rego.compile: Generate data filters

The rego.compile() built-in function allows you to programatically generate data filters and column masks from Rego rules.

The function partially evaluates a filter rule, checking that the rule is valid, and consequently compiling it into the desired output syntax, such as SQL strings or UCAST. It also evaluates any paired column masking rule, and will include the results in the output.

warning

rego.compile() is intended for use in testing (see here for examples), and it is therefore only available in eopa test and eopa eval.

rego.compile

Example Usage

rego.compile({
  "query": "data.filters.include",
  "target": "sql+postgresql",
}) with input.username as "jane"

Parameters

Parameter	Type	Required (default)	Description
query	String	Yes	The query to partially evaluate, e.g. data.filters.include.
target	String	Yes. One of sql+mysql, sql+postgresql, sql+sqlserver	The target for query translation.
mappings	Object	No ({})	Table and column name mappings for query translation (see below).
raise_error	Bool	No (true)	If set to false, translation-related errors are returned as an object with key error.

Mappings

Mappings allow you to rename fields during query compilation.

Below we will be reusing an example featuring a fruits table with a column name

When writing rules without mappings, input.<table_name>.<col_name> expressions are directly translated into SQL clauses.

A rego expression of input.fruits.name = "banana" would generate a SQL filter of

WHERE fruits.name = E'banana'

Renaming tables and columns

Using the same example expression as above, input.fruits.name could be adjusted to yield a constraint like

WHERE FRUIT.NAME = E'banana'

by passing a mapping object like this:

{
  "postgresql": {
    "fruits": {
      "$self": "FRUIT",
      "name": "NAME"
    }
  }
}

Renaming short references to tables

Furthermore, mappings enable the translation of short references, so an expression like input.name = "banana" could become a query like

WHERE fruits.name = E'banana'

by passing a mapping object like this:

{
  "postgresql": {
    "name": {
      "$table": "fruits",
    }
  }
}

Renaming short references to columns

If need be, we can also translate the column name of a short reference, so an expression like input.label = "banana" could become a query like:

WHERE fruits.name = E'banana'

by passing a mapping object like this:

{
  "postgresql": {
    "label": {
      "$table": "fruits"
    }
    "fruits": {
      "label": "name"
    }
  }
}