result += '%ssummary: %s\n' % (current_indent, self.summary)
        if self.description:
            result += '%sdescription: %s\n' % (current_indent, self.description)
        if self._has_range_choices():
            result += '%schoices: %s\n' % (current_indent, ', '.join(self.choices))
        elif self._has_range_minmax():
            result += '%srange: %s\n' % (current_indent, '%s..%s' % (self.range[0] or '', self.range[1] or ''))
        return result

    def get_xml_node(self):
        key_node = ET.Element('key')
        key_node.set('name', self.name)
        key_node.set('type', self.type)
        default_node = ET.SubElement(key_node, 'default')
        default_node.text = self.default
        if self.l10n:
            default_node.set('l10n', self.l10n)
            if self.l10n_context:
                default_node.set('context', self.l10n_context)
        if self.summary:
            summary_node = ET.SubElement(key_node, 'summary')
            summary_node.text = self.summary
        if self.description:
            description_node = ET.SubElement(key_node, 'description')
            description_node.text = self.description
        if self._has_range_choices():
            choices_node = ET.SubElement(key_node, 'choices')
            for choice in self.choices:
                choice_node = ET.SubElement(choices_node, 'choice')
                choice_node.set('value', choice)
        elif self._has_range_minmax():
            (min, max) = self.range
            range_node = ET.SubElement(key_node, 'range')
            min_node = ET.SubElement(range_node, 'min')
            if min:
                min_node.text = min
            max_node = ET.SubElement(range_node, 'max')
            if max:
                max_node.text = max
        return key_node


######################################


class SimpleSchemaParser:

    allowed_tokens = {
      ''               : [ 'gettext-domain', 'schema' ],
      'gettext-domain' : [ ],
      'schema'         : [ 'gettext-domain', 'path', 'child', 'key' ],
      'path'           : [ ],
      'child'          : [ 'gettext-domain', 'child', 'key' ],
      'key'            : [ 'l10n', 'summary', 'description', 'choices', 'range' ],
      'l10n'           : [ ],
      'summary'        : [ ],
      'description'    : [ ],
      'choices'        : [ ],
      'range'          : [ ]
    }

    allowed_separators = [ ':', '=' ]

    def __init__(self, file):
        self.file = file

        self.root = GSettingsSchemaRoot()

        # this is just a convenient helper to remove the leading indentation
        # that should be common to all lines
        self.leading_indent = None

        self.indent_stack = []
        self.token_stack = []
        self.object_stack = [ self.root ]

        self.previous_token = None
        self.current_token = None
        self.unparsed_line = ''

    def _eat_indent(self):
        line = self.unparsed_line
        i = 0
        buf = ''
        previous_max_index = len(self.indent_stack) - 1
        index = -1

        while i < len(line) - 1 and line[i].isspace():
            buf += line[i]
            i += 1
            if previous_max_index > index:
                if buf == self.indent_stack[index + 1]:
                    buf = ''
                    index += 1
                    continue
                elif self.indent_stack[index + 1].startswith(buf):
                    continue
                else:
                    raise GSettingsSchemaConvertException('Inconsistent indentation.')
            else:
                continue

        if buf and previous_max_index > index:
            raise GSettingsSchemaConvertException('Inconsistent indentation.')
        elif buf and previous_max_index <= index:
            self.indent_stack.append(buf)
        elif previous_max_index > index:
            self.indent_stack = self.indent_stack[:index + 1]

        self.unparsed_line = line[i:]

    def _parse_word(self):
        line = self.unparsed_line
        i = 0
        while i < len(line) and not line[i].isspace() and not line[i] in self.allowed_separators:
            i += 1
        self.unparsed_line = line[i:]
        return line[:i]

    def _word_to_token(self, word):
        lower = word.lower()
        if lower and lower in self.allowed_tokens.keys():
            return lower
        raise GSettingsSchemaConvertException('\'%s\' is not a valid token.' % lower)

    def _token_allow_separator(self):
        return self.current_token in [ 'gettext-domain', 'path', 'l10n', 'summary', 'description', 'choices', 'range' ]

