Greg Marr

I would say that the equality of the numerical form is more important than equality of the string form for floathing point values.

@barcode @nlohmann I think this just needs some documentation.

@barcode This needs to be updated for the new type safety namespace introduced in 3.11.0 through 3.11.2.

What is the benefit of having the version part a separate namespace that may or may not be used instead of just a part of the name that may or...

@imwhocodes Do you have a test case that demonstrates this error?

> > @imwhocodes Do you have a test case that demonstrates this error? > > See quote below. I can share my test case which fails to reproduce the issue...

I've looked at the discussion, and thought a lot about this, and done some researching, and I'm still not sure if it's possible to "fix" this use case at this...

@falbrechtskirchinger >Because your implementation reads on the advance, you'll inevitably run into problems. You're violating the expected semantics and complain that the result isn't what you want. That's what `std::istream_iterator`...

>The real scenario problem it is not about knowing if the stream is closed, but it is that on the current input_adapter implementation one char is always read ahead, this...

@falbrechtskirchinger I just went looking through the discussions in this PR for it and couldn't find it. Can you remind me again which case is broken with this change?