Added ability to escape double quotes in double quotes

pkg/yqlib/doc/Boolean Operators.md

@@ -1,5 +1,14 @@
-The `or` and `and` operators take two parameters and return a boolean result. `not` flips a boolean from true to false, or vice versa. These are most commonly used with the `select` operator to filter particular nodes.
-## OR example
+The `or` and `and` operators take two parameters and return a boolean result. 
+
+`not` flips a boolean from true to false, or vice versa. 
+
+`any` will return `true` if there are any `true` values in a array sequence, and `all` will return true if _all_ elements in an array are true.
+
+`any_c(condition)` and `all_c(condition)` are like `any` and `all` but they take a condition expression that is used against each element to determine if it's `true`. Note: in `jq` you can simply pass a condition to `any` or `all` and it simply works - `yq` isn't that clever..yet
+
+These are most commonly used with the `select` operator to filter particular nodes.
+
+## `or` example
 Running
 ```bash
 yq eval --null-input 'true or false'
@@ -9,7 +18,7 @@ will output
 true
 ```
 
-## AND example
+## `and` example
 Running
 ```bash
 yq eval --null-input 'true and false'
@@ -41,6 +50,104 @@ will output
   b: fly
 ```
 
+## `any` returns true if any boolean in a given array is true
+Given a sample.yml file of:
+```yaml
+- false
+- true
+```
+then
+```bash
+yq eval 'any' sample.yml
+```
+will output
+```yaml
+true
+```
+
+## `any` returns false for an empty array
+Given a sample.yml file of:
+```yaml
+[]
+```
+then
+```bash
+yq eval 'any' sample.yml
+```
+will output
+```yaml
+false
+```
+
+## `any_c` returns true if any element in the array is true for the given condition.
+Given a sample.yml file of:
+```yaml
+a:
+  - rad
+  - awesome
+b:
+  - meh
+  - whatever
+```
+then
+```bash
+yq eval '.[] |= any_c(. == "awesome")' sample.yml
+```
+will output
+```yaml
+a: true
+b: false
+```
+
+## `all` returns true if all booleans in a given array are true
+Given a sample.yml file of:
+```yaml
+- true
+- true
+```
+then
+```bash
+yq eval 'all' sample.yml
+```
+will output
+```yaml
+true
+```
+
+## `all` returns true for an empty array
+Given a sample.yml file of:
+```yaml
+[]
+```
+then
+```bash
+yq eval 'all' sample.yml
+```
+will output
+```yaml
+true
+```
+
+## `all_c` returns true if all elements in the array are true for the given condition.
+Given a sample.yml file of:
+```yaml
+a:
+  - rad
+  - awesome
+b:
+  - meh
+  - 12
+```
+then
+```bash
+yq eval '.[] |= all_c(tag == "!!str")' sample.yml
+```
+will output
+```yaml
+a: true
+b: false
+```
+
 ## Not true is false
 Running
 ```bash

pkg/yqlib/doc/Entries.md

@@ -35,6 +35,19 @@ will output
   value: b
 ```
 
+## to_entries null
+Given a sample.yml file of:
+```yaml
+null
+```
+then
+```bash
+yq eval 'to_entries' sample.yml
+```
+will output
+```yaml
+```
+
 ## from_entries map
 Given a sample.yml file of:
 ```yaml

pkg/yqlib/doc/Unique.md

@@ -0,0 +1,82 @@
+This is used to filter out duplicated items in an array.
