John Keiser

Related: #1382.

What's your read on having this for 1.0 now? It will likely take some ingenuity to do right ...

We are so close to being great for embedded systems, that this seems like a low cost way to let a whole segment of users in. Good stretch for 1.0.

@h5rdly most particularly, you either need to #include "simdjson.cpp" or compile it into your executable on the command line.

This is closer to what I'd expect. Those are new errors we haven't seen before!

@h5rdly note that if you specify c++17, it works. Does 2a have a bunch of compatibility breaks?

Yeah, this is a C++20 thing. They made == reversible and introduced rules to prevent ambiguity therein, which is fine; but it seems like it's interacting badly with templates here?...

Wow! We really really worry about c++17 errors, so I'd like to know more. It looks like system-specific errors, honestly (which is weird because I'd expect c++20 to fail the...

Oh! So the c++17 errors are due to running clang on Windows. That indeed makes sense, given that we test for Windows using _MSC_VER. Can you open a separate issue...

One thing that might be useful is a `max_size()` that tells you the maximum possible number of elements in the array. This would allow bindings that want to create arrays...