path: root/src/pydrilla/pydrilla.py

# SPDX-License-Identifier: AGPL-3.0-or-later

# Main repository logic.
#
# This file is part of Hydrilla
#
# Copyright (C) 2021 Wojtek Kosior
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as
# published by the Free Software Foundation, either version 3 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 Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program.  If not, see <https://www.gnu.org/licenses/>.
#
#
# I, Wojtek Kosior, thereby promise not to sue for violation of this
# file's license. Although I request that you do not make use this code
# in a proprietary program, I am not going to enforce this in court.

from flask import Flask, Blueprint, current_app, url_for, abort, request
from jinja2 import Environment, PackageLoader
import re
#from hashlib import sha256
import os
import pathlib
import json
import gettext
import logging

SCHEMA_VERSION = [0, 2]

strip_comment_re = re.compile(r'''
^ # match from the beginning of each line
( # catch the part before '//' comment
  (?: # this group matches either a string or a single out-of-string character
    [^"/] |
    "
    (?: # this group matches any in-a-string character
      [^"\\] |          # match any normal character
      \\[^u] |          # match any escaped character like '\f' or '\n'
      \\u[a-fA-F0-9]{4} # match an escape
    )*
    "
  )*
)
# expect either end-of-line or a comment:
# * unterminated strings will cause matching to fail
# * bad comment (with '/' instead of '//') will be indicated by second group
#   having length 1 instead of 2 or 0
(//?|$)
''', re.VERBOSE)

def strip_json_comments(text):
    processed = 0
    stripped_text = []
    for line in text.split('\n'):
        match = strip_comment_re.match(line)

        if match is None: # unterminated string
            # ignore this error, let json module report it
            stripped = line
        elif len(match[2]) == 1:
            raise json.JSONDecodeError('bad comment', text,
                                       processed + len(match[1]))
        else:
            stripped = match[1]

        stripped_text.append(stripped)
        processed += len(line) + 1

    return '\n'.join(stripped_text)

here = pathlib.Path(__file__).resolve().parent

bp = Blueprint('bp', __package__)

def load_config(config_path):
    config = {}
    to_load = [config_path]
    failures_ok = [False]

    while to_load:
        path = to_load.pop()
        can_fail = failures_ok.pop()

        try:
            with open(config_path) as config_file:
                new_config = json.loads(strip_json_comments(config_file.read()))
        except Exception as e:
            if can_fail:
                continue
            raise e from None

        config.update(new_config)

        for key, failure_ok in [('try_configs', True), ('use_configs', False)]:
            paths = new_config.get(key, [])
            paths.reverse()
            to_load.extend(paths)
            failures_ok.extend([failure_ok] * len(paths))

    for key in ['try_configs', 'use_configs']:
        if key in config:
            config.pop(key)

    return config

def get_content_file_path(path):
    if os.path.sep != '/':
        path.replace('/', os.path.sep)

    path = pathlib.Path(path)
    if path.is_absolute():
        raise ValueError(_('path_is_absolute_{}').format(path))

    return path

class MyNotImplError(NotImplementedError):
    '''Raised when a planned but not-yet-completed feature is used.'''
    def __init__(self, what, where):
        super().__init__(_('not_implemented_{what}_{where}')
                         .format(what=what, where=where))

def normalize_version(ver):
    '''
    ver is an array of integers. Strip right-most zeroes from ver.