+
+## Unique array of scalars (string/numbers)
+Given a sample.yml file of:
+```yaml
+- 1
+- 2
+- 3
+- 2
+```
+then
+```bash
+yq eval 'unique' sample.yml
+```
+will output
+```yaml
+- 1
+- 2
+- 3
+```
+
+## Unique nulls
+Unique works on the node value, so it considers different representations of nulls to be different
+
+Given a sample.yml file of:
+```yaml
+- ~
+- null
+- ~
+- null
+```
+then
+```bash
+yq eval 'unique' sample.yml
+```
+will output
+```yaml
+- ~
+- null
+```
+
+## Unique all nulls
+Run against the node tag to unique all the nulls
+
+Given a sample.yml file of:
+```yaml
+- ~
+- null
+- ~
+- null
+```
+then
+```bash
+yq eval 'unique_by(tag)' sample.yml
+```
+will output
+```yaml
+- ~
+```
+
+## Unique array object fields
+Given a sample.yml file of:
+```yaml
+- name: harry
+  pet: cat
+- name: billy
+  pet: dog
+- name: harry
+  pet: dog
+```
+then
+```bash
+yq eval 'unique_by(.name)' sample.yml
+```
+will output
+```yaml
+- name: harry
+  pet: cat
+- name: billy
+  pet: dog
+```
+

pkg/yqlib/doc/aa.md

@@ -62,13 +62,13 @@ cat
 
 Note that this node holds not only its value 'cat', but comments and metadata too, including path and parent information.
 
-The `=` operator then pipes the 'root' context through the `rhs` expression of `.b` to retun the node
+The `=` operator then pipes the 'root' context through the `rhs` expression of `.b` to return the node
 
 ```yaml
 dog
 ```
 
-Both sides have now been evaluated, so the operator performs its actual operation of assignment, and copies across the value from the RHS to the value on the LHS, and it returns the now updated 'root' context:
+Both sides have now been evaluated, so now the operator copies across the value from the RHS to the value on the LHS, and it returns the now updated context:
 
 ```yaml
 a: cat
@@ -76,6 +76,41 @@ b: dog
 ```
 
 # Relative update (e.g. `|=`)
-There is another form of the `=` operator which we call the relative form. It's very similar to `=` but with one key differnce when evaluating the RHS expression.
+There is another form of the `=` operator which we call the relative form. It's very similar to `=` but with one key difference when evaluating the RHS expression.
 
-In the plain form, we pass in the 'root' level context to the RHS expresssion. In relative form, we pass in each result of the LHS to the RHS expression. Let's go through an example.
+In the plain form, we pass in the 'root' level context to the RHS expression. In relative form, we pass in _each result of the LHS_ to the RHS expression. Let's go through an example.
+
+Given a document like:
+
+```yaml
+a: 1
+b: thing
+```
+
+with an expression:
+
+```
+.a |= . + 1
+```
+
+Similar to the `=` operator, `|=` takes two operands, the LHS and RHS.
+
+It pipes the current context (the whole document) through the LHS expression  of `.a` to get the node value:
+
+```
+1
+```
+
+Now it pipes _that LHS context_ into the RHS expression `. + 1` (whereas in the `=` plain form it piped the original document context into the RHS) to yield:
+
+
+```
+2
+```
+
+The assignment operator then copies across the value from the RHS to the value on the LHS, and it returns the now updated 'root' context:
+
+```yaml
+a: 2
+b: thing
+```

pkg/yqlib/doc/headers/Boolean Operators.md

@@ -1 +1,9 @@
-The `or` and `and` operators take two parameters and return a boolean result. `not` flips a boolean from true to false, or vice versa. These are most commonly used with the `select` operator to filter particular nodes.
+The `or` and `and` operators take two parameters and return a boolean result. 
+
+`not` flips a boolean from true to false, or vice versa. 
+
+`any` will return `true` if there are any `true` values in a array sequence, and `all` will return true if _all_ elements in an array are true.
+
+`any_c(condition)` and `all_c(condition)` are like `any` and `all` but they take a condition expression that is used against each element to determine if it's `true`. Note: in `jq` you can simply pass a condition to `any` or `all` and it simply works - `yq` isn't that clever..yet
+
+These are most commonly used with the `select` operator to filter particular nodes.

pkg/yqlib/doc/headers/Unique.md

@@ -0,0 +1 @@
+This is used to filter out duplicated items in an array.

