guava collections issue

[Martin Charlesworth]

Hi Graham, I was going to reply to your long message regarding my pull request, but I can't find it! Anyway I think the main point was that we can't use reflection in ConfigurationBuilder due to the CLI version, you had another idea using the bytecode? I've not done that before so that'd be good to try. What would be the first thing to do/look at?

martin

[Graham Allan]

Hi Martin,

The comment was on the now closed pull request:

https://github.com/MutabilityDetector/MutabilityDetector/pull/50#issuecomment-26224887

I think we can get away with trivial validations of user input (not null, not blank, etc). Failing fast would be a nice-to-have, I expect we could create another issue out of validating the configuration.

I'm currently more concerned about the Guava collection type being part of the public interface, as we can't release with that, and a replacement will have to be found. I'd suggest looking at that first. 

That okay?

Cheers,

Graham

On 16 October 2013 20:19, Martin Charlesworth <martincha...@gmail.com> wrote:

Hi Graham, I was going to reply to your long message regarding my pull request, but I can't find it! Anyway I think the main point was that we can't use reflection in ConfigurationBuilder due to the CLI version, you had another idea using the bytecode? I've not done that before so that'd be good to try. What would be the first thing to do/look at?

martin

--
You received this message because you are subscribed to the Google Groups "mutability-detector" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mutability-dete...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Martin Charlesworth]

On Thursday, 17 October 2013 18:20:56 UTC+1, Graham Allan wrote:

Hi Martin,

The comment was on the now closed pull request:

https://github.com/MutabilityDetector/MutabilityDetector/pull/50#issuecomment-26224887

Thanks!

 

I think we can get away with trivial validations of user input (not null, not blank, etc). Failing fast would be a nice-to-have, I expect we could create another issue out of validating the configuration.

OK, I'll do that. I actually only used reflection because I thought I needed the method object to get the type description string, but now I've looked at Type interface I see there are lots of other ways.

 

I'm currently more concerned about the Guava collection type being part of the public interface, as we can't release with that, and a replacement will have to be found. I'd suggest looking at that first. 

No problem. So you're suggesting we put out own wrapper class around the ImmutableMultimap?

I'm curious about CollectionTypeWrappedInUnmodifiableIdiomChecker.FIELD_TYPE_TO_COPY_METHODS - do we actually need it to be a map? ie do we need the type of field being assigned, or could it just be a List<CopyMethod>?

Cheers,

Martin

 

That okay?

Cheers,

Graham

On 16 October 2013 20:19, Martin Charlesworth <martincha...@gmail.com> wrote:

Hi Graham, I was going to reply to your long message regarding my pull request, but I can't find it! Anyway I think the main point was that we can't use reflection in ConfigurationBuilder due to the CLI version, you had another idea using the bytecode? I've not done that before so that'd be good to try. What would be the first thing to do/look at?

martin

--
You received this message because you are subscribed to the Google Groups "mutability-detector" group.

To unsubscribe from this group and stop receiving emails from it, send an email to mutability-detector+unsub...@googlegroups.com.