    def _parse_id_without_separator(self):
        line = self.unparsed_line
        if line[-1] in self.allowed_separators:
            line = line[:-1].strip()
        if not is_schema_id_valid(line):
            raise GSettingsSchemaConvertException('\'%s\' is not a valid schema id.' % line)

        self.unparsed_line = ''
        return line

    def _parse_key(self):
        line = self.unparsed_line

        split = False
        for separator in self.allowed_separators:
            items = line.split(separator)
            if len(items) == 2:
                split = True
                break

        if not split:
            raise GSettingsSchemaConvertException('Key \'%s\' cannot be parsed.' % line)

        name = items[0].strip()
        if not is_key_name_valid(name):
            raise GSettingsSchemaConvertException('\'%s\' is not a valid key name.' % name)

        type = ''
        value = items[1].strip()
        if value[0] == '@':
            i = 1
            while not value[i].isspace():
                i += 1
            type = value[1:i]
            value = value[i:].strip()
            if not value:
                raise GSettingsSchemaConvertException('No value specified for key \'%s\' (\'%s\').' % (name, line))

        self.unparsed_line = ''

        object = GSettingsSchemaKey()
        object.name = name
        object.type = type
        object.default = value

        return object

    def _parse_l10n(self):
        line = self.unparsed_line

        items = [ item.strip() for item in line.split(' ', 1) if item.strip() ]
        if not items:
            self.unparsed_line = ''
            return (None, None)
        if len(items) == 1:
            self.unparsed_line = ''
            return (items[0], None)
        if len(items) == 2:
            self.unparsed_line = ''
            return (items[0], items[1])

        raise GSettingsSchemaConvertException('Internal error: more items than expected for localization \'%s\'.' % line)

    def _parse_choices(self, object):
        if object.type not in TYPES_FOR_CHOICES:
            raise GSettingsSchemaConvertException('Key \'%s\' of type \'%s\' cannot have choices.' % (object.name, object.type))

        line = self.unparsed_line
        choices = [ item.strip() for item in line.split(',') ]
        if not are_choices_valid(choices):
            raise GSettingsSchemaConvertException('\'%s\' is not a valid choice.' % line)

        self.unparsed_line = ''
        return choices

    def _parse_range(self, object):
        if object.type not in TYPES_FOR_RANGE:
            raise GSettingsSchemaConvertException('Key \'%s\' of type \'%s\' cannot have a range.' % (object.name, object.type))

        line = self.unparsed_line
        minmax = [ item.strip() for item in line.split('..') ]

        if len(minmax) != 2:
            raise GSettingsSchemaConvertException('Range \'%s\' cannot be parsed.' % line)
        if not is_range_valid(minmax):
            raise GSettingsSchemaConvertException('\'%s\' is not a valid range.' % line)

        self.unparsed_line = ''
        return tuple(minmax)

    def parse_line(self, line):
        # make sure that lines with only spaces are ignored and considered as
        # empty lines
        self.unparsed_line = line.rstrip()

        # ignore empty line
        if not self.unparsed_line:
            return

        # look at the indentation to know where we should be
        self._eat_indent()
        if self.leading_indent is None:
            self.leading_indent = len(self.indent_stack)

        # ignore comments
        if self.unparsed_line[0] == '#':
            return

        word = self._parse_word()
        if self.current_token:
            self.previous_token = self.current_token
        self.current_token = self._word_to_token(word)
        self.unparsed_line = self.unparsed_line.lstrip()

        allow_separator = self._token_allow_separator()
        if len(self.unparsed_line) > 0 and self.unparsed_line[0] in self.allowed_separators:
            if allow_separator:
                self.unparsed_line = self.unparsed_line[1:].lstrip()
            else:
                raise GSettingsSchemaConvertException('Separator \'%s\' is not allowed after \'%s\'.' % (self.unparsed_line[0], self.current_token))

        new_level = len(self.indent_stack) - self.leading_indent
        old_level = len(self.token_stack)

        if new_level > old_level + 1:
            raise GSettingsSchemaConvertException('Internal error: stacks not in sync.')
        elif new_level <= old_level:
            self.token_stack = self.token_stack[:new_level]
            # we always have the root
            self.object_stack = self.object_stack[:new_level + 1]

        if new_level == 0:
            parent_token = ''
        else:
            parent_token = self.token_stack[-1]