    Returns a *new* array. Doesn't modify its argument.
    '''
    new_len = 0
    for i, num in enumerate(ver):
        if num != 0:
            new_len = i + 1

    return ver[:new_len]

def parse_version(ver_str):
    '''
    Convert ver_str into an array representation, e.g. for ver_str="4.6.13.0"
    return [4, 6, 13, 0].
    '''
    return [int(num) for num in ver_str.split('.')]

def version_string(ver, rev=None):
    '''
    ver is an array of integers. rev is an optional integer. Produce string
    representation of version (optionally with revision number), like:
        1.2.3-5
    No version normalization is performed.
    '''
    return '.'.join([str(n) for n in ver]) + ('' if rev is None else f'-{rev}')

class VersionedContentItem:
    '''Stores definitions of multiple versions of website content item.'''
    def __init__(self):
        self.uuid = None
        self.identifier = None
        self.by_version = {}
        self.known_versions = []

    def register_item(self, item):
        if self.identifier is None:
            self.identifier = item['identifier']
            self.uuid = item['uuid']
        elif self.uuid != item['uuid']:
            raise ValueError(_('uuid_mismatch_{identifier}')
                             .format(identifier=self.identifier))

        ver = item['version']
        ver_str = version_string(ver)

        if ver_str in self.by_version:
            raise ValueError(_('version_clash_{identifier}_{version}')
                             .format(identifier=self.identifier,
                                     version=ver_str))

        self.by_version[ver_str] = item
        self.known_versions.append(ver)

class PatternTreeNode:
    '''
    "Pattern Tree" is how we refer to the data structure used for querying
    Haketilo patterns. Those look like 'https://*.example.com/ab/***'. The goal
    is to make it possible for given URL to quickly retrieve all known patterns
    that match it.
    '''
    def __init__(self):
        self.wildcard_matches = [None, None, None]
        self.literal_match    = None
        self.children         = {}

    def search(self, segments):
        '''
        Yields all matches of this segments sequence against the tree that
        starts at this node. Results are produces in order from greatest to
        lowest pattern specificity.
        '''
        nodes = [self]

        for segment in segments:
            next_node = nodes[-1].children.get(segment)
            if next_node is None:
                break

            nodes.append(next_node)

        nsegments = len(segments)
        cond_literal = lambda: len(nodes)     == nsegments
        cond_wildcard = [
            lambda: len(nodes) + 1 == nsegments and segments[-1] != '*',
            lambda: len(nodes) + 1 <  nsegments,
            lambda: len(nodes) + 1 != nsegments or  segments[-1] != '***'
        ]

        while nodes:
            node = nodes.pop()

            for item, condition in [(node.literal_match, cond_literal),
                                    *zip(node.wildcard_matches, cond_wildcard)]:
                if item is not None and condition():
                    yield item

    def add(self, segments, item_instantiator):
        '''
        Make item queryable through (this branch of) the Pattern Tree. If there
        was not yet any item associated with the tree path designated by
        segments, create a new one using item_instantiator() function. Return
        all items matching this path (both the ones that existed and the ones
        just created).
        '''
        node = self

        for i, segment in enumerate(segments):
            wildcards = node.wildcard_matches

            child = node.children.get(segment) or PatternTreeNode()
            node.children[segment] = child
            node = child

        if node.literal_match is None:
            node.literal_match = item_instantiator()

        if segment not in ('*', '**', '***'):
            return [node.literal_match]

        if wildcards[len(segment) - 1] is None:
            wildcards[len(segment) - 1] = item_instantiator()

        return [node.literal_match, wildcards[len(segment) - 1]]

proto_regex  = re.compile(r'^(?P<proto>\w+)://(?P<rest>.*)$')
user_re      = r'[^/?#@]+@' # r'(?P<user>[^/?#@]+)@' # discarded for now
query_re     = r'\??[^#]*'  # r'\??(?P<query>[^#]*)' # discarded for now
domain_re    = r'(?P<domain>[^/?#]+)'
path_re      = r'(?P<path>[^?#]*)'
http_regex   = re.compile(f'{domain_re}{path_re}{query_re}.*')
ftp_regex    = re.compile(f'(?:{user_re})?{domain_re}{path_re}.*')

class UrlError(ValueError):
    pass

class DeconstructedUrl:
    '''Represents a deconstructed URL or URL pattern'''
    def __init__(self, url):
        self.url = url