pkg/yqlib/expression_processing_test.go

@@ -12,6 +12,11 @@ var pathTests = []struct {
 	expectedTokens  []interface{}
 	expectedPostFix []interface{}
 }{
+	{
+		`"\""`,
+		append(make([]interface{}, 0), "\" (string)"),
+		append(make([]interface{}, 0), "\" (string)"),
+	},
 	{
 		`[]|join(".")`,
 		append(make([]interface{}, 0), "[", "EMPTY", "]", "PIPE", "JOIN", "(", ". (string)", ")"),

pkg/yqlib/expression_tokeniser.go

@@ -189,6 +189,7 @@ func stringValue(wrapped bool) lex.Action {
 		if wrapped {
 			value = unwrap(value)
 		}
+		value = strings.ReplaceAll(value, "\\\"", "\"")
 		return &token{TokenType: operationToken, Operation: createValueOperation(value, value)}, nil
 	}
 }
@@ -259,6 +260,8 @@ func initLexer() (*lex.Lexer, error) {
 	lexer.Add([]byte(`sortKeys`), opToken(sortKeysOpType))
 	lexer.Add([]byte(`select`), opToken(selectOpType))
 	lexer.Add([]byte(`has`), opToken(hasOpType))
+	lexer.Add([]byte(`unique`), opToken(uniqueOpType))
+	lexer.Add([]byte(`unique_by`), opToken(uniqueByOpType))
 	lexer.Add([]byte(`explode`), opToken(explodeOpType))
 	lexer.Add([]byte(`or`), opToken(orOpType))
 	lexer.Add([]byte(`and`), opToken(andOpType))
@@ -274,6 +277,11 @@ func initLexer() (*lex.Lexer, error) {
 	lexer.Add([]byte(`join`), opToken(joinStringOpType))
 	lexer.Add([]byte(`sub`), opToken(subStringOpType))
 
+	lexer.Add([]byte(`any`), opToken(anyOpType))
+	lexer.Add([]byte(`any_c`), opToken(anyConditionOpType))
+	lexer.Add([]byte(`all`), opToken(allOpType))
+	lexer.Add([]byte(`all_c`), opToken(allConditionOpType))
+
 	lexer.Add([]byte(`split`), opToken(splitStringOpType))
 	lexer.Add([]byte(`keys`), opToken(keysOpType))
 
@@ -327,7 +335,7 @@ func initLexer() (*lex.Lexer, error) {
 	lexer.Add([]byte(`[Nn][Uu][Ll][Ll]`), nullValue())
 	lexer.Add([]byte(`~`), nullValue())
 
-	lexer.Add([]byte(`"[^"]*"`), stringValue(true))
+	lexer.Add([]byte(`"([^"\\]*(\\.[^"\\]*)*)"`), stringValue(true))
 	lexer.Add([]byte(`strenv\([^\)]+\)`), envOp(true))
 	lexer.Add([]byte(`env\([^\)]+\)`), envOp(false))
 

pkg/yqlib/lib.go

@@ -61,6 +61,11 @@ var shortPipeOpType = &operationType{Type: "SHORT_PIPE", NumArgs: 2, Precedence:
 var lengthOpType = &operationType{Type: "LENGTH", NumArgs: 0, Precedence: 50, Handler: lengthOperator}
 var collectOpType = &operationType{Type: "COLLECT", NumArgs: 0, Precedence: 50, Handler: collectOperator}
 
+var anyOpType = &operationType{Type: "ANY", NumArgs: 0, Precedence: 50, Handler: anyOperator}
+var allOpType = &operationType{Type: "ALL", NumArgs: 0, Precedence: 50, Handler: allOperator}
+var anyConditionOpType = &operationType{Type: "ANY_CONDITION", NumArgs: 1, Precedence: 50, Handler: anyOperator}
+var allConditionOpType = &operationType{Type: "ALL_CONDITION", NumArgs: 1, Precedence: 50, Handler: allOperator}
+
 var toEntriesOpType = &operationType{Type: "TO_ENTRIES", NumArgs: 0, Precedence: 50, Handler: toEntriesOperator}
 var fromEntriesOpType = &operationType{Type: "FROM_ENTRIES", NumArgs: 0, Precedence: 50, Handler: fromEntriesOperator}
 var withEntriesOpType = &operationType{Type: "WITH_ENTRIES", NumArgs: 1, Precedence: 50, Handler: withEntriesOperator}
