UTF-16 support for JSON and option to skip BOMs

gen_testdata.go

@@ -0,0 +1,56 @@
+// +build ignore
+
+// gen_testdata clones the utf-8 tests data to the other
+// unicode encodings and adds BOM variants of each.
+package main
+
+import (
+	"io/ioutil"
+	"log"
+	"os"
+	"path/filepath"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/unicode"
+)
+
+func main() {
+	var xforms = []struct {
+		dir, bom string
+		enc      encoding.Encoding
+	}{
+		{"testdata/utf-16be", "\xFE\xFF", unicode.UTF16(unicode.BigEndian, unicode.IgnoreBOM)},
+		{"testdata/utf-16le", "\xFF\xFE", unicode.UTF16(unicode.LittleEndian, unicode.IgnoreBOM)},
+	}
+
+	paths, _ := filepath.Glob("testdata/utf-8/*")
+	for _, p := range paths {
+		src, err := ioutil.ReadFile(p)
+		if err != nil {
+			log.Fatal(err)
+		}
+
+		write("testdata/utf-8_bom", p, "\xEF\xBB\xBF", src)
+		for _, xform := range xforms {
+			dst, err := xform.enc.NewEncoder().Bytes(src)
+			if err != nil {
+				log.Fatal(err)
+			}
+			write(xform.dir, p, "", dst)
+			write(xform.dir+"_bom", p, xform.bom, dst)
+		}
+	}
+}
+
+func write(dir, orig, bom string, buf []byte) {
+	f, err := os.Create(filepath.Join(dir, filepath.Base(orig)))
+	if err != nil {
+		log.Fatal(err)
+	}
+	if _, err = f.Write([]byte(bom)); err != nil {
+		log.Fatal(err)
+	}
+	if _, err = f.Write(buf); err != nil {
+		log.Fatal(err)
+	}
+}

main.go

@@ -2,8 +2,11 @@
 // a provided JSON Schema - https://json-schema.org/
 package main
 
+//go:generate go run gen_testdata.go
+
 import (
 	"bufio"
+	"bytes"
 	"flag"
 	"fmt"
 	"io"
@@ -15,21 +18,37 @@ import (
 	"strings"
 	"sync"
 
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/unicode"
+
 	"github.com/ghodss/yaml"
 	"github.com/mitchellh/go-homedir"
 	"github.com/xeipuuv/gojsonschema"
 )
 
 var (
-	version     = "v1.3.0-dev"
+	version     = "v1.4.0-dev"
 	schemaFlag  = flag.String("s", "", "primary JSON schema to validate against, required")
 	quietFlag   = flag.Bool("q", false, "quiet, only print validation failures and errors")
 	versionFlag = flag.Bool("v", false, "print version and exit")
+	bomFlag     = flag.Bool("b", false, "allow BOM in JSON files, error if seen and unset")
 
 	listFlags stringFlags
 	refFlags  stringFlags
 )
 
+// https://en.wikipedia.org/wiki/Byte_order_mark#Byte_order_marks_by_encoding
+const (
+	bomUTF8    = "\xEF\xBB\xBF"
+	bomUTF16BE = "\xFE\xFF"
+	bomUTF16LE = "\xFF\xFE"
+)
+
+var (
+	encUTF16BE = unicode.UTF16(unicode.BigEndian, unicode.IgnoreBOM)
+	encUTF16LE = unicode.UTF16(unicode.LittleEndian, unicode.IgnoreBOM)
+)
+
 func init() {
 	flag.Var(&listFlags, "l", "validate JSON documents from newline separated paths and/or globs in a text file (relative to the basename of the file itself)")
 	flag.Var(&refFlags, "r", "referenced schema(s), can be globs and/or used multiple times")
@@ -60,7 +79,7 @@ func realMain(args []string, w io.Writer) int {
 		dir := filepath.Dir(list)
 		f, err := os.Open(list)
 		if err != nil {
-			log.Fatalf("%s: %s\n", list, err)
+			return schemaError("%s: %s", list, err)
 		}
 		defer f.Close()
 
@@ -74,7 +93,7 @@ func realMain(args []string, w io.Writer) int {
 			docs = append(docs, glob(pattern)...)
 		}
 		if err := scanner.Err(); err != nil {
-			log.Fatalf("%s: invalid file list: %s\n", list, err)
+			return schemaError("%s: invalid file list: %s", list, err)
 		}
 	}
 	if len(docs) == 0 {
@@ -85,13 +104,13 @@ func realMain(args []string, w io.Writer) int {
 	sl := gojsonschema.NewSchemaLoader()
 	schemaPath, err := filepath.Abs(*schemaFlag)
 	if err != nil {
-		log.Fatalf("%s: unable to convert to absolute path: %s\n", *schemaFlag, err)
+		return schemaError("%s: unable to convert to absolute path: %s", *schemaFlag, err)
 	}
 	for _, ref := range refFlags {
 		for _, p := range glob(ref) {
 			absPath, err := filepath.Abs(p)
 			if err != nil {
-				log.Fatalf("%s: unable to convert to absolute path: %s\n", absPath, err)
+				return schemaError("%s: unable to convert to absolute path: %s", absPath, err)
 			}
 
