Binary stream I/O

[fatcerberus]

Latest development build of minisphere adds a DataReader object which can be used to read binary data from files and sockets:
https://github.com/fatcerberus/minisphere/blob/master/docs/miniRT-api.txt#L514-L615

Do we also want to have a DataWriter, or should we leave that out in order to push developers to use JSON instead for new formats?

[FlyingJester]

I think that if you have a binary data reader, it only makes sense to include a binary data writer.

[fatcerberus]

I was thinking along the lines of the TextEncoder/Decoder standard, where a whole bunch of legacy formats are supported for decoding, but only UTF-8 in the other direction. But I guess binary data isn't really "legacy" in any real sense - it's useful to be able to send binary data over a socket, for instance.

[fatcerberus]

One thing I do want to do for the standard binary module is simplify the schema format used by readStruct(). The schema format used in the miniRT module has a lot of fancy features like backreferences, conditional sections, etc. which allow me to read entire Sphere v1 formats in one go (including RMP and all RSS versions, which is quite an accomplishment). That's a bit too much complexity to standardize on, I think.

[FlyingJester]

Perhaps something closer to the format I used in TurboSphere (using the json formats I put under the Spehre-Group name)? It's true it required the script to determine what kind of block came next, but I found that provided a pretty simple and straightforward mechanism to load the Sphere formats at least.