relax taggedalgebraic dependency

[John-Colvin]

it's a real pain otherwise due to dependency clashes.

[joseph-wakeling-frequenz]

Hah, is it that simple? I thought there had been some breaking changes which needed more intrusive version-specific changes.

[John-Colvin]

Well, I don't really know, but the tests pass ¯\(ツ)/¯

Hoping that someone who does know a bit better chooses whether to merge this or not.

[s-ludwig]

Should be fine, the breaking changes should be limited to rarely encountered corner cases. BTW, it may even make sense to require 0.11.x at some point and make use of TaggedUnion and This, which would make the code a bit prettier.

[John-Colvin]

Of course, to really get the benefit of this change, a new release here would be necessary.

[John-Colvin]

ping, still need a new release to get people to be able to use this

[joseph-wakeling-frequenz]

@John-Colvin

Well, I don't really know, but the tests pass ¯(ツ)/¯

Apologies for the late feedback -- I think the issues I was seeing may have been about compiler version rather than taggedalgebraic version. I'll submit a CI PR shortly that may illustrate the issue.

[joseph-wakeling-frequenz]

Oh wait, it's worse than that. CI is still testing with taggedalgebraic v0.10.9, because of the fixed dub.selections.json. So it's never actually validating that v0.11.x works.

Going by my own local testing we have a double bind here:

• taggedalgebraic v0.11.x works with the compilers used in current CI

• it will fail with DMD >= 2.086

That's what I saw in my own local testing when I first filed #44, which is why I didn't follow up quickly at the time.

[joseph-wakeling-frequenz]

In short, please do NOT create a new release until these issues are resolved (I've written up details in #44). Otherwise there will be a lot of unhappy users of std_data_json ;-)