 			if absPath == schemaPath {
@@ -100,22 +119,22 @@ func realMain(args []string, w io.Writer) int {
 
 			loader, err := jsonLoader(absPath)
 			if err != nil {
-				log.Fatalf("%s: unable to load schema ref: %s\n", *schemaFlag, err)
+				return schemaError("%s: unable to load schema ref: %s", *schemaFlag, err)
 			}
 
 			if err := sl.AddSchemas(loader); err != nil {
-				log.Fatalf("%s: invalid schema: %s\n", p, err)
+				return schemaError("%s: invalid schema: %s", p, err)
 			}
 		}
 	}
 
 	schemaLoader, err := jsonLoader(schemaPath)
 	if err != nil {
-		log.Fatalf("%s: unable to load schema: %s\n", *schemaFlag, err)
+		return schemaError("%s: unable to load schema: %s", *schemaFlag, err)
 	}
 	schema, err := sl.Compile(schemaLoader)
 	if err != nil {
-		log.Fatalf("%s: invalid schema: %s\n", *schemaFlag, err)
+		return schemaError("%s: invalid schema: %s", *schemaFlag, err)
 	}
 
 	// Validate the schema against each doc in parallel, limiting simultaneous
@@ -131,7 +150,6 @@ func realMain(args []string, w io.Writer) int {
 			sem <- 0
 			defer func() { <-sem }()
 
-
 			loader, err := jsonLoader(path)
 			if err != nil {
 				msg := fmt.Sprintf("%s: error: load doc: %s", path, err)
@@ -190,19 +208,62 @@ func jsonLoader(path string) (gojsonschema.JSONLoader, error) {
 	}
 	switch filepath.Ext(path) {
 	case ".yml", ".yaml":
+		// TODO YAML requires the precense of a BOM to detect UTF-16
+		// text. Is there a decent hueristic to detect UTF-16 text
+		// missing a BOM so we can provide a better error message?
 		buf, err = yaml.YAMLToJSON(buf)
+	default:
+		buf, err = jsonDecodeCharset(buf)
 	}
 	if err != nil {
 		return nil, err
 	}
+	// TODO What if we have an empty document?
 	return gojsonschema.NewBytesLoader(buf), nil
 }
 
+// jsonDecodeCharset attempts to detect UTF-16 (LE or BE) JSON text and
+// decode as appropriate. It also skips a BOM at the start of the buffer
+// if `-b` was specified. Presence of a BOM is an error otherwise.
+func jsonDecodeCharset(buf []byte) ([]byte, error) {
+	if len(buf) < 2 { // UTF-8
+		return buf, nil
+	}
+
+	bom := ""
+	var enc encoding.Encoding
+	switch {
+	case bytes.HasPrefix(buf, []byte(bomUTF8)):
+		bom = bomUTF8
+	case bytes.HasPrefix(buf, []byte(bomUTF16BE)):
+		bom = bomUTF16BE
+		enc = encUTF16BE
+	case bytes.HasPrefix(buf, []byte(bomUTF16LE)):
+		bom = bomUTF16LE
+		enc = encUTF16LE
+	case buf[0] == 0:
+		enc = encUTF16BE
+	case buf[1] == 0:
+		enc = encUTF16LE
+	}
+
+	if bom != "" {
+		if !*bomFlag {
+			return nil, fmt.Errorf("unexpected BOM, see `-b` flag")
+		}
+		buf = buf[len(bom):]
+	}
+	if enc != nil {
+		return enc.NewDecoder().Bytes(buf)
+	}
+	return buf, nil
+}
+
 func printUsage() {
 	fmt.Fprintf(os.Stderr, `Usage: %s -s schema.(json|yml) [options] document.(json|yml) ...
 
-  yajsv validates JSON and YAML document(s) against a schema. One of three statuses are
-  reported per document:
+  yajsv validates JSON and YAML document(s) against a schema. One of three status
+  results are reported per document:
 
     pass: Document is valid relative to the schema
     fail: Document is invalid relative to the schema
@@ -212,7 +273,8 @@ func printUsage() {
   schema validation failure.
 
   Sets the exit code to 1 on any failures, 2 on any errors, 3 on both, 4 on
-  invalid usage. Otherwise, 0 is returned if everything passes validation.
+  invalid usage, 5 on schema definition or file-list errors. Otherwise, 0 is
+  returned if everything passes validation.
 
 Options:
 
@@ -227,6 +289,11 @@ func usageError(msg string) int {
 	return 4
 }
 
+func schemaError(format string, args ...interface{}) int {
+	fmt.Fprintf(os.Stderr, format+"\n", args...)
+	return 5
+}
+
 // glob is a wrapper that also resolves `~` since we may be skipping
 // the shell expansion when single-quoting globs at the command line
 func glob(pattern string) []string {