Support validating a CSV file against a simple JSON schema

[ldodds]

Allow a CSV file to be validated against a single JSON schema that describes the structure of the CSV.

Initially we should base this on JSON Table Schema, with some of the validation extension described here

There are several parts to this:

• Generic field validation -- presence, length, pattern
• Type mapping -- mapping to fixed set of datatypes (integer, float, URI, boolean), with errors if the values don't map
• Date formatting mappings -- xsd:date, xsd:dateTime, xsd:time, xsd:year, others?
• Type specific validation, e.g. minimum and maximum for date and number ranges

For the type mapping we can support a type key in the field definition which is an XML Schema URI. Initially we should aim to support integers, floats, URIs, date, date-time. To parse dates we can read a strftime format in a datePattern key.

[ldodds]

Other items to support:

• support a unique constraint to check whether values in a column should be unique
• add error/warnings if there are fewer columns in the data, than the schema
• add error/warnings if there are more columns in the data, than the schema
• if there are fewer columns in the data than in the schema, ensure we apply the validations from the missing columns, treating the column as empty. (Currently the code just skips the columns)
• if there's a header in the data then check column names match the schema

[JeniT]

Are we supporting optional columns?

[ldodds]

If something is not required and has no min length it should pass