bminer/schemer

Lightweight and flexible data encoding and decoding library

schemer's Issues

Copy function comments from `Schema` interface to each implementation

This will fix a majority of golint issues

test coverage for tags

Make a new .GO module called "tag-options_test.go" which contains a test to make sure that Schemer can parse struct tags properly. Have the test include all variations of tags.

cannot decode structures into empty interface

It doesn't appear possible to decode structures into empty interfaces. This might be related to the fact that DefaultGOType() for FixedObjects does not return an appropriate structure. (It currently only returns a blank struct.)

JSON Schema Decoder bugs

• Fix: bits and length fields should be allowed to be null
• Fix: field name should be allowed to be a string or an array of strings (currently only array of strings works)

Document custom schema IDs

There is enough room for 64 custom schema IDs. Probably have 0-31 be well-defined types and 32-63 be custom user-defined types

Replace `DecodeSchema` and `DecodeSchemaJSON` with custom Decoder

Replace DecodeSchema and DecodeSchemaJSON with something like NewSchemaDecoder and NewJSONSchemaDecoder.

I'd like to mirror the JSON decoder API to provide a Buffered() method that returns an io.Reader.

• func (Decoder) Buffered() bufio.Reader
• func (Decoder) Decode() (Schema, error)

This makes it clearer that the io.Reader is a buffered reader, and it exposes the buffered but unconsumed contents of the Reader.

better test coverage for decoding binary data

It might be a good idea to include some binary schemas and encoded data to use during unit testing. The binary data should be generated by encoding data with Schemer.

Then during unit tests, where changes could potentially be made to encoding / decoding logic, it is important to check the behavior of both the encoding and decoding logic.

On the encoding side, the tests should compare binary data generated with know good binary data.

On the decoding side, the test should make sure the decoder can always decode the known binary data to known values.

Add schemer version to schema

Adding the version number of schemer itself to the encoded schemas will allow even more compatibility as schemer evolves

ParseStructTag should use regular expressions

ParseStructTag() in module tag-options currently uses an ad hoc method to parse the schemer tags on structures. A better approach is to utilize regular expressions to parse the tags.

The syntax of the tags should be checked. Currently I believe they are in the format like this:

tag := (alias)?("," option)*
alias := identifier
"["identifier(","identifier)*"]"
option := "weak", "null", "not null"

An example tag is:
schemer:"readings"

finish example programs

The example programs in the examples directory need finished. Having good examples is important because they illustrate how to use the library, and show why Schemer is useful in a concrete way. Therefore, having them could be helpful to programmers considering using Schemer.

Reduce memory allocations on decode

Perhaps built-in Schema implementations can re-use buffers internally to reduce memory allocations for each Decode call?

Schemas should all implement `encoding.BinaryMarshaler`

Implement more custom schema types

• Duration
• Regular expression?
• IPv6?

Consider new folder / package layout

• schemer
  • BoolSchema
  • ...
  • ObjectField
  • Schema
  • SchemaOptions
  • Register
• schemer/decoder
  • decoder.NewSchemer
  • decoder.NewJSON
  • decoder.SchemaOf
  • decoder.SchemaOfType
• schemer/util
  • PreEncode
  • PreDecode
  • ReadUvarint
  • WriteUvarint
  • StructTag

Add link to benchmarks repo

javascript interface

A major use case for this library is probably to communicate from a GO server to Browser UI. Therefore, having a java script implementation seems important and worth implementing.