Added ability to sort by multiple fields #1541

pkg/yqlib/doc/operators/sort.md

@@ -25,6 +25,28 @@ will output
 - a: cat
 ```
 
+## Sort by multiple fields
+Given a sample.yml file of:
+```yaml
+- a: dog
+- a: cat
+  b: banana
+- a: cat
+  b: apple
+```
+then
+```bash
+yq 'sort_by(.a, .b)' sample.yml
+```
+will output
+```yaml
+- a: cat
+  b: apple
+- a: cat
+  b: banana
+- a: dog
+```
+
 ## Sort descending by string field
 Use sort with reverse to sort in descending order.
 

pkg/yqlib/operator_sort.go

@@ -42,18 +42,7 @@ func sortByOperator(d *dataTreeNavigator, context Context, expressionNode *Expre
 				return Context{}, err
 			}
 
-			nodeToCompare := &yaml.Node{Kind: yaml.ScalarNode, Tag: "!!null"}
-			if compareContext.MatchingNodes.Len() > 0 {
-				nodeToCompare = compareContext.MatchingNodes.Front().Value.(*CandidateNode).Node
-			}
-
-			log.Debug("going to compare %v by %v", NodeToString(candidate.CreateReplacement(originalNode)), NodeToString(candidate.CreateReplacement(nodeToCompare)))
-
-			sortableArray[i] = sortableNode{Node: originalNode, NodeToCompare: nodeToCompare, dateTimeLayout: context.GetDateTimeLayout()}
-
-			if nodeToCompare.Kind != yaml.ScalarNode {
-				return Context{}, fmt.Errorf("sort only works for scalars, got %v", nodeToCompare.Tag)
-			}
+			sortableArray[i] = sortableNode{Node: originalNode, CompareContext: compareContext, dateTimeLayout: context.GetDateTimeLayout()}
 
 		}
 
@@ -72,7 +61,7 @@ func sortByOperator(d *dataTreeNavigator, context Context, expressionNode *Expre
 
 type sortableNode struct {
 	Node           *yaml.Node
-	NodeToCompare  *yaml.Node
+	CompareContext Context
 	dateTimeLayout string
 }
 
@@ -82,9 +71,28 @@ func (a sortableNodeArray) Len() int      { return len(a) }
 func (a sortableNodeArray) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
 
 func (a sortableNodeArray) Less(i, j int) bool {
-	lhs := a[i].NodeToCompare
-	rhs := a[j].NodeToCompare
+	lhsContext := a[i].CompareContext
+	rhsContext := a[j].CompareContext
 
+	rhsEl := rhsContext.MatchingNodes.Front()
+	for lhsEl := lhsContext.MatchingNodes.Front(); lhsEl != nil && rhsEl != nil; lhsEl = lhsEl.Next() {
+		lhs := lhsEl.Value.(*CandidateNode)
+		rhs := rhsEl.Value.(*CandidateNode)
+
+		result := a.compare(lhs.Node, rhs.Node, a[i].dateTimeLayout)
+
+		if result < 0 {
+			return true
+		} else if result > 0 {
+			return false
+		}
+
+		rhsEl = rhsEl.Next()
+	}
+	return false
+}
+
+func (a sortableNodeArray) compare(lhs *yaml.Node, rhs *yaml.Node, dateTimeLayout string) int {
 	lhsTag := lhs.Tag
 	rhsTag := rhs.Tag
 
@@ -99,7 +107,7 @@ func (a sortableNodeArray) Less(i, j int) bool {
 	}
 
 	isDateTime := lhsTag == "!!timestamp" && rhsTag == "!!timestamp"
-	layout := a[i].dateTimeLayout
+	layout := dateTimeLayout
 	// if the lhs is a string, it might be a timestamp in a custom format.