@@ -99,6 +104,8 @@ var recursiveDescentOpType = &operationType{Type: "RECURSIVE_DESCENT", NumArgs:
 
 var selectOpType = &operationType{Type: "SELECT", NumArgs: 1, Precedence: 50, Handler: selectOperator}
 var hasOpType = &operationType{Type: "HAS", NumArgs: 1, Precedence: 50, Handler: hasOperator}
+var uniqueOpType = &operationType{Type: "UNIQUE", NumArgs: 0, Precedence: 50, Handler: unique}
+var uniqueByOpType = &operationType{Type: "UNIQUE_BY", NumArgs: 1, Precedence: 50, Handler: uniqueBy}
 var deleteChildOpType = &operationType{Type: "DELETE", NumArgs: 1, Precedence: 40, Handler: deleteChildOperator}
 
 type Operation struct {

pkg/yqlib/operator_booleans.go

@@ -2,14 +2,13 @@ package yqlib
 
 import (
 	"container/list"
+	"fmt"
 
 	yaml "gopkg.in/yaml.v3"
 )
 
-func isTruthy(c *CandidateNode) (bool, error) {
-	node := unwrapDoc(c.Node)
+func isTruthyNode(node *yaml.Node) (bool, error) {
 	value := true
-
 	if node.Tag == "!!null" {
 		return false, nil
 	}
@@ -23,6 +22,11 @@ func isTruthy(c *CandidateNode) (bool, error) {
 	return value, nil
 }
 
+func isTruthy(c *CandidateNode) (bool, error) {
+	node := unwrapDoc(c.Node)
+	return isTruthyNode(node)
+}
+
 type boolOp func(bool, bool) bool
 
 func performBoolOp(op boolOp) func(d *dataTreeNavigator, context Context, lhs *CandidateNode, rhs *CandidateNode) (*CandidateNode, error) {
@@ -61,6 +65,73 @@ func performBoolOp(op boolOp) func(d *dataTreeNavigator, context Context, lhs *C
 	}
 }
 
+func findBoolean(wantBool bool, d *dataTreeNavigator, context Context, expressionNode *ExpressionNode, sequenceNode *yaml.Node) (bool, error) {
+	for _, node := range sequenceNode.Content {
+
+		if expressionNode != nil {
+			//need to evaluate the expression against the node
+			candidate := &CandidateNode{Node: node}
+			rhs, err := d.GetMatchingNodes(context.SingleChildContext(candidate), expressionNode)
+			if err != nil {
+				return false, err
+			}
+			if rhs.MatchingNodes.Len() > 0 {
+				node = rhs.MatchingNodes.Front().Value.(*CandidateNode).Node
+			} else {
+				// no results found, ignore this entry
+				continue
+			}
+		}
+
+		truthy, err := isTruthyNode(node)
+		if err != nil {
+			return false, err
+		}
+		if truthy == wantBool {
+			return true, nil
+		}
+	}
+	return false, nil
+}
+
+func allOperator(d *dataTreeNavigator, context Context, expressionNode *ExpressionNode) (Context, error) {
+	var results = list.New()
+
+	for el := context.MatchingNodes.Front(); el != nil; el = el.Next() {
+		candidate := el.Value.(*CandidateNode)
+		candidateNode := unwrapDoc(candidate.Node)
+		if candidateNode.Kind != yaml.SequenceNode {
+			return Context{}, fmt.Errorf("any only supports arrays, was %v", candidateNode.Tag)
+		}
+		booleanResult, err := findBoolean(false, d, context, expressionNode.Rhs, candidateNode)
+		if err != nil {
+			return Context{}, err
+		}
+		result := createBooleanCandidate(candidate, !booleanResult)
+		results.PushBack(result)
+	}
+	return context.ChildContext(results), nil
+}
+
+func anyOperator(d *dataTreeNavigator, context Context, expressionNode *ExpressionNode) (Context, error) {
+	var results = list.New()
+
+	for el := context.MatchingNodes.Front(); el != nil; el = el.Next() {