        match = proto_regex.match(url)
        if not match:
            raise UrlError(_('invalid_URL_{}').format(url))

        self.proto = match.group('proto')
        if self.proto not in ('http', 'https', 'ftp'):
            raise UrlError(_('disallowed_protocol_{}').format(proto))

        if self.proto == 'ftp':
            match = ftp_regex.match(match.group('rest'))
        elif self.proto in ('http', 'https'):
            match = http_regex.match(match.group('rest'))

        if not match:
            raise UrlError(_('invalid_URL_{}').format(url))

        self.domain = match.group('domain').split('.')
        self.domain.reverse()
        self.path = [*filter(None, match.group('path').split('/'))]

class MappingItem:
    '''
    A mapping, together with one of its patterns, as stored in Pattern Tree.
    '''
    def __init__(self, pattern, mapping):
        self.pattern = pattern
        self.mapping = mapping

    def register(self, patterns_by_proto):
        '''
        Make self queryable through the Pattern Tree that starts with the
        protocols dictionary passed in the argument.
        '''
        deco = DeconstructedUrl(self.pattern)

        domain_tree = patterns_by_proto.get(deco.proto) or PatternTreeNode()
        patterns_by_proto[deco.proto] = domain_tree

        for path_tree in domain_tree.add(deco.domain, PatternTreeNode):
            for match_list in path_tree.add(deco.path, list):
                match_list.append(self)

class Content:
    '''Stores serveable website content.'''
    def __init__(self):
        self.resources = {}
        self.mappings  = {}
        self.licenses  = {}
        self.indexes   = {}
        self.definition_processors = {
            'resource': self.process_resource_or_mapping,
            'mapping': self.process_resource_or_mapping,
            'license': self.process_license
        }
        self.patterns_by_proto = {}

    @staticmethod
    def register_item(dict, item):
        '''
        Helper function used to add a versioned item definition to content
        data structures.
        '''
        identifier = item['identifier']
        versioned_item = dict.get(identifier)
        if versioned_item is None:
            versioned_item = VersionedContentItem()
            dict[identifier] = versioned_item

        versioned_item.register_item(item)

    @staticmethod
    def _process_copyright_and_license(definition):
        '''Helper function used by other process_*() methods.'''
        for field in ['copyright', 'licenses']:
            if definition[field] == 'auto':
                raise MyNotImplError(f'"{{field}}": "auto"',
                                     definition['source_name'])

    def process_resource_or_mapping(self, definition, index):
        '''
        Sanitizes, autocompletes and registers serveable mapping/resource
        definition.
        '''
        definition['version'] = normalize_version(definition['version'])

        if definition['type'] == 'resource':
            self._process_copyright_and_license(definition)
            definition['dependencies'] = definition.get('dependencies', [])
            self.register_item(self.resources, definition)
        else:
            self.register_item(self.mappings, definition)

    def process_license(self, license, index):
        '''Sanitizes and registers serveable license definition.'''
        identifier = license['identifier']
        if identifier in self.licenses:
            raise ValueError(_('license_clash_{}').format(identifier))

        self.licenses[identifier] = license

    def process_index(self, index, source_name):
        '''
        Sanitizes, autocompletes and registers data from a loaded index.json
        file.
        '''
        schema_ver = normalize_version(index['schema_version'])
        index['schema_version'] = schema_ver
        if schema_ver != SCHEMA_VERSION:
            raise ValueError('index_json_schema_mismatch_{found}_{required}'
                             .format(found=version_string(schema_ver),
                                     required=version_string(SCHEMA_VERSION)))

        if source_name in self.indexes:
            raise ValueError(_('source_name_clash_{}').format(source_name))

        index['source_name'] = source_name

        self._process_copyright_and_license(index)

