Pattern for more detailed types

[rufuspollock]

Many platforms/tools including SQL have a more detailed set of types (e.g. long integer) than Table Schema. This may be relevant for serialization and deserialization reliably. We have intentionally not included lots of types into Table Schema as it quickly gets complex (and you still don't support everyone).

What is proposed here is some standard pattern for adding additional information on types to Table Schema, esp a model of subtypes e.g. that something is a long integer cf #277 .

Tasks

• Review existing important tools and their type lists e.g. SQL, Spark, BigQuery, ...
• Develop the pattern

[khughitt]

Adding this as a reference for any future consideration of explicit types to support:

[pandas/web/pandas/pdeps/0013-logical-type-system.md at 8a246978e9812a2448f43d0df24a82d51e850d53 · pandas-dev/pandas](PDEP-13: The pandas Logical Type System)

The focus of the proposal is on handling of different technical implementations of types (e.g. string vs. string[pyarrow]) while providing a simple/unified interface to users, but it also can be useful simple when trying to think carefully about all of the types one might want to explicitly handle in Frictionless.

Logical types intended to be supported in the proposal:

• Signed Integer
• Unsigned Integer
• Floating Point
• Fixed Point
• Boolean
• Date
• Datetime
• Duration
• Interval
• Period
• Binary
• String
• Dictionary
• List
• Struct