doku2git/doku2git.py

#!/usr/bin/env python3


###############################################################################
from typing import List, Mapping, Iterable, Callable, Generator
from operator import attrgetter

import os.path
import os
import collections
import itertools


###############################################################################
import logging

logging.basicConfig(level = logging.WARNING)
log = logging.getLogger()


###############################################################################
def find_index(haystack: Iterable, needle, comparator: Callable) -> int:
    """
    Find the index of the first value within 'haystack' that compares as equal
    to 'needle' using the supplied 'comparator'
    """
    for index, data in zip(itertools.count(), haystack):
        if comparator(data, needle):
            return index

    raise ValueError("Needle not found")


###############################################################################
class User(object):
    """
    Represents a single user within the Dokuwiki namespace.

    We deliberately do not store the password hash.
    """

    login: str
    real_name: str
    email: str

    def __init__(self, login, real_name, email):
        self.login = login
        self.real_name = real_name
        self.email = email

    def author(self) -> str:
        """
        Return authorship information formatted for `git commit`.
        eg. 'A U Thor <author@example.com>'
        """
        return f"{self.real_name} <{self.email}>"


##-----------------------------------------------------------------------------
class Change(object):
    """
    The base class for all change records in the Dokuwiki instance.

    This class stores all the required data fields, but subclasses must
    provide some required functionality; eg, path lookups, and history
    storage, are different for Pages and Media.
    """

    timestamp: str
    ip: str
    operation: str
    page: str
    user: str
    description: str

    KNOWN_OPERATIONS = "CEeD"

    def __init__(self, timestamp, ip, operation, page, user, description):
        assert operation in Change.KNOWN_OPERATIONS, f"Operation '{operation}' is not handle"

        self.timestamp = timestamp
        self.ip = ip
        self.operation = operation
        self.page = page
        self.user = user
        self.description = description

    def __str__(self):
        return f'{self.user}@{self.timestamp}:{self.page} "{self.description}"'

    def _delete(self, src: str, dst: str) -> Generator[str, None, None]:
        """
        Generates the required commands to delete a given page
        :param src: The Dokuwiki instance the change belongs to
        :param dst: The path to the root of the git repository.
        """
        localpath = self.page.replace(':', os.path.sep)
        yield f'git rm --quiet "{localpath}.txt"'

    def apply(self, src, dst: str) -> Generator[str, None, None]:
        """
        Generate a list of commands to enact this changeset from the Dokuwiki
        instance; eg, create, edit, or delete a page.

        The actual operation is not performed within this function. Instead, it's
        dispatched to (potentially overridden) member functions.

        Some basic configure time checks are performed; primarily for the
        existence of the required source files.

        :param src: The Dokuwiki instance this change comes from
        :param dst: The path to the root of the git repository.
        """
        log.info(f"Applying {self.operation}: '{self.description}'")

        # If we're not trying to delete a page then ensure we can see the
        # source material in the Dokuwiki attic
        if self.operation not in ('D'):
            attic = os.path.join(src.root, self._attic(src))
            if not os.path.isfile(attic):
                log.error(f'Source file {attic} not present for {self}; ignoring')
                return

        # Find the function that will apply this update. Deliberately
        # construct this to throw if the operation isn't handled.
        func = {
            'C': self._update,
            'E': self._update,
            'e': self._update,
            'D': self._delete
        }[self.operation]

        # Generate the subcommands using the update function, but protect
        # against failure.
        for cmd in func(src, dst):
            yield f'{cmd} || die "{self.timestamp}"'

        # Finalise the update with a commit
        yield " ".join([
            'git commit',
            '--quiet',
            '--allow-empty',
            '--allow-empty-message',
            f'--author="{src.users[self.user].author()}"',
            f' --date="{self.timestamp} +0000"',
            f'-m "{self.description}"',
            f'|| die "{self.timestamp}"'
        ])


##-----------------------------------------------------------------------------
class PageChange(Change):
    """
    Represents one change to one page in the Dokuwiki instance.
    """
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def _attic(self, src) -> str:
        """
        Find the path to the history data within the Dokuwiki directories.
        :param src: The Dokuwiki instance this change belongs to.
        :return: The relative path from the root of Dokuwiki to the compressed
        data. This is always gzip compressed.
        """
        local = self.page.replace(':', os.path.sep)

        return os.path.join(
            'data',
            'attic',
            f'{local}.{self.timestamp}.txt.gz'
        )

    def _update(self, src, dst: str) -> Generator[str, None, None]:
        """
        Yield the commands required to unpack this change.

        It is assumed we don't need to remove old data and can just uncompress
        over the top of the old data.
        :param src: The Dokuwiki instance this change belongs to.
        :param dst: The absolute path to the root of the git repository
        :return:
        """
        localpath = self.page.replace(':', os.path.sep)

        dstpath = os.path.join(dst, localpath)
        dstdir = os.path.dirname(dstpath)

        attic = os.path.join(src.root, self._attic(src))

        cmds = [
            f'mkdir -p "{dstdir}"',
            f'gunzip -c "{attic}" > "{dstpath}.txt"',
            f'git add "{localpath}".txt',

        ]

        for c in cmds:
            yield c


##-----------------------------------------------------------------------------
class MediaChange(Change):
    """
    Represents changes to one media file in a Dokuwiki instance.
    """
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def _attic(self, src):
        """
        Calculates the path to the source data required for this change.
        :param src: The Dokuwiki instance this change belongs to
        :return: The absolute to the data
        """
        local = self.page.replace(':', os.path.sep)

