Hume is a service for performing empirical data validation tests against previously seen and "known good" values/distributions/etc.

Metrics are configurable functions that can be applied to files of fields to output statsitics that can then be persisted and/or evaluated.

Metric names are globally-unique, so they can be easily referenced in a data source configuration, and are associated with a spcific metric type. Some metric types may require configuration while others do not. For example, a metric to calculate the number of fields of a data set (important for file processing):

{
	"name":"Field Count",
	"type":"FieldCount"
}

But a metric for, say, checking the date format of a field would look like:

{
	"name":"Date Format: YYYY-MM-DD",
	"type":"DateFormat",
	"format":"%Y-%m-%d"
}

Sources define an input source, such as a file or a sql query, their expected fields, any metrics and associated evaluators.

{
	"name":"Example File",
	"type":"File",
	"delimiter":"|",
	"fields":[
		"id",
		"value"
	],
	"has_header":True,
	"tests":[
		{
			"metric":"Field Count",
			"persist":True,
			"evaulate":[
				{
					"evaluator":"SingleValueMatch",
					"value":2
					"threshhold":1.00
				}
			]
		}
	]
}