        self.indexes[source_name] = index

        for definition in index['definitions']:
            try:
                definition['source_name'] = source_name
                definition['source_copyright'] = index['copyright']
                definition['source_licenses'] = index['licenses']
                processor = self.definition_processors[definition['type']]
                processor(definition, index)
            except Exception as e:
                if current_app._pydrilla_werror:
                    raise e from None
                logging.error(_('couldnt_load_definition_from_%s'), subdir_path,
                              exc_info=True)
    @staticmethod
    def all_items(versioned_items_dict):
        '''Iterator over all registered versions of all items.'''
        for versioned_item in versioned_items_dict.values():
            for item in versioned_item.by_version.values():
                yield item

    def report_missing(self):
        '''
        Use logger to print information about items that are referenced but
        were not loaded.
        '''
        def report_missing_license(object, object_type, lic):
            if object_type == 'index':
                logging.error(_('no_index_license_%(source)s_%(lic)s'),
                              source=object['source_name'], lic=lic)
                return

            ver_str = version_string(object['version'])
            kwargs = {object_type: object['identifier'], ver: ver_str, lic: lic}
            if object_type == 'resource':
                fmt = _('no_resource_license_%(resource)s_%(ver)s_%(lic)s')
            else:
                fmt = _('no_mapping_license_%(mapping)s_%(ver)s_%(lic)s')

            logging.error(fmt, **kwargs)

        for object_type, iterable in [
                ('index',    self.indexes.values()),
                ('resource', self.all_items(self.resources))
        ]:
            for object in iterable:
                to_process = [object['licenses']]
                licenses = []
                while to_process:
                    term = to_process.pop()

                    if type(term) is str:
                        if term not in ['or', 'and'] and \
                           term not in self.licenses:
                            report_missing_license(object, object_type, lic)
                        continue

                    to_process.extend(term)

        def report_missing_dependency(resource, dep):
            logging.error(_('no_dep_%(resource)s_%(ver)s_%(dep)s'),
                          dep=dep, resource=resource['identifier'],
                          ver=version_string(resource['version']))

        for resource in self.all_items(self.resources):
            for dep in resource['dependencies']:
                if dep not in self.resources:
                    report_missing_dependency(resource, dep)

        def report_missing_payload(mapping, payload):
            logging.error(_('no_payload_%(mapping)s_%(ver)s_%(payload)s'),
                          mapping=mapping['identifier'], payload=payload,
                          ver=version_string(mapping['version']))

        for mapping in self.all_items(self.mappings):
            for payload in mapping['payloads']:
                payload = payload['payload']
                if payload not in self.resources:
                    report_missing_payload(mapping, payload)

    def finalize(self):
        '''
        Initialize structures needed to serve queries. Called once after all
        data gets loaded.
        '''
        for dict in [self.resources, self.mappings]:
            for versioned_item in dict.values():
                versioned_item.known_versions.sort()

        for mapping in self.all_items(self.mappings):
            for payload in mapping['payloads']:
                pattern = payload['pattern']
                try:
                    MappingItem(pattern, mapping)\
                        .register(self.patterns_by_proto)
                except Exception as e:
                    if current_app._pydrilla_werror:
                        raise e from None
                    logging.error(
                        _('couldnt_register_%(mapping)s_%(ver)s_%(pattern)s'),
                        mapping=mapping['identifier'], pattern=pattern,
                        ver=version_string(mapping['version'])
                    )

    def find_item(self, type, identifier, ver=None):
        '''
        Find and return definition of the newest version of resource/mapping
        named by identifier. If no such resource/mapping exists, return None.

        If ver is specified, instead find and return definition of that version
        of the item (or None is absent).
        '''
        dict = self.resources if type == 'resource' else self.mappings
        versioned_item = dict.get(identifier)
        if not versioned_item:
            return None

        ver = version_string(ver or versioned_item.known_versions[-1])

        return versioned_item.by_version.get(ver)

    def query(self, url, max=0):
        '''
        Return return registered patterns and mappings (available as
        MappingItems) that match url. The maximum number of items yielded may be
        limited by using the optional max argument. Its default value, 0, causes
        no limit to be imposed.