        # there's new indentation, but no token is allowed under the previous
        # one
        if new_level == old_level + 1 and self.previous_token != parent_token:
            raise GSettingsSchemaConvertException('\'%s\' is not allowed under \'%s\'.' % (self.current_token, self.previous_token))

        if not self.current_token in self.allowed_tokens[parent_token]:
            if parent_token:
                error = '\'%s\' is not allowed under \'%s\'.' % (self.current_token, parent_token)
            else:
                error = '\'%s\' is not allowed at the root level.' % self.current_token
            raise GSettingsSchemaConvertException(error)

        current_object = self.object_stack[-1]

        new_object = None
        if self.current_token == 'gettext-domain':
            current_object.gettext_domain = self.unparsed_line
        elif self.current_token == 'schema':
            name = self._parse_id_without_separator()
            new_object = GSettingsSchema()
            new_object.id = name
            current_object.schemas.append(new_object)
        elif self.current_token == 'path':
            current_object.path = self.unparsed_line
        elif self.current_token == 'child':
            if not isinstance(current_object, GSettingsSchema):
                raise GSettingsSchemaConvertException('Internal error: child being processed with no parent schema.')
            name = self._parse_id_without_separator()
            new_object = GSettingsSchema()
            new_object.id = '%s.%s' % (current_object.id, name)
            if current_object.path:
                new_object.path = '%s%s/' % (current_object.path, name)
            new_object.name = name
            current_object.children.append(new_object)
        elif self.current_token == 'key':
            new_object =  self._parse_key()
            current_object.keys.append(new_object)
        elif self.current_token == 'l10n':
            (current_object.l10n, current_object.l10n_context) = self._parse_l10n()
        elif self.current_token == 'summary':
            current_object.summary = self.unparsed_line
        elif self.current_token == 'description':
            current_object.description = self.unparsed_line
        elif self.current_token == 'choices':
            current_object.choices = self._parse_choices(current_object)
        elif self.current_token == 'range':
            current_object.range = self._parse_range(current_object)

        if new_object:
            self.token_stack.append(self.current_token)
            self.object_stack.append(new_object)

    def parse(self):
        f = open(self.file, 'r')
        lines = [ line[:-1] for line in f.readlines() ]
        f.close()

        try:
            current_line_nb = 0
            for line in lines:
                current_line_nb += 1
                self.parse_line(line)
        except GSettingsSchemaConvertException, e:
            raise GSettingsSchemaConvertException('%s:%s: %s' % (os.path.basename(self.file), current_line_nb, e))

        return self.root


######################################


class XMLSchemaParser:

    def __init__(self, file):
        self.file = file

        self.root = None

    def _parse_key(self, key_node, schema):
        key = GSettingsSchemaKey()

        key.name = key_node.get('name')
        if not key.name:
            raise GSettingsSchemaConvertException('A key in schema \'%s\' has no name.' % schema.id)
        key.type = key_node.get('type')
        if not key.type:
            raise GSettingsSchemaConvertException('Key \'%s\' in schema \'%s\' has no type.' % (key.name, schema.id))

        default_node = key_node.find('default')
        if default_node is None or not default_node.text.strip():
            raise GSettingsSchemaConvertException('Key \'%s\' in schema \'%s\' has no default value.' % (key.name, schema.id))
        key.l10n = default_node.get('l10n')
        key.l10n_context = default_node.get('context')
        key.default = default_node.text.strip()

        summary_node = key_node.find('summary')
        if summary_node is not None:
            key.summary = summary_node.text.strip()
        description_node = key_node.find('description')
        if description_node is not None:
            key.description = description_node.text.strip()

        range_node = key_node.find('range')
        if range_node is not None:
            min = None
            max = None
            min_node = range_node.find('min')
            if min_node is not None:
                min = min_node.text.strip()
            max_node = range_node.find('max')
            if max_node is not None:
                max = max_node.text.strip()
            if min or max:
                self.range = (min, max)

