feat(#1381): Add a way to specify "inject-only" with @JacksonInject #5175
Conversation
…lization discard the injected value in favor of input (if any)
| @@ -1,4 +1,4 @@ | |||
| package com.fasterxml.jackson.databind.tofix; | |||
| package com.fasterxml.jackson.databind.deser.inject; | |||
There was a problem hiding this comment.
moved since this should not fail anymore
| } | ||
|
|
||
| @Test | ||
| @DisplayName("input YES, injectable YES, useInput DEFAULT|FALSE => injected") |
There was a problem hiding this comment.
According to useInput's javadocs:
Default is
OptBoolean.DEFAULT, which translates toOptBoolean.TRUE: this is
for backwards compatibility (2.8 and earlier always allow binding input value).
This combination: input YES, injectable YES, useInput DEFAULT should actually behave as if we had useInput = TRUE, that is dropping the injected value and returning the input.
Nevertheless, since useInput never worked that way, this would mean introducing breaking changes. My suggestion is to keep things as per this test and change the javadoc accordingly:
Default is
OptBoolean.DEFAULT, which translates toOptBoolean.FALSE
| import static org.junit.jupiter.api.Assertions.assertEquals; | ||
| import static org.junit.jupiter.api.Assertions.assertThrows; | ||
|
|
||
| class JacksonInject1381WithOptionalTest extends DatabindTestUtil { |
There was a problem hiding this comment.
This class shows how the behavior changes for all the combinations of the other test class when all injected fields are optional. Do you think we still need a required property in the annotation?
I feel like combining useInput and optional would cover all cases, also because optional should drive the same logic (but reversed) of required. Given this, adding a required property would make things unclear for the users, and the internal logic would be quite messy. Does this make sense?
useInput=TRUE in JacksonInject now make deserialization discard the injected value in favor of input (if any)