        If multiple versions of a mapping are applicable, only the most recent
        is included in the result.
        '''
        deco = DeconstructedUrl(url)

        domain_tree = self.patterns_by_proto.get(deco.proto) \
            or PatternTreeNode()
        for path_tree in domain_tree.search(deco.domain):
            for item in path_tree.search(deco.path):
                if url[-1] == '/' or item.pattern[-1] != '/':
                    yield item
                    max -= 1
                    if max == 0:
                        return

def load_content_from_subdir(subdir_path, source_name, content):
    index_path = subdir_path / 'index.json'
    with open(index_path) as index_file:
        index = json.loads(strip_json_comments(index_file.read()))

    content.process_index(index, source_name)

def load_content(path):
    path = pathlib.Path(path)
    if not path.is_dir():
        raise ValueError(_('content_dir_path_not_dir'))

    content = Content()

    for subdir_path in path.iterdir():
        if not subdir_path.is_dir():
            continue
        try:
            load_content_from_subdir(subdir_path, subdir_path.name, content)
        except Exception as e:
            if current_app._pydrilla_werror:
                raise e from None
            logging.error(_('couldnt_load_content_from_%s'), subdir_path,
                          exc_info=True)

    content.report_missing()
    content.finalize()

    return content

def create_app(config_path=(here / 'config.json'), flask_config={}):
    app = Flask(__package__)
    app.config.update(flask_config)

    language = flask_config.get('lang', 'en')
    translation = gettext.translation('pydrilla', localedir=(here / 'locales'),
                                      languages=[language])

    app._pydrilla_gettext = translation.gettext

    # https://stackoverflow.com/questions/9449101/how-to-stop-flask-from-initialising-twice-in-debug-mode
    if app.debug and os.environ.get('WERKZEUG_RUN_MAIN') != 'true':
        return app

    config = load_config(config_path)
    for key in ['static_resource_uri', 'content_dir']:
        if key not in config:
            raise ValueError(_('config_key_absent_{}').format(key))

    app._pydrilla_static_resource_uri = config['static_resource_uri']
    app._pydrilla_werror = config.get('werror', False)
    if 'hydrilla_parent' in config:
        raise MyNotImplError('hydrilla_parent', config_path.name)
    with app.app_context():
        app._pydrilla_content = load_content(config['content_dir'])

    app.register_blueprint(bp)

    return app

def _(text_key):
    return current_app._pydrilla_gettext(text_key)

def escaping_gettext(text_key):
    from markupsafe import escape

    return str(escape(_(text_key)))

class MyEnvironment(Environment):
    '''
    A wrapper class around jinja2.Environment that causes GNU gettext function
    (as '_' and '__') and url_for function to be passed to every call of each
    template's render() method.
    '''

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def get_template(self, *args, **kwargs):
        template = super().get_template(*args, **kwargs)
        old_render = template.render

        def new_render(*args, **kwargs):
            final_kwargs = {
                '_': escaping_gettext,
                '__': escaping_gettext,
                'url_for': url_for
            }
            final_kwargs.update(kwargs)

            return old_render(*args, **final_kwargs)

        template.render = new_render

        return template

j2env = MyEnvironment(loader=PackageLoader(__package__), autoescape=False)

indexpage = j2env.get_template('index.html')
@bp.route('/')
def index():
    return indexpage.render()

for item_type in ['mapping', 'resource']:
    def item(identifier, item_type=item_type):
        ver = request.args.get('ver')
        if ver is not None:
            try:
                ver = normalize_version(parse_version(ver))
            except:
                abort(400)

        definition = current_app._pydrilla_content\
                                .find_item(item_type, identifier, ver)
        if definition is None:
            abort(404)

        return json.dumps(definition)

    bp.add_url_rule(f'/{item_type}s/<string:identifier>', item_type, item)