Implement project operator

Fixes #7

pql.go

@@ -57,7 +57,7 @@ func splitQueries(expr *parser.TabularExpr) ([]*subquery, error) {
 	for i := 0; i < len(expr.Operators); i++ {
 		switch op := expr.Operators[i].(type) {
 		case *parser.SortOperator:
-			if lastSubquery != nil && lastSubquery.sort == nil && lastSubquery.take == nil {
+			if lastSubquery != nil && canAttachSort(lastSubquery.op) && lastSubquery.sort == nil && lastSubquery.take == nil {
 				lastSubquery.sort = op
 			} else {
 				lastSubquery = &subquery{
@@ -66,7 +66,7 @@ func splitQueries(expr *parser.TabularExpr) ([]*subquery, error) {
 				subqueries = append(subqueries, lastSubquery)
 			}
 		case *parser.TakeOperator:
-			if lastSubquery != nil && lastSubquery.take == nil {
+			if lastSubquery != nil && canAttachSort(lastSubquery.op) && lastSubquery.take == nil {
 				lastSubquery.take = op
 			} else {
 				lastSubquery = &subquery{
@@ -107,11 +107,39 @@ func splitQueries(expr *parser.TabularExpr) ([]*subquery, error) {
 	return subqueries, nil
 }
 
+// canAttachSort reports whether the given operator's subquery can have a sort clause attached.
+// This becomes significant for operators like "project"
+// because they change the identifiers in scope.
+func canAttachSort(op parser.TabularOperator) bool {
+	switch op.(type) {
+	case *parser.ProjectOperator, *parser.SummarizeOperator:
+		return false
+	default:
+		return true
+	}
+}
+
 func (sub *subquery) write(sb *strings.Builder) {
 	switch op := sub.op.(type) {
 	case nil:
 		sb.WriteString("SELECT * FROM ")
 		sb.WriteString(sub.sourceSQL)
+	case *parser.ProjectOperator:
+		sb.WriteString("SELECT ")
+		for i, col := range op.Cols {
+			if i > 0 {
+				sb.WriteString(", ")
+			}
+			if col.X == nil {
+				writeExpression(sb, col.Name)
+			} else {
+				writeExpression(sb, col.X)
+			}
+			sb.WriteString(" AS ")
+			quoteIdentifier(sb, col.Name.Name)
+		}
+		sb.WriteString(" FROM ")
+		sb.WriteString(sub.sourceSQL)
 	case *parser.WhereOperator:
 		sb.WriteString("SELECT * FROM ")
 		sb.WriteString(sub.sourceSQL)

testdata/Goldens/Project/input.pql

@@ -0,0 +1,2 @@
+StormEvents
+| project State, EventType, DamageProperty

testdata/Goldens/Project/output.csv

@@ -0,0 +1,5 @@
+ATLANTIC SOUTH,Waterspout,0
+FLORIDA,Heavy Rain,0
+FLORIDA,Tornado,6200000
+GEORGIA,Thunderstorm Wind,2000
+MISSISSIPPI,Thunderstorm Wind,20000

testdata/Goldens/Project/output.sql

@@ -0,0 +1 @@
+SELECT "State" AS "State", "EventType" AS "EventType", "DamageProperty" AS "DamageProperty" FROM "StormEvents";

testdata/Goldens/ProjectAfterTake/input.pql

@@ -0,0 +1,4 @@
+StormEvents
+| sort by EventId asc
+| take 3
+| project State, EventType, DamageProperty

testdata/Goldens/ProjectAfterTake/output.csv

@@ -0,0 +1,3 @@
+ATLANTIC SOUTH,Waterspout,0
+FLORIDA,Heavy Rain,0
+GEORGIA,Thunderstorm Wind,2000

testdata/Goldens/ProjectAfterTake/output.sql

@@ -0,0 +1,2 @@
+WITH "subquery0" AS (SELECT * FROM "StormEvents" ORDER BY "EventId" ASC NULLS FIRST LIMIT 3)
+SELECT "State" AS "State", "EventType" AS "EventType", "DamageProperty" AS "DamageProperty" FROM "subquery0";

testdata/Goldens/ProjectAssignment/input.pql

@@ -0,0 +1,2 @@
+StormEvents
+| project Id = EventId, DamagePropertyK = DamageProperty / 1000

testdata/Goldens/ProjectAssignment/output.csv

@@ -0,0 +1,5 @@
+11032,0
+11098,0
+60913,6200
+11503,2
+13913,20

testdata/Goldens/ProjectAssignment/output.sql

@@ -0,0 +1 @@
+SELECT "EventId" AS "Id", ("DamageProperty") / (1000) AS "DamagePropertyK" FROM "StormEvents";