As I don't want to change the shape of the DataTransportAbstraction immediately after we released it for the first time, this one would also need to go into 0.9.0.
This switches from two very separate "request started" and "request finished" events to a stream of events with a shared identifier between queries.
This has the benefit that in the future we can easily track which events are still "unhandled", e.g. for keeping the connection open for longer.
Also, for the first time we detect & communicate errors that happened in SSR to the browser, to restart queries there immediately.
As a positive side-effect of having a "query identifier", the query document is sent only once, not twice over. That already plays a bit into #210.
(Technically, it's sent one more time per hook, so we're down from three times to two).
phryneas
commented
1 month ago 
relativeci[bot]
commented
1 month ago 
As I don't want to change the shape of the DataTransportAbstraction immediately after we released it for the first time, this one would also need to go into 0.9.0.
This switches from two very separate "request started" and "request finished" events to a stream of events with a shared identifier between queries.
This has the benefit that in the future we can easily track which events are still "unhandled", e.g. for keeping the connection open for longer.
Also, for the first time we detect & communicate errors that happened in SSR to the browser, to restart queries there immediately.
As a positive side-effect of having a "query identifier", the query document is sent only once, not twice over. That already plays a bit into #210. (Technically, it's sent one more time per hook, so we're down from three times to two).