Connection Pool Implementation

[asanger]

THIS IS A WIP

Implements connection pool support using https://github.com/langchain-ai/langgraph/pull/2494/files#diff-44b94a2742a89c9e81630c89dd9d0454423c2f3c16f131652d3a68c86bca0104 as a reference.

[tjni]

I'm aligned with this in broad strokes. The one area I'd consider modifying is to by default still not use the pool to keep it the same as upstream's PostgresSQL implementation.

Correct me if I'm wrong, but I think upstream expects that creation of a pool is done outside of the from_conn_string convenience method. But I'm open to adding extra arguments to from_conn_string to make this more convenient: a boolean to decide whether to use a pool or not, and maybe room for extra arguments (such as size, pool recycle, etc.).

What do you think?

[tjni]

Alternatively, we can expose the query parsing logic as a static method for ease of use in constructing the pool separately. I don't have a strong opinion either way.

[asanger]

Honestly, I'm having a hard time groking how the Postgres version is handling pools. I think the Postgres lib allows you to create an AsyncConnection that can contain Pool properties, but with MySQL I think you have to use either a Conn or a Pool which are similar but a pool can't be used directly as a connection.

I think what we ultimately want is as close to a 1:1 of the Postgres implementation as possible because as they make changes it'll make things easier to update. But I know that due to the differences between dependencies, that won't always be possible.

Also, python is definitely not my primary language so I don't have a deep understanding of how MySQL/aiomysql/etc work with py. I'm just trying to figure it all out as I go :)

You are more than welcome to take what I have in my PR and make adjustments as needed, I just need the ability to use connection pools.

[tjni]

I apologize for taking so long on this. This week is busy for me at work, and I realized I need more focused time and brain power to understand how to incorporate those latest changes. (Specifically, I think I want to port those pregel tests over to test the race condition, so I can figure out how or if we need to duplicate the transaction logic.)

I expect to have time this Saturday to dig into this and hopefully get something working based on upstream to unblock you.

[tjni]

I think this has been addressed now, so optimistically closing this PR. Thank you for your help with this!