DataTypes: "baseType" and "scenario"

[francescoloconte]

As we continue the work to enable Orchestra to support binary (exchange) protocols, these questions have come up:

“baseType” Attribute

The definition of a datatype can specify a “baseType”, as indicated in this paragraph, to inherit its properties. However, it is not clear what properties are inherited since I can see that a DataType only has the following properties:

• Name: N.A. for inheritance
• BaseType: Does it inherit this?
• Pedigree: N.A. for inheritance
• Documentation: N.A. for inheritance
• AppInfo: N.A. for inheritance
• Children “MappedDataTypes”: Does it inherit these? And if it does, why? And what happens if the inheriting DataType specifies its own MappedDataTypes? Are they overridden?

In short, it would seem that the value of inheriting from another DataType is to inherit its BaseType. Is there anything else?

“scenario” Attribute

DataTypes also support scenarios, through the “scenario” attribute. Do we have a use-case for this? The only case I can think of where this is useful, is to crate a new scenario of a DataType to provide it with a different set of MappedDataTypes. Was this the requirement? Is there more to it? What happens to the scenario if the “base” scenario uses a “baseType” to inherit from another one? I think we should clarify it a bit better.

[francescoloconte]

Response from Don Mendelson:

The baseType attribute was carried over from Unified Repository. Not only was it not well defined, the values were often incorrect in my opinion. They were often based on the lexical space of tag-value encoding rather than value space. For example, the base type of UTCTimestamp is listed as String. But in SBE, time values are encoded with numeric values representing a number of time units; there is no semantic relationship to String. In my view, scenario is a better mechanism than baseType and the syntax is consistent with other elements in Orchestra. I recommend removing baseType.

As for the usage of scenarios for datatypes, I can give a few examples:

• The int datatype by itself is unbounded, but it may be useful to specify different ranges for different purposes. The number of entries in a book might be covered by a short integer < 256 while the number of trades in a day might be a long integer in the millions.
• Character data might be in different encodings, e.g. ASCII vs. Unicode UTF-16.
• Price, Qty, and Amt all are supposedly based on float. But monetary amounts should be expressed as exact decimals while true floating point ratios may be expressed with IEEE 754 floating point representation (binary).