Add docs for using Psycopg 3 connection pooling #12540
Conversation
…ssion sqlalchemy#12522) Signed-off-by: chrispy <chrispy@synopsys.com>
| downsizing for unused connections), connection health pre-checks, and support | ||
| for both synchronous and asynchronous code environments. | ||
|
|
||
| To take advantage of ``psycopg``'s pool, |
There was a problem hiding this comment.
I'm not a fan of this section since we have the example just below, but let's wait mike's opinion on this
There was a problem hiding this comment.
@CaselIT - this is patterned after the Oracle docs here. I am happy to make any changes requested.
There was a problem hiding this comment.
I just meant the list part. Let's wait mike opinion on this
| the Psycopg 3 documentation for ``psycopg_pool.ConnectionPool``, | ||
| which provides more information on available configuration parameters, | ||
| logging, and more. | ||
|
|
There was a problem hiding this comment.
since the async version is not 100% the same it may make sense to provide the example example also in async
There was a problem hiding this comment.
@CaselIT - I added an incremental example for async code. If you prefer a full example, I can do that too.
Signed-off-by: chrispy <chrispy@synopsys.com>
| @@ -117,6 +117,20 @@ def close(self): | |||
| creator=mypool.getconn, # Use Psycopg 3 connection pool to obtain connections | |||
| ) | |||
|
|
|||
| Usage of ``psycopg_pool.AsyncConnectionPool`` is similar, but it requires the | |||
| following ``psycopg.AsyncConnection`` subclass instead: | |||
There was a problem hiding this comment.
I think it make sense to mention also the use of the AsyncConnectionPool and async_creator
There was a problem hiding this comment.
@CaselIT - good feedback. There were indeed enough differences that a more complete example was needed. It's there now.
Signed-off-by: chrispy <chrispy@synopsys.com>
|
@dvarrazzo - when I use the asynchronous code structure in this pull request (complete runnable example here) with logging enabled, I get the following warning: How should the code change to resolve this? It is not clear to me. |
|
As I have re-opened psycopg/psycopg#1046 maybe we want to wait for which shape it takes before merging this? |
You must open the pool in closed state and use async def main():
mypool = psycopg_pool.AsyncConnectionPool(
connection_class=MyAsyncConnection,
conninfo=db_url.replace("+psycopg", ""),
open=False,
)
await mypool.open()This is documented here but maybe it's unclear that |
|
@dvarrazzo - thanks! I updated the asynchronous example there to open the pool outside the constructor. |
Signed-off-by: chrispy <chrispy@synopsys.com>
works for me! Or we can even have two example depending on the pool version. |
|
@CaselIT, @dvarrazzo - I am fine with waiting for the psycogp_pool 3.3.0 release, then releasing this documentation update with the more user-friendly example (noting the required psycogp_pool version). |
I doubt that psycopg pool 3.3 will bring any difference that anyone would prefer to use the older version. I think that, because asking to subclass the connection is pretty awkward, I think you can limit the documentation to the easier 3.3. If someone needs to use an older version, I am adding the subclassing example to our docs too, so I don't think you need to be responsible of presenting it. I will link you the updated docs for review as soon as the branch I am preparing is tested. |
|
@dvarrazzo - for completeness, I will also keep both the < 3.3 and >= 3.3 asynchronous examples here as a reference. |
makes sense. At worst we could link to that discussion comment from the documentation simply stating "In older version of psycopg see the example at [...]" |
Updates documentation to describe how to use Psycopg 3 connection pooling in SQLAlchemy. This is a follow-up to discussion #12522.
Description
The Postgres >
psycopgdialect documentation is updated. I used the Oracle connection pool docs as a reference to ensure consistent documentation style.Checklist