 	if lhsTag == "!!str" && layout != time.RFC3339 {
 		_, errLhs := parseDateTime(layout, lhs.Value)
@@ -108,13 +116,13 @@ func (a sortableNodeArray) Less(i, j int) bool {
 	}
 
 	if lhsTag == "!!null" && rhsTag != "!!null" {
-		return true
+		return -1
 	} else if lhsTag != "!!null" && rhsTag == "!!null" {
-		return false
+		return 1
 	} else if lhsTag == "!!bool" && rhsTag != "!!bool" {
-		return true
+		return -1
 	} else if lhsTag != "!!bool" && rhsTag == "!!bool" {
-		return false
+		return 1
 	} else if lhsTag == "!!bool" && rhsTag == "!!bool" {
 		lhsTruthy, err := isTruthyNode(lhs)
 		if err != nil {
@@ -125,20 +133,30 @@ func (a sortableNodeArray) Less(i, j int) bool {
 		if err != nil {
 			panic(fmt.Errorf("could not parse %v as boolean: %w", rhs.Value, err))
 		}
-
-		return !lhsTruthy && rhsTruthy
+		if lhsTruthy == rhsTruthy {
+			return 0
+		} else if lhsTruthy {
+			return 1
+		}
+		return -1
 	} else if isDateTime {
 		lhsTime, err := parseDateTime(layout, lhs.Value)
 		if err != nil {
 			log.Warningf("Could not parse time %v with layout %v for sort, sorting by string instead: %w", lhs.Value, layout, err)
-			return strings.Compare(lhs.Value, rhs.Value) < 0
+			return strings.Compare(lhs.Value, rhs.Value)
 		}
 		rhsTime, err := parseDateTime(layout, rhs.Value)
 		if err != nil {
 			log.Warningf("Could not parse time %v with layout %v for sort, sorting by string instead: %w", rhs.Value, layout, err)
-			return strings.Compare(lhs.Value, rhs.Value) < 0
+			return strings.Compare(lhs.Value, rhs.Value)
 		}
-		return lhsTime.Before(rhsTime)
+		if lhsTime.Equal(rhsTime) {
+			return 0
+		} else if lhsTime.Before(rhsTime) {
+			return -1
+		}
+
+		return 1
 	} else if lhsTag == "!!int" && rhsTag == "!!int" {
 		_, lhsNum, err := parseInt64(lhs.Value)
 		if err != nil {
@@ -148,7 +166,7 @@ func (a sortableNodeArray) Less(i, j int) bool {
 		if err != nil {
 			panic(err)
 		}
-		return lhsNum < rhsNum
+		return int(lhsNum - rhsNum)
 	} else if (lhsTag == "!!int" || lhsTag == "!!float") && (rhsTag == "!!int" || rhsTag == "!!float") {
 		lhsNum, err := strconv.ParseFloat(lhs.Value, 64)
 		if err != nil {
@@ -158,8 +176,14 @@ func (a sortableNodeArray) Less(i, j int) bool {
 		if err != nil {
 			panic(err)
 		}
-		return lhsNum < rhsNum
+		if lhsNum == rhsNum {
+			return 0
+		} else if lhsNum < rhsNum {
+			return -1
+		}
+
+		return 1
 	}
 
-	return strings.Compare(lhs.Value, rhs.Value) < 0
+	return strings.Compare(lhs.Value, rhs.Value)
 }

pkg/yqlib/operator_sort_test.go

@@ -11,6 +11,32 @@ var sortByOperatorScenarios = []expressionScenario{
 			"D0, P[], (!!seq)::[{a: apple}, {a: banana}, {a: cat}]\n",
 		},
 	},
+	{
+		description: "Sort by multiple fields",
+		document:    "[{a: dog},{a: cat, b: banana},{a: cat, b: apple}]",
+		expression:  `sort_by(.a, .b)`,
+		expected: []string{
+			"D0, P[], (!!seq)::[{a: cat, b: apple}, {a: cat, b: banana}, {a: dog}]\n",
+		},
+	},
+	{
+		description: "Sort by multiple fields",
+		skipDoc:     true,
+		document:    "[{a: dog, b: good},{a: cat, c: things},{a: cat, b: apple}]",
+		expression:  `sort_by(.a, .b)`,
+		expected: []string{
+			"D0, P[], (!!seq)::[{a: cat, c: things}, {a: cat, b: apple}, {a: dog, b: good}]\n",
+		},
+	},
+	{
+		description: "Sort by multiple fields",
+		skipDoc:     true,
+		document:    "[{a: dog, b: 0.1},{a: cat, b: 0.01},{a: cat, b: 0.001}]",
+		expression:  `sort_by(.a, .b)`,
+		expected: []string{
+			"D0, P[], (!!seq)::[{a: cat, b: 0.001}, {a: cat, b: 0.01}, {a: dog, b: 0.1}]\n",
+		},
+	},
 	{
 		description:    "Sort descending by string field",
 		subdescription: "Use sort with reverse to sort in descending order.",