You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It seems that the SharedInformerFactory is designed for the openapi client, with the CallGenerator returning an okhttp3.Call.
ProtoClient doesn't seem designed in way to hook in the same way, at least at the moment, as it doesn't return okhttp3.Call. Rather, it invokes a request synchronously which wouldn't work with watches.
I'm not sure would be natural for it to return io.kubernetes.client.openapi.ApiException from generate, but that could be fine since ProtoClient depends on the openapi generated ApiClient anyway.
I think it could be something we can rejig to make it happen, by changing SharedInformerFactory or extracting some base class from it. This wouldn't unhook the gson dependency as still ApiClient pins it, but it could be possible to still use the proto infrastructure and drive informers more efficiently.
The text was updated successfully, but these errors were encountered:
Happy to have work done here. As expressed elsewhere, ProtoClient was hacked together a long time ago (before SharedInformerFactory, fwiw) and then has mostly lain dormant.
It seems that the
SharedInformerFactory
is designed for the openapi client, with theCallGenerator
returning anokhttp3.Call
.ProtoClient
doesn't seem designed in way to hook in the same way, at least at the moment, as it doesn't returnokhttp3.Call
. Rather, it invokes a request synchronously which wouldn't work with watches.I'm not sure would be natural for it to return
io.kubernetes.client.openapi.ApiException
fromgenerate
, but that could be fine sinceProtoClient
depends on the openapi generatedApiClient
anyway.I think it could be something we can rejig to make it happen, by changing
SharedInformerFactory
or extracting some base class from it. This wouldn't unhook the gson dependency as stillApiClient
pins it, but it could be possible to still use the proto infrastructure and drive informers more efficiently.The text was updated successfully, but these errors were encountered: