Minor updates, and docs, for schema delta files (#11823)

* Make functions in python deltas optional

It's annoying to always have to write stubs for these.

* Documentation for delta files

* changelog
This commit is contained in:
Richard van der Hoff 2022-01-25 14:11:13 +00:00 committed by GitHub
parent 4210143f53
commit fc8598bc87
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 61 additions and 3 deletions

1
changelog.d/11823.misc Normal file
View File

@ -0,0 +1 @@
Minor updates and documentation for database schema delta files.

View File

@ -96,6 +96,60 @@ Ensure postgres is installed, then run:
NB at the time of writing, this script predates the split into separate `state`/`main`
databases so will require updates to handle that correctly.
## Delta files
Delta files define the steps required to upgrade the database from an earlier version.
They can be written as either a file containing a series of SQL statements, or a Python
module.
Synapse remembers which delta files it has applied to a database (they are stored in the
`applied_schema_deltas` table) and will not re-apply them (even if a given file is
subsequently updated).
Delta files should be placed in a directory named `synapse/storage/schema/<database>/delta/<version>/`.
They are applied in alphanumeric order, so by convention the first two characters
of the filename should be an integer such as `01`, to put the file in the right order.
### SQL delta files
These should be named `*.sql`, or — for changes which should only be applied for a
given database engine — `*.sql.posgres` or `*.sql.sqlite`. For example, a delta which
adds a new column to the `foo` table might be called `01add_bar_to_foo.sql`.
Note that our SQL parser is a bit simple - it understands comments (`--` and `/*...*/`),
but complex statements which require a `;` in the middle of them (such as `CREATE
TRIGGER`) are beyond it and you'll have to use a Python delta file.
### Python delta files
For more flexibility, a delta file can take the form of a python module. These should
be named `*.py`. Note that database-engine-specific modules are not supported here
instead you can write `if isinstance(database_engine, PostgresEngine)` or similar.
A Python delta module should define either or both of the following functions:
```python
import synapse.config.homeserver
import synapse.storage.engines
import synapse.storage.types
def run_create(
cur: synapse.storage.types.Cursor,
database_engine: synapse.storage.engines.BaseDatabaseEngine,
) -> None:
"""Called whenever an existing or new database is to be upgraded"""
...
def run_upgrade(
cur: synapse.storage.types.Cursor,
database_engine: synapse.storage.engines.BaseDatabaseEngine,
config: synapse.config.homeserver.HomeServerConfig,
) -> None:
"""Called whenever an existing database is to be upgraded."""
...
```
## Boolean columns
Boolean columns require special treatment, since SQLite treats booleans the

View File

@ -499,9 +499,12 @@ def _upgrade_existing_database(
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module) # type: ignore
logger.info("Running script %s", relative_path)
module.run_create(cur, database_engine) # type: ignore
if not is_empty:
if hasattr(module, "run_create"):
logger.info("Running %s:run_create", relative_path)
module.run_create(cur, database_engine) # type: ignore
if not is_empty and hasattr(module, "run_upgrade"):
logger.info("Running %s:run_upgrade", relative_path)
module.run_upgrade(cur, database_engine, config=config) # type: ignore
elif ext == ".pyc" or file_name == "__pycache__":
# Sometimes .pyc files turn up anyway even though we've