        # If we're the most recent iteration of the file then we return a path
        # directly to the live media directory.
        #
        # Keep in mind that we search from the newest-to-oldest (due to the
        # reversed sorting in media list records).
        #
        # There are some situations where we get two edits with the same
        # timestamp; but we can't have additional entries in `media_attic` for
        # the same timestamp (and there doesn't appear to be an alternative
        # method to store these changes)
        #
        # We treat this as if they can be collapsed into one change for the
        # purposes of path lookups.
        changes = src.media[self.page]
        index = find_index(changes, self, lambda a, b: a.timestamp == b.timestamp)
        if index == 0:
            return os.path.join(src.root, 'data', 'media', local)

        # We're a historical entry, so we need to find the backup within
        # media_attic
        basename, extension = os.path.splitext(local)
        return os.path.join(
            src.root,
            'data',
            'media_attic',
            f'{basename}.{self.timestamp}{extension}'
        )

    def _update(self, src: str, dst: str) -> Generator[str, None, None]:
        """
        Yields the commands required to unpack this one change.
        :param src: The Dokuwiki instance this change belongs to.
        :param dst: The absolute path to the root of the git repository
        :return:
        """
        localpath = self.page.replace(':', os.path.sep)
        localdir = os.path.dirname(localpath)

        yield f'cp "{self._attic(src)}" "{os.path.join(dst, localdir)}"'


###############################################################################
class Dokuwiki:
    """
    Represents data for a single Dokuwiki instance.

    It is not safe to use this on an installation that may be receiving
    updates; but it should be safe to use on a live site without active
    editors. Either way it will never rewrite any data.
    """

    users: Mapping[str, User]
    media: Mapping[str, List[Change]]
    changes: List[Change]

    def _record_media(self, change):
        self.media[change.page].append(change)

    def __init__(self, root: str):
        self.root = root
        self.users = self._find_users()
        self.media = collections.defaultdict(list)

        page_generator = self._find_meta(os.path.join (self.root, 'data', 'meta'), PageChange)
        media_generator = self._find_meta(os.path.join (self.root, 'data', 'media_meta'), MediaChange)

        self.changes = list(page_generator)
        for entry in media_generator:
            self._record_media(entry)
            self.changes.append(entry)

        self.changes.sort(key=attrgetter('timestamp'))
        for k,v in self.media.items():
            v.sort(key=attrgetter('timestamp'), reverse=True)

    def _find_users(self) -> Mapping[str, User]:
        """
        Parses the users.auth.php file to discover all listed users.
        :return: A mapping of login to User objects
        """
        found = {}

        auth_path = os.path.join(self.root, 'conf', 'users.auth.php')
        with open(auth_path, 'r') as auth_file:
            for line in auth_file:
                # Make sure the line actually contains some data
                line = line.rstrip()
                if len(line) == 0:
                    continue
                if line[0] == '#':
                    continue

                login, password, real_name, email, groups, *tail = line.split(':')
                if len(tail) > 0:
                    log.warning("Found extra components when parsing users.auth.php")

                found[login] = User(login=login, real_name=real_name, email=email)

        found[None] = User(login="system", real_name="system", email="system@mlug-au.org")
        return found


    def _find_meta(self, root: str, type):
        for dirpath, dirnames, filenames in os.walk(root):
            for file in filenames:
                # We don't really care about 'indexed' or 'meta' files.
                base, ext = os.path.splitext(file)
                if ext != '.changes':
                    continue

                # Some paths are autogenerated listings of content that aren't
                # relevant to us (and drastically complicate processing)
                if base in ('_media', '_dokuwiki'):
                    continue

                # Actually read the metadata
                path = os.path.join(dirpath, file)
                for change in self._read_meta(path, type):
                    yield change


    def _read_meta(self, path: str, type):
        log.debug("extracting meta from %s", path)
        with open(path, 'r') as data:
            for entry in data:
                timestamp, ip, operation, page, user, description, *tail = entry.split('\t')

                if len(tail) > 1:
                    log.warning("Dropping meta change data")
                else:
                    assert tail[0] == '\n'

                if not user:
                    if description != "external edit":
                        raise RuntimeError("Empty user doesn't correspond to 'external edit'")
                    user = None

                yield type(
                    timestamp=timestamp,
                    ip=ip,
                    operation=operation,
                    page=page,
                    user=user,
                    description=description
                )


###############################################################################
if __name__ == "__main__":
    import argparse
    import json

    header = [
        '#!/usr/bin/env bash',
        'die() { echo "$*" 1>&2 ; exit 1; }',
    ]

    def preamble(src, dst: str):
        cmds = [
            f'git init "{dst}"',
            f'cd "{dst}"',
        ]

        for c in cmds:
            yield c

    def finish(src, dst: str):
        arguments = [
            f'git commit',
            ' --quiet',
            '--allow-empty',
            '--author="doku2git@mlug-au.org <doku2git>"',
            '-m "Converted dokuwiki to git"'
        ]
        yield " ".join(arguments)

    def main():
        parser = argparse.ArgumentParser()
        parser.add_argument("--src", required=True, help="The root directory of the dokuwiki source")
        parser.add_argument("--dst", required=True, help="The git output directory")
        parser.add_argument("--users", type=str, help="A JSON serialised list of supplementary users")

        args = parser.parse_args()

        src = Dokuwiki(root=args.src)

        if args.users:
            with open(args.users, 'r') as users:
                for user in json.load(users):
                    src.users[user['login']] = User(**user)

        todo = [
            header,
            preamble(src, args.dst),
            *(change.apply(src, args.dst) for change in src.changes),
            finish(src, args.dst)
        ]

        for cmd in itertools.chain(*todo):
            print(cmd)

    main()