JS: Handling Infinity values for Float fields

[serchavalos]

Hello!

I noticed that in encoder.js, the method does not allow writeFloat does not accept Infinity as its value

https://github.com/protocolbuffers/protobuf/blob/d9ccd0c0e6bbda9bf4476088eeb46b02d7dcd327/js/binary/encoder.js#L365-L375

/**
 * Writes a single-precision floating point value to the buffer. Numbers
 * requiring more than 32 bits of precision will be truncated.
 * @param {number} value The value to write.
 */
jspb.BinaryEncoder.prototype.writeFloat = function(value) {
  goog.asserts.assert((value >= -jspb.BinaryConstants.FLOAT32_MAX) &&
                      (value <= jspb.BinaryConstants.FLOAT32_MAX));
  jspb.utils.splitFloat32(value);
  this.writeUint32(jspb.utils.split64Low);
};

And its comment seems misleading because it says Numbers requiring more than 32 bits of precision will be truncated. but in reality, it throws an error.

Shouldn't writeFloat remove this restriction or at least this assert?

Thanks in advance,

[dibenede]

Yes, the API is a bit misleading and we should clean this up.