        choices_node = key_node.find('choices')
        if choices_node is not None:
            self.choices = []
            for choice_node in choices_node.findall('choice'):
                value = choice_node.get('value')
                if value:
                    self.choices.append(value)
                else:
                    raise GSettingsSchemaConvertException('A choice for key \'%s\' in schema \'%s\' has no value.' % (key.name, schema.id))

        return key

    def _parse_schema(self, schema_node):
        schema = GSettingsSchema()

        schema._children = []

        schema.id = schema_node.get('id')
        if not schema.id:
            raise GSettingsSchemaConvertException('A schema has no id.')
        schema.path = schema_node.get('path')
        schema.gettext_domain = schema_node.get('gettext-domain')

        for key_node in schema_node.findall('key'):
            key = self._parse_key(key_node, schema)
            schema.keys.append(key)

        for child_node in schema_node.findall('child'):
            child_name = child_node.get('name')
            if not child_name:
                raise GSettingsSchemaConvertException('A child of schema \'%s\' has no name.' % schema.id)
            child_schema = child_node.get('schema')
            if not child_schema:
                raise GSettingsSchemaConvertException('Child \'%s\' of schema \'%s\' has no schema.' % (child_name, schema.id))

            expected_id = schema.id + '.' + child_name
            if child_schema != expected_id:
                raise GSettingsSchemaConvertException('\'%s\' is too complex for this tool: child \'%s\' of schema \'%s\' has a schema that is not the expected one (\'%s\' vs \'%s\').' % (os.path.basename(self.file), child_name, schema.id, child_schema, expected_id))

            schema._children.append((child_schema, child_name))

        return schema

    def parse(self):
        self.root = GSettingsSchemaRoot()
        schemas = []
        parent = {}

        schemalist_node = ET.parse(self.file).getroot()
        self.root.gettext_domain = schemalist_node.get('gettext-domain')

        for schema_node in schemalist_node.findall('schema'):
            schema = self._parse_schema(schema_node)

            for (child_schema, child_name) in schema._children:
                if parent.has_key(child_schema):
                    raise GSettingsSchemaConvertException('Child \'%s\' is declared by two different schemas: \'%s\' and \'%s\'.' % (child_schema, parent[child_schema], schema.id))
                parent[child_schema] = schema

            schemas.append(schema)

        # now let's move all schemas where they should leave
        for schema in schemas:
            if parent.has_key(schema.id):
                parent_schema = parent[schema.id]

                # check that the paths of parent and child are supported by
                # this tool
                found = False
                for (child_schema, child_name) in parent_schema._children:
                    if child_schema == schema.id:
                        found = True
                        break

                if not found:
                    raise GSettingsSchemaConvertException('Internal error: child not found in parent\'s children.')

                schema.name = child_name
                parent_schema.children.append(schema)
            else:
                self.root.schemas.append(schema)

        return self.root


######################################


def map_gconf_type_to_variant_type(gconftype, gconfsubtype):
    typemap = { 'string': 's', 'int': 'i', 'float': 'd', 'bool': 'b', 'list': 'a' }
    try:
        result = typemap[gconftype]
    except KeyError:
        raise GSettingsSchemaConvertException('Type \'%s\' is not a known gconf type.' % gconftype)

    if gconftype == 'list':
        try:
            result = result + typemap[gconfsubtype]
        except KeyError:
            raise GSettingsSchemaConvertException('Type \'%s\' is not a known gconf type.' % gconfsubtype)

    return result


def fix_value_for_simple_gconf_type(gconftype, gconfvalue):
    '''If there is no value, then we choose a 'neutral' value (false, 0, empty
       string).
    '''
    if gconftype == 'string':
        if not gconfvalue:
            return '\'\''
        return '\'' + gconfvalue.replace('\'', '\\\'') + '\''
    elif gconftype == 'int':
        if not gconfvalue:
            return '0'

        try:
            int(gconfvalue)
        except ValueError:
            raise GSettingsSchemaConvertException()

