Closed istathar closed 2 years ago
The original implementation based on resource-pool required hacks to accommodate to the requirements of resource-pool. It had something to do with exceptions but I don't remember precisely. There were several issues caused by those hacks, which you can see in the closed issues of this tracker. At some point I realised that it would be easier to just implement the pool directly. As you can see in the source code, the whole implementation now is like 80 lines of code.
@istathar FYI, as I still need to rely on resource-pool
in my projects, I released hasql-resource-pool fork of v0.5.2.2 that was the last one using resource-pool
internally. I intend making a switch to the recent v0.3.x release maintained by scrive (I only need to port the missing stats thing). So if you're interested in a resource-pool
based implementation, feel free to contribute.
@nikita-volkov given resource-pool
is redundant now, and your previous suggestion to use it for things that require dynamic settings acquisition, would you be open to consider IO-based connection settings API in the new hasql-pool
implementation? The current version still doesn't allow me to use it with AWS RDS token-based authentication, and this is a predominantly mandatory setup in corporate environments I've been working with.
@avanov I've answered to your question in the according issue linked above. Thanks for bringing this up!
Hey Nikita. I hesitated to ask this because I can see you've been working really hard to refine hasql-pool. I'm sure you're aware (but maybe you're not) that resource-pool was adopted and resuscitated by the community. Notably
with the result that a 0.3.0 has been released to Hackage with a number of performance and correctness fixes.
I assume you want nothing to do with it at this point, but we're using resource-pool elsewhere in our system and so noticed that suddenly hasql-pool wasn't anymore. Which is entirely fine but hey, thought I'd mention. Go ahead and close this
:)
. All the best.