Filter Rules ​

Permissions, validation, and the API's filter parameter all rely on a specific JSON structure to define their rules. This page describes the syntax for creating flat, relational, or complex filter rules.

Syntax ​

• Field — Any valid root field, relational field, or logical operator
• Operator — Any valid filter operator
• Value — Any valid static value, or dynamic variable

{
	<field>: {
		<operator>: <value>
	}
}

{
	<field>: {
		<operator>: <value>
	}
}

Examples ​

{
	"title": {
		"_contains": "Directus"
	}
}

{
	"title": {
		"_contains": "Directus"
	}
}

{
	"owner": {
		"_eq": "$CURRENT_USER"
	}
}

{
	"owner": {
		"_eq": "$CURRENT_USER"
	}
}

{
	"datetime": {
		"_lte": "$NOW"
	}
}

{
	"datetime": {
		"_lte": "$NOW"
	}
}

{
	"category": {
		"_null": true
	}
}

{
	"category": {
		"_null": true
	}
}

Filter Operators ​

The following operator has no Title on the Filter Interface as it is only available in validation permissions:

^[1] Only available on Geometry types.
^[2] JavaScript "flavor" regex. Make sure to escape backslashes.

Relational ​

You can target related values by nesting field names. For example, if you have a relational Many-to-One author field, you can set a rule for the author.name field using the following syntax.

{
	"author": {
		"name": {
			"_eq": "Rijk van Zanten"
		}
	}
}

{
	"author": {
		"name": {
			"_eq": "Rijk van Zanten"
		}
	}
}

When using M2M relationships, a junction table will be created and the filter applies to the junction table itself. For example, if you have a books collection, with a M2M relationship to authors of each book, the junction collection will probably be named books_authors and have 3 fields : id, books_id and authors_id. To filter specific books depending on their authors you must go through the junction table and the authors_id field :

{
	"authors": {
		"authors_id": {
			"name": {
				"_eq": "Rijk van Zanten"
			}
		}
	}
}

{
	"authors": {
		"authors_id": {
			"name": {
				"_eq": "Rijk van Zanten"
			}
		}
	}
}

Logical Operators ​

You can nest or group multiple rules using the _and or _or logical operators. Each logical operator holds an array of Filter Rules, allowing for more complex filtering. Also note in the example that Logical Operators can be sub-nested into Logical Operators. However, they cannot be sub-nested into Filter Rules.

{
	"_or": [
		{
			"_and": [
				{
					"user_created": {
						"_eq": "$CURRENT_USER"
					}
				},
				{
					"status": {
						"_in": ["published", "draft"]
					}
				}
			]
		},
		{
			"_and": [
				{
					"user_created": {
						"_neq": "$CURRENT_USER"
					}
				},
				{
					"status": {
						"_in": ["published"]
					}
				}
			]
		}
	]
}

{
	"_or": [
		{
			"_and": [
				{
					"user_created": {
						"_eq": "$CURRENT_USER"
					}
				},
				{
					"status": {
						"_in": ["published", "draft"]
					}
				}
			]
		},
		{
			"_and": [
				{
					"user_created": {
						"_neq": "$CURRENT_USER"
					}
				},
				{
					"status": {
						"_in": ["published"]
					}
				}
			]
		}
	]
}

Some vs None in One-to-Many ​

When applying filters to a one-to-many field, Directus will default to a "some" search, for example in:

{
	"categories": {
		"name": {
			"_eq": "Recipe"
		}
	}
}

{
	"categories": {
		"name": {
			"_eq": "Recipe"
		}
	}
}

the top level parent will be returned if one of the categories has the name Recipe. This behavior can be overridden by using the explicit _some and _none operators, for example:

{
	"categories": {
		"_none": {
			"name": {
				"_eq": "Recipe"
			}
		}
	}
}

{
	"categories": {
		"_none": {
			"name": {
				"_eq": "Recipe"
			}
		}
	}
}

will fetch all parent items that don't have the category "Recipe"

Dynamic Variables ​

In addition to static values, you can also filter against dynamic values using the following variables.

• $CURRENT_USER — The primary key of the currently authenticated user
• $CURRENT_ROLE — The primary key of the role for the currently authenticated user
• $NOW — The current timestamp
• $NOW(<adjustment>) - The current timestamp plus/minus a given distance, for example $NOW(-1 year), $NOW(+2 hours)