+		candidate := el.Value.(*CandidateNode)
+		candidateNode := unwrapDoc(candidate.Node)
+		if candidateNode.Kind != yaml.SequenceNode {
+			return Context{}, fmt.Errorf("any only supports arrays, was %v", candidateNode.Tag)
+		}
+		booleanResult, err := findBoolean(true, d, context, expressionNode.Rhs, candidateNode)
+		if err != nil {
+			return Context{}, err
+		}
+		result := createBooleanCandidate(candidate, booleanResult)
+		results.PushBack(result)
+	}
+	return context.ChildContext(results), nil
+}
+
 func orOperator(d *dataTreeNavigator, context Context, expressionNode *ExpressionNode) (Context, error) {
 	log.Debugf("-- orOp")
 	return crossFunction(d, context, expressionNode, performBoolOp(

pkg/yqlib/operator_booleans_test.go

@@ -6,7 +6,7 @@ import (
 
 var booleanOperatorScenarios = []expressionScenario{
 	{
-		description: "OR example",
+		description: "`or` example",
 		expression:  `true or false`,
 		expected: []string{
 			"D0, P[], (!!bool)::true\n",
@@ -29,7 +29,7 @@ var booleanOperatorScenarios = []expressionScenario{
 		},
 	},
 	{
-		description: "AND example",
+		description: "`and` example",
 		expression:  `true and false`,
 		expected: []string{
 			"D0, P[], (!!bool)::false\n",
@@ -43,6 +43,70 @@ var booleanOperatorScenarios = []expressionScenario{
 			"D0, P[], (!!seq)::- {a: bird, b: dog}\n- {a: cat, b: fly}\n",
 		},
 	},
+	{
+		description: "`any` returns true if any boolean in a given array is true",
+		document:    `[false, true]`,
+		expression:  "any",
+		expected: []string{
+			"D0, P[], (!!bool)::true\n",
+		},
+	},
+	{
+		description: "`any` returns false for an empty array",
+		document:    `[]`,
+		expression:  "any",
+		expected: []string{
+			"D0, P[], (!!bool)::false\n",
+		},
+	},
+	{
+		description: "`any_c` returns true if any element in the array is true for the given condition.",
+		document:    "a: [rad, awesome]\nb: [meh, whatever]",
+		expression:  `.[] |= any_c(. == "awesome")`,
+		expected: []string{
+			"D0, P[], (doc)::a: true\nb: false\n",
+		},
+	},
+	{
+		skipDoc:    true,
+		document:   `[false, false]`,
+		expression: "any",
+		expected: []string{
+			"D0, P[], (!!bool)::false\n",
+		},
+	},
+	{
+		description: "`all` returns true if all booleans in a given array are true",
+		document:    `[true, true]`,
+		expression:  "all",
+		expected: []string{
+			"D0, P[], (!!bool)::true\n",
+		},
+	},
+	{
+		skipDoc:    true,
+		document:   `[false, true]`,
+		expression: "all",
+		expected: []string{
+			"D0, P[], (!!bool)::false\n",
+		},
+	},
+	{
+		description: "`all` returns true for an empty array",
+		document:    `[]`,
+		expression:  "all",
+		expected: []string{
+			"D0, P[], (!!bool)::true\n",
+		},
+	},
+	{
+		description: "`all_c` returns true if all elements in the array are true for the given condition.",
+		document:    "a: [rad, awesome]\nb: [meh, 12]",
+		expression:  `.[] |= all_c(tag == "!!str")`,
+		expected: []string{
+			"D0, P[], (doc)::a: true\nb: false\n",
+		},
+	},
 	{
 		skipDoc:    true,
 		expression: `false or false`,

pkg/yqlib/operator_entries.go

@@ -59,7 +59,9 @@ func toEntriesOperator(d *dataTreeNavigator, context Context, expressionNode *Ex
