Removed usage of getters for api endpoints #19876
Conversation
ref ENG-761 ref https://linear.app/tryghost/issue/ENG-761 By removing getters we ensure that each endpoint is only instantiated once. This _may_ have an adverse effect on boot time, so we're going to keep an eye on that, and make any necessary adjustments to the implementation.
|
Out of curiosity, is there any memory overhead with doing this? i.e. now that the |
|
@vikaspotluri123 There is an initial memory overhead yes, but as soon as the API is accessed, we're back to the same footprint |
|
Right, but I'm wondering if/how many APIs are rarely/never instantiated during a boot lifetime. E.g. webhook management might not be done often. (With the asterisk that I'm not sure how the API is accessed within the core, just thinking from an Admin/Public API aspect) |
|
The way the API is mounted means that - once any Admin API endpoint is mounted - every Admin API endpoint file is loaded - same for the Content API - so it's not granular on a endpoint by endpoint basis |
|
Closing for now, in favour of #19880 We have a lot of circular dependencies which are hidden by the lazy-loading of files here - and cleaning it up should be part of a larger task than optimising controller instantiation. |
allouis
deleted the
egg-eng-761-api-pipeline-is-cpu-intensive-and-created-on-the-fly
branch
ref ENG-761
ref https://linear.app/tryghost/issue/ENG-761
By removing getters we ensure that each endpoint is only instantiated once. This may have an adverse effect on boot time, so we're going to keep an eye on that, and make any necessary adjustments to the implementation.