Various changes

[wezm]

I made some fairly bold changes to the code. I will understand if you don't want to merge them but figured I'd raise the PR for you to consider. The commits are pretty self-contained and do the following:

• Drop the thiserror dependency in favour of manually implementing the std::error::Error impl.
• Drop the byteorder dependency in favour of the functions in std.
• Fix a warning in some example code.
• Avoid unwrapping in the send macro:
  • This seemed like a bit of a footgun since the unwrap was hidden.
  • I used this change to potentially avoid re-panicking inside the panic handler.
• Drop the write_output method since send_message was identical.

I'd also like to run rustfmt on the code, if you're ok with that I'll do it in a separate PR.

[neon64]

Hey! As you can probably tell I haven't been super-actively maintaining this library, but I'm very happy to accept your contributions. These changes mostly sound reasonable! The redundant byteorder dep is interesting - I think I must have originally written this before the equivalent stdlib functions were added.

The one change I'm not immediately sold on is dropping the thiserror dependency. My understanding is that thiserror reduces some of the boilerplate of writing the std::error::Error and Display impls, but isn't visible at all in the public API. Admittedly, this library is so small we're only saving ~20 lines of code. Is there a particular advantage you see in avoiding the thiserror dependency?

[wezm]

Admittedly, this library is so small we're only saving ~20 lines of code. Is there a particular advantage you see in avoiding the thiserror dependency?

I think fewer dependencies is generally better (when it makes sense) as there's fewer things outside of your control. Nothing against thiserror, just didn't seem worth pulling in an extra dependency (and its dependencies) to save 20 lines as you say.

thiserror v1.0.31
└── thiserror-impl v1.0.31 (proc-macro)
    ├── proc-macro2 v1.0.39
    │   └── unicode-ident v1.0.1
    ├── quote v1.0.18
    │   └── proc-macro2 v1.0.39 (*)
    └── syn v1.0.96
        ├── proc-macro2 v1.0.39 (*)
        ├── quote v1.0.18 (*)
        └── unicode-ident v1.0.1

It might also help this crate compiler faster as there's no proc macros involved now—and everyone loves faster Rust compile times :)

[neon64]

Thanks. Indeed that does look quite heavy with all the recursive deps. I'm curious how many of those would already be included your average Rust project (e.g.: other crates like serde also pull in syn, quote, etc...). I suspect different versions could pose a problem (i.e.: multiple separate copies of syn because different projects request different versions).

I think with that in mind I'll just merge this PR.