Bikeshedding type identifiers: align with TJSON?

[tarcieri]

ObjectHash presently uses the following single-letter identifiers for its types:

• b: boolean
• d: dict
• f: float
• i: integer
• l: list
• s: set
• u: Unicode string

In TJSON I have slightly different single-letter identifiers which express a slightly larger range of types, including binary data (see also #22), and have names that align with the JSON data model:

• A: array
• O: object
• S: set
• b: boolean
• d: binary data
• f: floating point
• i: signed integer
• s: string
• t: timestamp

Note the use of a capital to denote a non-scalar type.

I think it'd be great to agree on a common set of tags. I began working on an ObjectHash-alike using these identifiers, but I'd definitely prefer to use ObjectHash itself if possible.

If ObjectHash were to adopt the TJSON identifiers outright, it would require the following changes (old => new):

• d => O
• l => A
• s => S
• u => s

This would also free up d for use with binary data, which I think is a particularly nice thing to support (i.e. Unicode strings can be normalized prior to hashing, but binary data is hashed as-is).

[benlaurie]

We should definitely support binary data.

I am planning to start a more production-ready and less dependent on JSON version of objecthash - perhaps leave this change for that?

[tarcieri]

Sure, sounds good