        return gconfvalue
    elif gconftype == 'float':
        if not gconfvalue:
            return '0.0'

        try:
            float(gconfvalue)
        except ValueError:
            raise GSettingsSchemaConvertException()

        return gconfvalue
    elif gconftype == 'bool':
        if not gconfvalue:
            return 'false'

        value = gconfvalue.lower()
        # gconf schemas can have 0/1 for false/true
        if value == '0':
            return 'false'
        elif value == '1':
            return 'true'
        elif value in ['false', 'true']:
            return value
        else:
            raise GSettingsSchemaConvertException()
    else:
        return gconfvalue


class GConfSchema:

    def __init__(self, node):
        locale_node = node.find('locale')

        self.key = node.find('key').text
        self.type = node.find('type').text
        if self.type == 'list':
            self.list_type = node.find('list_type').text
        else:
            self.list_type = None
        self.varianttype = map_gconf_type_to_variant_type(self.type, self.list_type)

        applyto_node = node.find('applyto')
        if applyto_node is not None:
            self.applyto = node.find('applyto').text
            self.applyto.strip()
            self.keyname = self.applyto[self.applyto.rfind('/')+1:]
            self.prefix = self.applyto[:self.applyto.rfind('/')+1]
        else:
            self.applyto = None
            self.key.strip()
            self.keyname = self.key[self.key.rfind('/')+1:]
            self.prefix = self.key[:self.key.rfind('/')+1]
        self.prefix = os.path.normpath(self.prefix)

        try:
            self.default = locale_node.find('default').text
            self.localized = 'messages'
        except:
            try:
                self.default = node.find('default').text
            except:
                self.default = ''
            self.localized = None
        self.typed_default = None

        self.short = self._get_value_with_locale(node, locale_node, 'short')
        self.long = self._get_value_with_locale(node, locale_node, 'long')

        if self.short:
            self.short = self._oneline(self.short)
        if self.long:
            self.long = self._oneline(self.long)

        # Fix the default value to be parsable by GVariant
        if self.type == 'list':
            l = self.default.strip()
            if not l:
                l = '[]'
            elif not (l[0] == '[' and l[-1] == ']'):
                raise GSettingsSchemaConvertException('Cannot parse default list value \'%s\' for key \'%s\'.' % (self.default, self.applyto or self.key))
            values = l[1:-1].strip()
            if not values:
                self.default = '[]'
                self.typed_default = '@%s []' % self.varianttype
            else:
                items = [ item.strip() for item in values.split(',') ]
                try:
                    items = [ fix_value_for_simple_gconf_type(self.list_type, item) for item in items ]
                except GSettingsSchemaConvertException:
                    raise GSettingsSchemaConvertException('Invalid item(s) of type \'%s\' in default list \'%s\' for key \'%s\'.' % (self.list_type, self.default, self.applyto or self.key))
                values = ', '.join(items)
                self.default = '[ %s ]' % values
        else:
            try:
                self.default = fix_value_for_simple_gconf_type(self.type, self.default)
            except GSettingsSchemaConvertException:
                raise GSettingsSchemaConvertException('Invalid default value \'%s\' of type \'%s\' for key \'%s\'.' % (self.default, self.type, self.applyto or self.key))

    def _get_value_with_locale(self, node, locale_node, element):
        element_node = None
        if locale_node is not None:
            element_node = locale_node.find(element)
        if element_node is None:
            element_node = node.find(element)
        if element_node is not None:
            return element_node.text
        else:
            return None

    def _oneline(self, s):
        lines = s.splitlines()
        result = ''
        for line in lines:
            result += ' ' + line.lstrip()
        return result.strip()

    def convert_underscores(self):
        self.prefix = self.prefix.replace('_', '-')
        self.keyname = self.keyname.replace('_', '-')

    def get_gsettings_schema_key(self):
        key = GSettingsSchemaKey()
        key.fill(self.keyname, self.varianttype, self.default, self.typed_default, self.localized, self.keyname, self.short, self.long, None, None)
        return key