 		case yaml.SequenceNode:
 			results.PushBack(toEntriesfromSeq(candidate))
 		default:
-			return Context{}, fmt.Errorf("%v has no keys", candidate.Node.Tag)
+			if candidateNode.Tag != "!!null" {
+				return Context{}, fmt.Errorf("%v has no keys", candidate.Node.Tag)
+			}
 		}
 	}
 
@@ -133,26 +135,26 @@ func withEntriesOperator(d *dataTreeNavigator, context Context, expressionNode *
 	//to_entries on the context
 	toEntries, err := toEntriesOperator(d, context, expressionNode)
 	if err != nil {
-		return Context{}, nil
+		return Context{}, err
 	}
 
 	//run expression against entries
 	// splat toEntries and pipe it into Rhs
 	splatted, err := splat(d, toEntries, traversePreferences{})
 	if err != nil {
-		return Context{}, nil
+		return Context{}, err
 	}
 
 	result, err := d.GetMatchingNodes(splatted, expressionNode.Rhs)
 	log.Debug("expressionNode.Rhs %v", expressionNode.Rhs.Operation.OperationType)
 	log.Debug("result %v", result)
 	if err != nil {
-		return Context{}, nil
+		return Context{}, err
 	}
 
 	collected, err := collectOperator(d, result, expressionNode)
 	if err != nil {
-		return Context{}, nil
+		return Context{}, err
 	}
 
 	//from_entries on the result

pkg/yqlib/operator_entries_test.go

@@ -21,6 +21,12 @@ var entriesOperatorScenarios = []expressionScenario{
 			"D0, P[], (!!seq)::- key: 0\n  value: a\n- key: 1\n  value: b\n",
 		},
 	},
+	{
+		description: "to_entries null",
+		document:    `null`,
+		expression:  `to_entries`,
+		expected:    []string{},
+	},
 	{
 		description: "from_entries map",
 		document:    `{a: 1, b: 2}`,

pkg/yqlib/operator_equals.go

@@ -10,14 +10,26 @@ func equalsOperator(d *dataTreeNavigator, context Context, expressionNode *Expre
 func isEquals(flip bool) func(d *dataTreeNavigator, context Context, lhs *CandidateNode, rhs *CandidateNode) (*CandidateNode, error) {
 	return func(d *dataTreeNavigator, context Context, lhs *CandidateNode, rhs *CandidateNode) (*CandidateNode, error) {
 		value := false
-
+		log.Debugf("-- isEquals cross function")
 		if lhs == nil && rhs == nil {
 			owner := &CandidateNode{}
 			return createBooleanCandidate(owner, !flip), nil
 		} else if lhs == nil {
-			return createBooleanCandidate(rhs, flip), nil
+			log.Debugf("lhs nil, but rhs is not")
+			rhsNode := unwrapDoc(rhs.Node)
+			value := rhsNode.Tag == "!!null"
+			if flip {
+				value = !value
+			}
+			return createBooleanCandidate(rhs, value), nil
 		} else if rhs == nil {
-			return createBooleanCandidate(lhs, flip), nil
+			log.Debugf("lhs not nil, but rhs is")
+			lhsNode := unwrapDoc(lhs.Node)
+			value := lhsNode.Tag == "!!null"
+			if flip {
+				value = !value
+			}
+			return createBooleanCandidate(lhs, value), nil
 		}
 
 		lhsNode := unwrapDoc(lhs.Node)
@@ -37,6 +49,6 @@ func isEquals(flip bool) func(d *dataTreeNavigator, context Context, lhs *Candid
 }
 
 func notEqualsOperator(d *dataTreeNavigator, context Context, expressionNode *ExpressionNode) (Context, error) {
-	log.Debugf("-- equalsOperation")
+	log.Debugf("-- notEqualsOperator")
 	return crossFunction(d, context, expressionNode, isEquals(true), true)
 }

