Making doc extraction more reliable (#17842)

* Making doc extraction more reliable

* fix error message

* Adding test doc file

* Removing test that not works on CI because the lack of a binary

Co-authored-by: Mattermod <mattermod@users.noreply.github.com>

services/docextractor/docextractor_test.go

@@ -150,7 +150,7 @@ func TestExtract(t *testing.T) {
 		require.Equal(t, "", text)
 	})
 
-	t.Run("Wrong extension", func(t *testing.T) {
+	t.Run("Wrong docx extension", func(t *testing.T) {
 		data, err := testutils.ReadTestFile("sample-doc.pdf")
 		require.NoError(t, err)
 		text, err := Extract("sample-doc.docx", bytes.NewReader(data), ExtractSettings{})

services/docextractor/documents.go

@@ -31,7 +31,14 @@ func (de *documentExtractor) Match(filename string) bool {
 	return ok
 }
 
-func (de *documentExtractor) Extract(filename string, r io.ReadSeeker) (string, error) {
+func (de *documentExtractor) Extract(filename string, r io.ReadSeeker) (out string, outErr error) {
+	defer func() {
+		if r := recover(); r != nil {
+			out = ""
+			outErr = errors.New("error extracting document text")
+		}
+	}()
+
 	extension := strings.TrimPrefix(path.Ext(filename), ".")
 	converter, ok := doconvConverterByExtensions[extension]
 	if !ok {