######################################


class GConfSchemaParser:

    def __init__(self, file, default_gettext_domain, default_schema_id, keep_underscores):
        self.file = file
        self.default_gettext_domain = default_gettext_domain
        self.default_schema_id = default_schema_id
        self.keep_underscores = keep_underscores

        self.root = None
        self.default_schema_id_count = 0

    def _insert_schema(self, gconf_schema):
        if not self.keep_underscores:
            gconf_schema.convert_underscores()

        schemas_only = (gconf_schema.applyto is None)

        dirpath = gconf_schema.prefix
        if dirpath[0] != '/':
            raise GSettingsSchemaConvertException('Key \'%s\' has a relative path. There is no relative path in GSettings schemas.' % gconf_schema.applyto or gconf_schema.key)

        # remove leading 'schemas/' for schemas-only keys
        if schemas_only and dirpath.startswith('/schemas/'):
            dirpath = dirpath[len('/schemas'):]

        if len(dirpath) == 1:
            raise GSettingsSchemaConvertException('Key \'%s\' is a toplevel key. Toplevel keys are not accepted in GSettings schemas.' % gconf_schema.applyto or gconf_schema.key)

        # remove trailing slash because we'll split the string
        if dirpath[-1] == '/':
            dirpath = dirpath[:-1]
        # and also remove leading slash when splitting
        hierarchy = dirpath[1:].split('/')

        # we don't want to put apps/ and desktop/ keys in the same schema,
        # so we have a first step where we make sure to create a new schema
        # to avoid this case if necessary
        gsettings_schema = None
        for schema in self.root.schemas:
            if schemas_only:
                schema_path = schema._hacky_path
            else:
                schema_path = schema.path
            if dirpath.startswith(schema_path):
                gsettings_schema = schema
                break
        if not gsettings_schema:
            gsettings_schema = GSettingsSchema()
            if schemas_only:
                gsettings_schema._hacky_path = '/' + hierarchy[0] + '/'
            else:
                gsettings_schema.path = '/' + hierarchy[0] + '/'
            self.root.schemas.append(gsettings_schema)

        # we create the schema hierarchy that leads to this key
        gsettings_dir = gsettings_schema
        for item in hierarchy[1:]:
            subdir = None
            for child in gsettings_dir.children:
                if child.name == item:
                    subdir = child
                    break
            if not subdir:
                subdir = GSettingsSchema()
                # note: the id will be set later on
                if gsettings_dir.path:
                    subdir.path = '%s%s/' % (gsettings_dir.path, item)
                subdir.name = item
                gsettings_dir.children.append(subdir)
            gsettings_dir = subdir

        # we have the final directory, so we can put the key there
        gsettings_dir.keys.append(gconf_schema.get_gsettings_schema_key())

    def _set_children_id(self, schema):
        for child in schema.children:
            child.id = '%s.%s' % (schema.id, child.name)
            self._set_children_id(child)

    def _fix_hierarchy(self):
        for schema in self.root.schemas:
            # we created one schema per level, starting at the root level;
            # however, we don't need to go that far and we can simplify the
            # hierarchy
            while len(schema.children) == 1 and not schema.keys:
                child = schema.children[0]
                schema.children = child.children
                schema.keys = child.keys
                if schema.path:
                    schema.path += child.name + '/'

            # now that we have a toplevel schema, set the id
            if self.default_schema_id:
                schema.id = self.default_schema_id
                if self.default_schema_id_count > 0:
                    schema.id += '.FIXME-%s' % self.default_schema_id_count
                self.default_schema_id_count += 1
            else:
                schema.id = 'FIXME'
            self._set_children_id(schema)

    def parse(self):
        # reset the state of the parser
        self.root = GSettingsSchemaRoot()
        if self.default_gettext_domain:
            self.root.gettext_domain = self.default_gettext_domain
        self.default_schema_id_count = 0

        gconfschemafile_node = ET.parse(self.file).getroot()
        for schemalist_node in gconfschemafile_node.findall('schemalist'):
            for schema_node in schemalist_node.findall('schema'):
                gconf_schema = GConfSchema(schema_node)
                if gconf_schema.localized and not self.root.gettext_domain:
                    self.root.gettext_domain = 'FIXME'
                self._insert_schema(gconf_schema)

