Issue 2925: export schema with special chars

worker/export.go

@@ -27,6 +27,7 @@ import (
 	"strconv"
 	"strings"
 	"time"
+	"unicode"
 
 	"github.com/golang/glog"
 	"golang.org/x/net/context"
@@ -54,6 +55,20 @@ var rdfTypeMap = map[types.TypeID]string{
 	types.PasswordID: "xs:password",
 }
 
+// Having '<' and '>' around all predicates makes the exported schema harder
+// for humans to look at, so only put them on predicates containing "exotic"
+// characters (i.e. ones not in this list).
+var predNonSpecialChars = unicode.RangeTable{
+	R16: []unicode.Range16{
+		// Ranges must be in order.
+		unicode.Range16{'.', '.', 1},
+		unicode.Range16{'0', '9', 1},
+		unicode.Range16{'A', 'Z', 1},
+		unicode.Range16{'_', '_', 1},
+		unicode.Range16{'a', 'z', 1},
+	},
+}
+
 func toRDF(pl *posting.List, prefix string, readTs uint64) (*bpb.KVList, error) {
 	var buf bytes.Buffer
 
@@ -142,7 +157,10 @@ func toRDF(pl *posting.List, prefix string, readTs uint64) (*bpb.KVList, error)
 func toSchema(attr string, update pb.SchemaUpdate) (*bpb.KVList, error) {
 	// bytes.Buffer never returns error for any of the writes. So, we don't need to check them.
 	var buf bytes.Buffer
-	if strings.ContainsRune(attr, ':') {
+	isSpecial := func(r rune) bool {
+		return !(unicode.In(r, &predNonSpecialChars))
+	}
+	if strings.IndexFunc(attr, isSpecial) >= 0 {
 		buf.WriteRune('<')
 		buf.WriteString(attr)
 		buf.WriteRune('>')

worker/export_test.go

@@ -273,6 +273,81 @@ func TestToSchema(t *testing.T) {
 			},
 			expected: "<Alice:best>:string @reverse @lang . \n",
 		},
+		{
+			skv: &skv{
+				attr: "username/password",
+				schema: pb.SchemaUpdate{
+					Predicate: "",
+					ValueType: pb.Posting_STRING,
+					Directive: pb.SchemaUpdate_NONE,
+					List:      false,
+					Count:     false,
+					Upsert:    false,
+					Lang:      false,
+				},
+			},
+			expected: "<username/password>:string . \n",
+		},
+		{
+			skv: &skv{
+				attr: "B*-tree",
+				schema: pb.SchemaUpdate{
+					Predicate: "",
+					ValueType: pb.Posting_UID,
+					Directive: pb.SchemaUpdate_REVERSE,
+					List:      true,
+					Count:     false,
+					Upsert:    false,
+					Lang:      false,
+				},
+			},
+			expected: "<B*-tree>:[uid] @reverse . \n",
+		},
+		{
+			skv: &skv{
+				attr: "base_de_données",
+				schema: pb.SchemaUpdate{
+					Predicate: "",
+					ValueType: pb.Posting_STRING,
+					Directive: pb.SchemaUpdate_NONE,
+					List:      false,
+					Count:     false,
+					Upsert:    false,
+					Lang:      true,
+				},
+			},
+			expected: "<base_de_données>:string @lang . \n",
+		},
+		{
+			skv: &skv{
+				attr: "data_base",
+				schema: pb.SchemaUpdate{
+					Predicate: "",
+					ValueType: pb.Posting_STRING,
+					Directive: pb.SchemaUpdate_NONE,
+					List:      false,
+					Count:     false,
+					Upsert:    false,
+					Lang:      true,
+				},
+			},
+			expected: "data_base:string @lang . \n",
+		},
+		{
+			skv: &skv{
+				attr: "data.base",
+				schema: pb.SchemaUpdate{
+					Predicate: "",
+					ValueType: pb.Posting_STRING,
+					Directive: pb.SchemaUpdate_NONE,
+					List:      false,
+					Count:     false,
+					Upsert:    false,
+					Lang:      true,
+				},
+			},
+			expected: "data.base:string @lang . \n",
+		},
 	}
 	for _, testCase := range testCases {
 		list, err := toSchema(testCase.skv.attr, testCase.skv.schema)