anonymousland-synapse/synapse/storage/prepare_database.py

# -*- coding: utf-8 -*-
# Copyright 2014 - 2016 OpenMarket Ltd
# Copyright 2018 New Vector Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import fnmatch
import imp
import logging
import os
import re

from synapse.storage.engines.postgres import PostgresEngine

logger = logging.getLogger(__name__)


# Remember to update this number every time a change is made to database
# schema files, so the users will be informed on server restarts.
SCHEMA_VERSION = 56

dir_path = os.path.abspath(os.path.dirname(__file__))


class PrepareDatabaseException(Exception):
    pass


class UpgradeDatabaseException(PrepareDatabaseException):
    pass


def prepare_database(db_conn, database_engine, config):
    """Prepares a database for usage. Will either create all necessary tables
    or upgrade from an older schema version.

    If `config` is None then prepare_database will assert that no upgrade is
    necessary, *or* will create a fresh database if the database is empty.

    Args:
        db_conn:
        database_engine:
        config (synapse.config.homeserver.HomeServerConfig|None):
            application config, or None if we are connecting to an existing
            database which we expect to be configured already
    """
    try:
        cur = db_conn.cursor()
        version_info = _get_or_create_schema_state(cur, database_engine)

        if version_info:
            user_version, delta_files, upgraded = version_info

            if config is None:
                if user_version != SCHEMA_VERSION:
                    # If we don't pass in a config file then we are expecting to
                    # have already upgraded the DB.
                    raise UpgradeDatabaseException("Database needs to be upgraded")
            else:
                _upgrade_existing_database(
                    cur, user_version, delta_files, upgraded, database_engine, config
                )
        else:
            _setup_new_database(cur, database_engine)

        # check if any of our configured dynamic modules want a database
        if config is not None:
            _apply_module_schemas(cur, database_engine, config)

        cur.close()
        db_conn.commit()
    except Exception:
        db_conn.rollback()
        raise


def _setup_new_database(cur, database_engine):
    """Sets up the database by finding a base set of "full schemas" and then
    applying any necessary deltas.

    The "full_schemas" directory has subdirectories named after versions. This
    function searches for the highest version less than or equal to
    `SCHEMA_VERSION` and executes all .sql files in that directory.

    The function will then apply all deltas for all versions after the base
    version.

    Example directory structure:

        schema/
            delta/
                ...
            full_schemas/
                3/
                    test.sql
                    ...
                11/
                    foo.sql
                    bar.sql
                ...

    In the example foo.sql and bar.sql would be run, and then any delta files
    for versions strictly greater than 11.
    """
    current_dir = os.path.join(dir_path, "schema", "full_schemas")
    directory_entries = os.listdir(current_dir)

    valid_dirs = []
    pattern = re.compile(r"^\d+(\.sql)?$")

    if isinstance(database_engine, PostgresEngine):
        specific = "postgres"
    else:
        specific = "sqlite"

    specific_pattern = re.compile(r"^\d+(\.sql." + specific + r")?$")

    for filename in directory_entries:
        match = pattern.match(filename) or specific_pattern.match(filename)
        abs_path = os.path.join(current_dir, filename)
        if match and os.path.isdir(abs_path):
            ver = int(match.group(0))
            if ver <= SCHEMA_VERSION:
                valid_dirs.append((ver, abs_path))
        else:
            logger.debug("Ignoring entry '%s' in 'full_schemas'", filename)

    if not valid_dirs:
        raise PrepareDatabaseException(
            "Could not find a suitable base set of full schemas"
        )

    max_current_ver, sql_dir = max(valid_dirs, key=lambda x: x[0])

    logger.debug("Initialising schema v%d", max_current_ver)

    directory_entries = os.listdir(sql_dir)

    for filename in sorted(
        fnmatch.filter(directory_entries, "*.sql")
        + fnmatch.filter(directory_entries, "*.sql." + specific)
    ):
        sql_loc = os.path.join(sql_dir, filename)
        logger.debug("Applying schema %s", sql_loc)
        executescript(cur, sql_loc)

    cur.execute(
        database_engine.convert_param_style(
            "INSERT INTO schema_version (version, upgraded)" " VALUES (?,?)"
        ),
        (max_current_ver, False),
    )

    _upgrade_existing_database(
        cur,
        current_version=max_current_ver,
        applied_delta_files=[],
        upgraded=False,
        database_engine=database_engine,
        config=None,
        is_empty=True,
    )


def _upgrade_existing_database(
    cur,
    current_version,
    applied_delta_files,
    upgraded,
    database_engine,
    config,
    is_empty=False,
):
    """Upgrades an existing database.