        self._fix_hierarchy()

        return self.root


######################################


def main(args):
    parser = optparse.OptionParser()

    parser.add_option("-o", "--output", dest="output",
                      help="output file")
    parser.add_option("-g", "--gconf", action="store_true", dest="gconf",
                      default=False, help="convert a gconf schema file")
    parser.add_option("-d", "--gettext-domain", dest="gettext_domain",
                      help="default gettext domain to use when converting gconf schema file")
    parser.add_option("-i", "--schema-id", dest="schema_id",
                      help="default schema ID to use when converting gconf schema file")
    parser.add_option("-u", "--keep-underscores", action="store_true", dest="keep_underscores",
                      help="keep underscores in key names instead of replacing them with dashes when converting gconf schema file")
    parser.add_option("-s", "--simple", action="store_true", dest="simple",
                      default=False, help="use the simple schema format as output")
    parser.add_option("-x", "--xml", action="store_true", dest="xml",
                      default=False, help="use the xml schema format as output")
    parser.add_option("-f", "--force", action="store_true", dest="force",
                      default=False, help="overwrite output file if already existing")

    (options, args) = parser.parse_args()

    if len(args) < 1:
        print >> sys.stderr, 'Need a filename to work on.'
        return 1
    elif len(args) > 1:
        print >> sys.stderr, 'Too many arguments.'
        return 1

    if options.simple and options.xml:
        print >> sys.stderr, 'Too many output formats requested.'
        return 1

    if not options.gconf and options.gettext_domain:
        print >> sys.stderr, 'Default gettext domain can only be specified when converting a gconf schema.'
        return 1

    if not options.gconf and options.schema_id:
        print >> sys.stderr, 'Default schema ID can only be specified when converting a gconf schema.'
        return 1

    if not options.gconf and options.keep_underscores:
        print >> sys.stderr, 'The --keep-underscores option can only be specified when converting a gconf schema.'
        return 1

    argfile = os.path.expanduser(args[0])
    if not os.path.exists(argfile):
        print >> sys.stderr, '\'%s\' does not exist.' % argfile
        return 1

    if options.output:
        options.output = os.path.expanduser(options.output)

    try:
        if options.output and not options.force and os.path.exists(options.output):
            raise GSettingsSchemaConvertException('\'%s\' already exists. Use --force to overwrite it.' % options.output)

        if options.gconf:
            if not options.simple and not options.xml:
                options.xml = True

            try:
                parser = GConfSchemaParser(argfile, options.gettext_domain, options.schema_id, options.keep_underscores)
                schema_root = parser.parse()
            except SyntaxError, e:
                raise GSettingsSchemaConvertException('\'%s\' does not look like a valid gconf schema file: %s' % (argfile, e))
        else:
            # autodetect if file is XML or not
            try:
                parser = XMLSchemaParser(argfile)
                schema_root = parser.parse()
                if not options.simple and not options.xml:
                    options.simple = True
            except SyntaxError, e:
                parser = SimpleSchemaParser(argfile)
                schema_root = parser.parse()
                if not options.simple and not options.xml:
                    options.xml = True

        if options.xml:
            node = schema_root.get_xml_node()
            try:
                output = ET.tostring(node, pretty_print = True)
            except TypeError:
                # pretty_print only works with lxml
                output = ET.tostring(node)
        else:
            output = schema_root.get_simple_string()

        if not options.output:
            sys.stdout.write(output)
        else:
            try:
                fout = open(options.output, 'w')
                fout.write(output)
                fout.close()
            except GSettingsSchemaConvertException, e:
                fout.close()
                if os.path.exists(options.output):
                    os.unlink(options.output)
                raise e

    except GSettingsSchemaConvertException, e:
        print >> sys.stderr, '%s' % e
        return 1

    return 0


if __name__ == '__main__':
    try:
        res = main(sys.argv)
        sys.exit(res)
    except KeyboardInterrupt:
        pass