Discussion: API design about hubspot-ruby HOT 4 CLOSED

An issue with wrapping requests automatically in a retry loop is that this may exacerbate any issues and potentially start triggering rate limiting by HubSpot. I think it's better for the user of the gem to decide for each request if they want to retry based on the failure type. If the failure is expressed as a specific exception class, these can easily be retried with existing functionality in things like ActiveJob or handled cleanly with Rails features like rescue_from, or with a simple Ruby begin...end with a retry statement. This gives the user of the gem more flexibility to explicitly retry a failed request.

from hubspot-ruby.

Sticking with the language and community's idiomatic approach would have the smallest learning curve for folks looking to adopt this gem. Any thoughts on (optionally) wrapping requests in a retry if they hit a 4xx or 5xx (but not 429) error code?

from hubspot-ruby.

I think the main reason for using exceptions when working with external APIs is that since it's external you can rely on normal mechanisms for undoing something you've done. For example, with ActiveRecord, you may modify the properties of an instance and change several associated records as part of the change. If something fails, the SQL transaction can be rolled back and everything is undone. With an API that's not possible and it becomes an exceptional case.

I think there is a middle ground we can have here, which follows the Ruby (Rails?) conventions. Having an update() method that returns false on error while also having an update!() method that raises an exception on error.

from hubspot-ruby.

Thank you all for your feedback! I'm going to close this issue as I think we're in agreement on the following:

Design Decisions:

• Avoid leaky abstractions (ex: surfacing HTTParty responses or errors to the gem user)
• Include the Hubspot API error message when a request fails. We want to pass along all insight that's provided by the HubSpot API.
• The library will raise for any non-200 response
• Discuss “handling failure” on the README so readers know what to expect when a request fails and how best to rescue

from hubspot-ruby.