Add getMultiWithErrors #17
Conversation
|
|
||
| const usedServerKeys = Object.keys(serverKeytoLookupKeys); | ||
| const errors: GetMultiError<Keys>[] = []; | ||
| const results = await Promise.all( |
There was a problem hiding this comment.
in case of error, will results contains undefined for the keys or missing?
There was a problem hiding this comment.
results will be an array of partial results and undefineds, so if we had 3 keys foo, bar, and baz on different servers and the server for bar failed, we'd get:
[{foo: {value: "1", extras: "", cas: "token"}},
undefined,
{baz: {value: "2", extras: "", cas: "token"}]
This would be unpacked/coalesced by Object.assign({}, ...results) to yield
{
foo: {value: "1", extras: "", cas: "token"},
baz: {value: "2", extras: "", cas: "token"}
}
from looking at the code/types, I think if bar were a cache miss instead of a failure we'd get
[{foo: {value: "1", extras: "", cas: "token"}},
{},
{baz: {value: "2", extras: "", cas: "token"}]
which would yield the same result
but I would need to verify and test
|
@leon-root added tests and updated branch. Let me know if it looks good on your end. |
Tested on my side and works fine.
|
if a single cache node involved in a
getMulticall fails/times out, the whole batch errors out. The alternative is to treat the failure as a cache miss and return the partial result. We want information on the failures so that we can monitor, react, and/or circuit break on our end.This is untested. If we want to go in this direction, we can add tests and clean up (it's also possible to implement getMulti in terms of getMultiWithErrors to remove the code duplication)