[
https://issues.apache.org/jira/browse/ISIS-804?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14032296#comment-14032296
]
Dan Haywood commented on ISIS-804:
----------------------------------
Oscar said (on dev mailing list):
Mmmm... this is a quite controversial proposal!!!
Nice to be able to discuss about this so disruptive implementation :-))
In my opinion, this can lead to quite misinterpretations. The usual
expectations for anybody coming into Isis (or migrating his current JDO - JPA
in the near future - into Isis) there would be for all properties with both
getter and setter to be writable.
If we analyze our entities, nearly all properties have getters and setters.
This proposal would force us to create modifyXXX and clearXXX methods for all
them, invalidating the model.
Perhaps this more restrictive model could be optional, but my opinion would be
to keep it as writeable by default (as it's assumed by nearly all other
platforms or frameworks).
Just my 2 cents...
Also, to implement modifyXXX and clearXX would not imply that tests are
available for all them, so the responsibility to properly testing them would
not be delegated to the platform. So the programmer would not avoid to write to
all tests.
Sorry for not agreeing (by now ..) :-))
> Make (entity) properties read-only by default.
> ----------------------------------------------
>
> Key: ISIS-804
> URL: https://issues.apache.org/jira/browse/ISIS-804
> Project: Isis
> Issue Type: Improvement
> Components: Core
> Affects Versions: core-1.5.0
> Reporter: Dan Haywood
> Assignee: Dan Haywood
> Fix For: core-2.0.0
>
>
> Currently properties are read/write by default; the programmer has to
> annotate with @Disabled (or equivalent) to make read-only.
> While this "subtractive programming" approach is nice for demo's, the truth
> is that with larger applications, if not rigorously and completely tested,
> the end-user may get the opportunity to change something that they ought not;
> potentially corrupting data.
> It would be better (and safer) for properties to be read-only by default.
> The presence of the modifyXxx(...)/clearXxx() supporting methods would then
> indicate that they are read-write.
> Another possible benefit is that - if we implement ISIS-273 (to read from
> fields) then the amount of boilerplate would substantially be reduced if a
> tool like Lombok was used. That is, read-only properties would consist
> solely of a private field; read-write properties would be the field plus the
> modify/clear.
> It might also make sense for the clearXxx() method to be optional; that is,
> to allow modifyXxx(..) to be called with null.
> NB: for view models, properties could (perhaps) remain as read-write; there
> are no side-effects.
> This change requires discussion on the mailing list.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
