freitagbr / crystal-uvarint Goto Github PK

In order to avoid namespace conflicts, the struct UVarInt definition should be placed inside a module.

For many use cases, a UVarInt will need to be created from an existing hex representation, usually as part of a string or from IO. In these cases, the representation might be included as part of a larger string, like in a multihash, and it may not be known how many bytes represent the varint. A constructor should be created to account for these use cases.

It would be really great to have a way to build a varint from an array of bytes where we may not know how long the Varint itself is. Imagine you have:

[1000 0001 , 1000 0010 , 0000 0001 , 1000 0001 , 0000 0001, 1000 0010]

It would be nice to have a constructor for Varints like Go's implementation:

value1, numBytesRead = Varint(bytes)
puts value1 # => 16641 (if my math is right)
puts numBytesRead # => 3

# Now we can get the second Varint from the array!
newBytes = getArrayStartingAt(bytes, numBytesRead) # => [1000 0001, 0000 0001, 1000 0010]

value2, numBytesRead = Varint(newBytes)
puts value2 # => 129 (if my math is right)
puts numBytesRead # => 2

lastByteArray = getArrayStartingAt(newBytes, numBytesRead) # => [1000 0010]
value3, numBytesRead = Varint(lastByteArray) # Throws an exception! We can't parse this

As with the final example, if we can't parse a VarInt from the byte array, then we should just throw an exception.

This could either be implemented as a class constructor or a module-level method that returns a Varint.

The README needs to be updated with the new constructor method.