Parsing CBOR data with numeric keys

Help & Support
1

For a project, I need to work a lot with CBOR data. This library has been great so far: GitHub - albertorestifo/crystal-cbor: [mirror] CBOR encoding and decoding in Crystal. Basically, it works like YAML and JSON serializer.

But I often run into problems when parsing objects with numeric keys. Here’s an example from the diagnostic:

{1: -8, 4: h'5d9bd15d2b241d42ae767c653e4aaa2ea9e3b2d2da66f0e09d13b9691d12f033', "address": h'01e4740b2690791de5ebad00ea555d8c6c31318ea980999eb7b13a24b955c29e1abea4593c52660b90b6fa20b3e5a42b105666522a3f14af29'}

Ideally, I would like to do something like this:

alias Header = NamedTuple(1: Int32, 4: Bytes, address: Bytes)
header = Header.from_cbor(cbor_string)

But a NamedTuple can’t have numeric keys. So I get around it by parsing the object into a Hash and wrapping the result into a struct to access the properties. This approach works with simple objects like the one above, but it becomes tedious with complex objects involving unions.

So I was wondering, what’s the reason why a NamedTuple can’t have numeric keys? And is there another solution I missed?

My last resort would be forking the library and making @[CBOR::Field(key: 1)] work so I can map the whole object to a Struct. But maybe I’m missing something.

2

In the meantime, I created a PR on the CBOR library to allow objects with numeric keys to be parsed to structs with CBOR::Field annotations:

struct Headers
  include CBOR::Serializable

  @[CBOR::Field(key: 1)]
  getter one : Int32
  @[CBOR::Field(key: 4)]
  getter four : Bytes
  getter address : Bytes
end

Although, I’m curious to know why a NamedTuple can’t have numeric keys.

3

NamedTuple mainly exists to support named argument passing, and named arguments can’t have number literals as their names. You have to turn numbers into strings similarly in that case:

def foo(*, 1 x) # Error: unexpected token: "1"
end

foo(1: "a") # Error: expecting token ')', not ':'

def foo(*, "1" x) # okay
end

foo("1": "a") # okay
2 Likes
4

Right, that makes a lot of sense. Thanks for explaining!