refactor: allow coercion inputs to work with async pipe #17640
Conversation
googlebot
added
the
cla: yes
devversion
force-pushed
the
refactor/coercion-inputs-accept-nuÃll-undefined
branch
from
e4c81ca to
711350d
Compare
|
|
||
| if (node.type.kind === ts.SyntaxKind.AnyKeyword) { | ||
| // if the type is "any", then it can be "null" and "undefined" too. | ||
| return; |
There was a problem hiding this comment.
Do we want to allow it to be any? Since it's something we control it might make sense to be a bit more strict.
There was a problem hiding this comment.
Some coercion members are set to any. i.e. the value input for mat-input's because otherwise value would only accept string while datepicker input accepts a date too.
So any is a special case that we allow.
We need to accept `null` and `undefined` for coercion members to make them work with the `async` pipe. The overall problem remains for other non-coercion inputs, but since the majority of inputs are coerced, we only make the change to these. The overall problem applies to _all_ inputs, but it's not clear yet how this issue can be solved for all inputs. This involves discussion with the framework team. More information: https://hackmd.io/@devversion/rkKOD8ZjB
devversion
force-pushed
the
refactor/coercion-inputs-accept-nuÃll-undefined
branch
from
711350d to
e6f3619
Compare
devversion
requested review from
andrewseguin,
jelbourn,
mmalerba and
a team
as code owners
devversion
added
the
P2
|
We could consider adding type aliases so that we don't repeat |
|
@mmalerba Sounds like a good idea to me. We can do that as a follow-up if we all think it's reasonable. I'd be happy to switch to that. |
|
Follow-up PR would be good |
devversion
added
P1
|
This issue has been automatically locked due to inactivity. Read more about our automatic conversation locking policy. This action has been performed automatically by a bot. |
We need to accept
nullandundefinedfor coercion membersto make them work with the
asyncpipe. The overall problemremains for other non-coercion inputs, but since the majority
of inputs are coerced, we only make the change to these.
The overall problem applies to all inputs, but it's not clear
yet how this issue can be solved for all inputs. This involves
discussion with the framework team.
More information: https://hackmd.io/@devversion/rkKOD8ZjB