pkg/yqlib/operator_equals_test.go

@@ -14,6 +14,28 @@ var equalsOperatorScenarios = []expressionScenario{
 			"D0, P[], (!!bool)::false\n",
 		},
 	},
+	{
+		skipDoc:    true,
+		document:   "{a: {b: 10}}",
+		expression: "select(.c != null)",
+		expected:   []string{},
+	},
+	{
+		skipDoc:    true,
+		document:   "{a: {b: 10}}",
+		expression: "select(.d == .c)",
+		expected: []string{
+			"D0, P[], (doc)::{a: {b: 10}}\n",
+		},
+	},
+	{
+		skipDoc:    true,
+		document:   "{a: {b: 10}}",
+		expression: "select(null == .c)",
+		expected: []string{
+			"D0, P[], (doc)::{a: {b: 10}}\n",
+		},
+	},
 	{
 		skipDoc:    true,
 		document:   "{a: { b: {things: \"\"}, f: [1], g: [] }}",

pkg/yqlib/operator_select.go

@@ -24,7 +24,9 @@ func selectOperator(d *dataTreeNavigator, context Context, expressionNode *Expre
 
 		if first != nil {
 			result := first.Value.(*CandidateNode)
+			log.Debugf("result %v", NodeToString(result))
 			includeResult, errDecoding := isTruthy(result)
+			log.Debugf("isTruthy %v", includeResult)
 			if errDecoding != nil {
 				return Context{}, errDecoding
 			}

pkg/yqlib/operator_unique.go

@@ -0,0 +1,59 @@
+package yqlib
+
+import (
+	"container/list"
+	"fmt"
+
+	"github.com/elliotchance/orderedmap"
+	yaml "gopkg.in/yaml.v3"
+)
+
+func unique(d *dataTreeNavigator, context Context, expressionNode *ExpressionNode) (Context, error) {
+	selfExpression := &ExpressionNode{Operation: &Operation{OperationType: selfReferenceOpType}}
+	uniqueByExpression := &ExpressionNode{Operation: &Operation{OperationType: uniqueByOpType}, Rhs: selfExpression}
+	return uniqueBy(d, context, uniqueByExpression)
+
+}
+
+func uniqueBy(d *dataTreeNavigator, context Context, expressionNode *ExpressionNode) (Context, error) {
+
+	log.Debugf("-- uniqueBy Operator")
+	var results = list.New()
+
+	for el := context.MatchingNodes.Front(); el != nil; el = el.Next() {
+		candidate := el.Value.(*CandidateNode)
+		candidateNode := unwrapDoc(candidate.Node)
+
+		if candidateNode.Kind != yaml.SequenceNode {
+			return Context{}, fmt.Errorf("Only arrays are supported for unique")
+		}
+
+		var newMatches = orderedmap.NewOrderedMap()
+		for _, node := range candidateNode.Content {
+			child := &CandidateNode{Node: node}
+			rhs, err := d.GetMatchingNodes(context.SingleChildContext(child), expressionNode.Rhs)
+
+			if err != nil {
+				return Context{}, err
+			}
+
+			first := rhs.MatchingNodes.Front()
+			keyCandidate := first.Value.(*CandidateNode)
+			keyValue := keyCandidate.Node.Value
+			_, exists := newMatches.Get(keyValue)
+
+			if !exists {
+				newMatches.Set(keyValue, child.Node)
+			}
+		}
+		resultNode := &yaml.Node{Kind: yaml.SequenceNode, Tag: "!!seq"}
+		for el := newMatches.Front(); el != nil; el = el.Next() {
+			resultNode.Content = append(resultNode.Content, el.Value.(*yaml.Node))
+		}
+
+		results.PushBack(candidate.CreateChild(nil, resultNode))
+	}
+
+	return context.ChildContext(results), nil
+
+}

pkg/yqlib/operator_unique_by_test.go

@@ -0,0 +1,49 @@
+package yqlib
+
+import (
+	"testing"
+)
+
+var uniqueOperatorScenarios = []expressionScenario{
+	{
+		description: "Unique array of scalars (string/numbers)",
