trailing comma

[Banyc]

It is also a common case

[tmccombs]

This would be pretty difficult to do with the current implementation.

The current implementation acts as a filter for a Read implementation and modifies the buffer in place.

For comments this works fine, since it can just replace the comments with spaces.

However for a trailing comma, it can't actually know if a comma is trailing until it encounters a "]" or "}". And that might not happen until the next block is read.

I think it could be done, by allocating a separate Vec to store results in across multiple reads. Or possibly it could use a fixed-size buffer if I also changed it to remove whitespace and comments instead of replacing with whitespace.

Either way, that adds a fair amount of complexity, and could hurt performance even if removing trailing commas isn't desired.

It might be cleaner to have a separate filter to remove trailing commas, either as part of this crate or a separate crate.

Or I suppose do what https://github.com/parcel-bundler/parcel/pull/9032/files#diff-194e737b9e37f7950c7d86d1508d15d18f158db5d9ec44084f2ece3dc6eac7d4 does and have a separate API that only works on a single String (or &[u8]) that contains the entire object, so it doesn't have to worry about handling individual blocks of data separately.