    Delta files can either be SQL stored in *.sql files, or python modules
    in *.py.

    There can be multiple delta files per version. Synapse will keep track of
    which delta files have been applied, and will apply any that haven't been
    even if there has been no version bump. This is useful for development
    where orthogonal schema changes may happen on separate branches.

    Different delta files for the same version *must* be orthogonal and give
    the same result when applied in any order. No guarantees are made on the
    order of execution of these scripts.

    This is a no-op of current_version == SCHEMA_VERSION.

    Example directory structure:

        schema/
            delta/
                11/
                    foo.sql
                    ...
                12/
                    foo.sql
                    bar.py
                ...
            full_schemas/
                ...

    In the example, if current_version is 11, then foo.sql will be run if and
    only if `upgraded` is True. Then `foo.sql` and `bar.py` would be run in
    some arbitrary order.

    Args:
        cur (Cursor)
        current_version (int): The current version of the schema.
        applied_delta_files (list): A list of deltas that have already been
            applied.
        upgraded (bool): Whether the current version was generated by having
            applied deltas or from full schema file. If `True` the function
            will never apply delta files for the given `current_version`, since
            the current_version wasn't generated by applying those delta files.
    """

    if current_version > SCHEMA_VERSION:
        raise ValueError(
            "Cannot use this database as it is too "
            + "new for the server to understand"
        )

    start_ver = current_version
    if not upgraded:
        start_ver += 1

    logger.debug("applied_delta_files: %s", applied_delta_files)

    if isinstance(database_engine, PostgresEngine):
        specific_engine_extension = ".postgres"
    else:
        specific_engine_extension = ".sqlite"

    specific_engine_extensions = (".sqlite", ".postgres")

    for v in range(start_ver, SCHEMA_VERSION + 1):
        logger.info("Upgrading schema to v%d", v)

        delta_dir = os.path.join(dir_path, "schema", "delta", str(v))

        try:
            directory_entries = os.listdir(delta_dir)
        except OSError:
            logger.exception("Could not open delta dir for version %d", v)
            raise UpgradeDatabaseException(
                "Could not open delta dir for version %d" % (v,)
            )

        directory_entries.sort()
        for file_name in directory_entries:
            relative_path = os.path.join(str(v), file_name)
            logger.debug("Found file: %s", relative_path)
            if relative_path in applied_delta_files:
                continue

            absolute_path = os.path.join(dir_path, "schema", "delta", relative_path)
            root_name, ext = os.path.splitext(file_name)
            if ext == ".py":
                # This is a python upgrade module. We need to import into some
                # package and then execute its `run_upgrade` function.
                module_name = "synapse.storage.v%d_%s" % (v, root_name)
                with open(absolute_path) as python_file:
                    module = imp.load_source(module_name, absolute_path, python_file)
                logger.info("Running script %s", relative_path)
                module.run_create(cur, database_engine)
                if not is_empty:
                    module.run_upgrade(cur, database_engine, config=config)
            elif ext == ".pyc" or file_name == "__pycache__":
                # Sometimes .pyc files turn up anyway even though we've
                # disabled their generation; e.g. from distribution package
                # installers. Silently skip it
                continue
            elif ext == ".sql":
                # A plain old .sql file, just read and execute it
                logger.info("Applying schema %s", relative_path)
                executescript(cur, absolute_path)
            elif ext == specific_engine_extension and root_name.endswith(".sql"):
                # A .sql file specific to our engine; just read and execute it
                logger.info("Applying engine-specific schema %s", relative_path)
                executescript(cur, absolute_path)
            elif ext in specific_engine_extensions and root_name.endswith(".sql"):
                # A .sql file for a different engine; skip it.
                continue
            else:
                # Not a valid delta file.
                logger.warning(
                    "Found directory entry that did not end in .py or .sql: %s",
                    relative_path,
                )
                continue