+		document:    `[1,2,3,2]`,
+		expression:  `unique`,
+		expected: []string{
+			"D0, P[], (!!seq)::- 1\n- 2\n- 3\n",
+		},
+	},
+	{
+		description:    "Unique nulls",
+		subdescription: "Unique works on the node value, so it considers different representations of nulls to be different",
+		document:       `[~,null, ~, null]`,
+		expression:     `unique`,
+		expected: []string{
+			"D0, P[], (!!seq)::- ~\n- null\n",
+		},
+	},
+	{
+		description:    "Unique all nulls",
+		subdescription: "Run against the node tag to unique all the nulls",
+		document:       `[~,null, ~, null]`,
+		expression:     `unique_by(tag)`,
+		expected: []string{
+			"D0, P[], (!!seq)::- ~\n",
+		},
+	},
+	{
+		description: "Unique array object fields",
+		document:    `[{name: harry, pet: cat}, {name: billy, pet: dog}, {name: harry, pet: dog}]`,
+		expression:  `unique_by(.name)`,
+		expected: []string{
+			"D0, P[], (!!seq)::- {name: harry, pet: cat}\n- {name: billy, pet: dog}\n",
+		},
+	},
+}
+
+func TestUniqueOperatorScenarios(t *testing.T) {
+	for _, tt := range uniqueOperatorScenarios {
+		testScenario(t, &tt)
+	}
+	documentScenarios(t, "Unique", uniqueOperatorScenarios)
+}

pkg/yqlib/operators.go

@@ -24,7 +24,17 @@ func emptyOperator(d *dataTreeNavigator, context Context, expressionNode *Expres
 
 type crossFunctionCalculation func(d *dataTreeNavigator, context Context, lhs *CandidateNode, rhs *CandidateNode) (*CandidateNode, error)
 
-func resultsForRhs(d *dataTreeNavigator, context Context, lhsCandidate *CandidateNode, rhs Context, calculation crossFunctionCalculation, results *list.List) error {
+func resultsForRhs(d *dataTreeNavigator, context Context, lhsCandidate *CandidateNode, rhs Context, calculation crossFunctionCalculation, results *list.List, calcWhenEmpty bool) error {
+
+	if calcWhenEmpty && rhs.MatchingNodes.Len() == 0 {
+		resultCandidate, err := calculation(d, context, lhsCandidate, nil)
+		if err != nil {
+			return err
+		}
+		results.PushBack(resultCandidate)
+		return nil
+	}
+
 	for rightEl := rhs.MatchingNodes.Front(); rightEl != nil; rightEl = rightEl.Next() {
 		log.Debugf("Applying calc")
 		rhsCandidate := rightEl.Value.(*CandidateNode)
@@ -52,14 +62,7 @@ func doCrossFunc(d *dataTreeNavigator, context Context, expressionNode *Expressi
 	}
 
 	if calcWhenEmpty && lhs.MatchingNodes.Len() == 0 {
-		if rhs.MatchingNodes.Len() == 0 {
-			resultCandidate, err := calculation(d, context, nil, nil)
-			if err != nil {
-				return Context{}, err
-			}
-			results.PushBack(resultCandidate)
-		}
-		err := resultsForRhs(d, context, nil, rhs, calculation, results)
+		err := resultsForRhs(d, context, nil, rhs, calculation, results, calcWhenEmpty)
 		if err != nil {
 			return Context{}, err
 		}
@@ -68,7 +71,7 @@ func doCrossFunc(d *dataTreeNavigator, context Context, expressionNode *Expressi
 	for el := lhs.MatchingNodes.Front(); el != nil; el = el.Next() {
 		lhsCandidate := el.Value.(*CandidateNode)
 
-		err := resultsForRhs(d, context, lhsCandidate, rhs, calculation, results)
+		err := resultsForRhs(d, context, lhsCandidate, rhs, calculation, results, calcWhenEmpty)
 		if err != nil {
 			return Context{}, err
 		}