            # Mark as done.
            cur.execute(
                database_engine.convert_param_style(
                    "INSERT INTO applied_schema_deltas (version, file) VALUES (?,?)"
                ),
                (v, relative_path),
            )

            cur.execute("DELETE FROM schema_version")
            cur.execute(
                database_engine.convert_param_style(
                    "INSERT INTO schema_version (version, upgraded) VALUES (?,?)"
                ),
                (v, True),
            )


def _apply_module_schemas(txn, database_engine, config):
    """Apply the module schemas for the dynamic modules, if any

    Args:
        cur: database cursor
        database_engine: synapse database engine class
        config (synapse.config.homeserver.HomeServerConfig):
            application config
    """
    for (mod, _config) in config.password_providers:
        if not hasattr(mod, "get_db_schema_files"):
            continue
        modname = ".".join((mod.__module__, mod.__name__))
        _apply_module_schema_files(
            txn, database_engine, modname, mod.get_db_schema_files()
        )


def _apply_module_schema_files(cur, database_engine, modname, names_and_streams):
    """Apply the module schemas for a single module

    Args:
        cur: database cursor
        database_engine: synapse database engine class
        modname (str): fully qualified name of the module
        names_and_streams (Iterable[(str, file)]): the names and streams of
            schemas to be applied
    """
    cur.execute(
        database_engine.convert_param_style(
            "SELECT file FROM applied_module_schemas WHERE module_name = ?"
        ),
        (modname,),
    )
    applied_deltas = set(d for d, in cur)
    for (name, stream) in names_and_streams:
        if name in applied_deltas:
            continue

        root_name, ext = os.path.splitext(name)
        if ext != ".sql":
            raise PrepareDatabaseException(
                "only .sql files are currently supported for module schemas"
            )

        logger.info("applying schema %s for %s", name, modname)
        for statement in get_statements(stream):
            cur.execute(statement)

        # Mark as done.
        cur.execute(
            database_engine.convert_param_style(
                "INSERT INTO applied_module_schemas (module_name, file)" " VALUES (?,?)"
            ),
            (modname, name),
        )


def get_statements(f):
    statement_buffer = ""
    in_comment = False  # If we're in a /* ... */ style comment

    for line in f:
        line = line.strip()

        if in_comment:
            # Check if this line contains an end to the comment
            comments = line.split("*/", 1)
            if len(comments) == 1:
                continue
            line = comments[1]
            in_comment = False

        # Remove inline block comments
        line = re.sub(r"/\*.*\*/", " ", line)

        # Does this line start a comment?
        comments = line.split("/*", 1)
        if len(comments) > 1:
            line = comments[0]
            in_comment = True

        # Deal with line comments
        line = line.split("--", 1)[0]
        line = line.split("//", 1)[0]

        # Find *all* semicolons. We need to treat first and last entry
        # specially.
        statements = line.split(";")

        # We must prepend statement_buffer to the first statement
        first_statement = "%s %s" % (statement_buffer.strip(), statements[0].strip())
        statements[0] = first_statement

        # Every entry, except the last, is a full statement
        for statement in statements[:-1]:
            yield statement.strip()

        # The last entry did *not* end in a semicolon, so we store it for the
        # next semicolon we find
        statement_buffer = statements[-1].strip()


def executescript(txn, schema_path):
    with open(schema_path, "r") as f:
        for statement in get_statements(f):
            txn.execute(statement)


def _get_or_create_schema_state(txn, database_engine):
    # Bluntly try creating the schema_version tables.
    schema_path = os.path.join(dir_path, "schema", "schema_version.sql")
    executescript(txn, schema_path)

    txn.execute("SELECT version, upgraded FROM schema_version")
    row = txn.fetchone()
    current_version = int(row[0]) if row else None
    upgraded = bool(row[1]) if row else None

    if current_version:
        txn.execute(
            database_engine.convert_param_style(
                "SELECT file FROM applied_schema_deltas WHERE version >= ?"
            ),
            (current_version,),
        )
        applied_deltas = [d for d, in txn]
        return current_version, applied_deltas, upgraded

    return None