change PdfPageLabels.getPageLabels() to scope prefixes to the range

they are defined for

While reading a PDF generated by Adobe InDesign CS6 using the Adobe
PDF Library 10.0.1 which included several page label sections I
noticed iText was keeping the prefix applied to subsequent sections
in the values return by PdfPageLabels.getPageLabels(), even when the
key was omitted in the following page label dictionary entries.
According to the PDF Reference 1.7, Table 8.10 says that /P is an
optional label prefix for page labels in this range. PDF Readers I
have installed scope the prfix only to the range it is specified.
This change empties out the prefix field to give behaviour I think
aligns with the reference and my experience of other PDF readers.

I am not aware of any potential regression issues introduced by this
change, existing page label dictionaries created by iText set /P to
an empty string so this will only affect reading PDFs with partially
prefixed page label ranges as far as I know.

src/main/java/com/itextpdf/text/pdf/PdfPageLabels.java

@@ -195,6 +195,9 @@ public static String[] getPageLabels(PdfReader reader) {
 				if (d.contains(PdfName.P)) {
 					prefix = ((PdfString)d.get(PdfName.P)).toUnicodeString();
 				}
+				else {
+					prefix = "";
+				}
 				if (d.contains(PdfName.S)) {
 					type = ((PdfName)d.get(PdfName.S)).toString().charAt(1);
 				}

src/test/java/com/itextpdf/text/pdf/PdfPageLabelsTest.java

@@ -0,0 +1,28 @@
+package com.itextpdf.text.pdf;
+
+import com.itextpdf.testutils.TestResourceUtils;
+import junit.framework.Assert;
+import org.junit.Test;
+
+import java.io.File;
+
+public class PdfPageLabelsTest {
+    @Test
+    public void testGetPageLabels() throws Exception {
+        File testFile = TestResourceUtils.getResourceAsTempFile(this, "test-prefix-reset.pdf");
+        String[] expectedPageLabels = new String[] {"i", "ii", "iii", "iv", "v", "1", "2", "3", "4", "5", "G1", "G2", "G3", "G4", "G5", "6", "7", "8", "9", "10"};
+
+        PdfReader reader = new PdfReader(testFile.getAbsolutePath());
+
+        String[] pageLabels = PdfPageLabels.getPageLabels(reader);
+
+        Assert.assertNotNull(pageLabels);
+        Assert.assertEquals(expectedPageLabels.length, pageLabels.length);
+
+        for (int page = 0; page < pageLabels.length; page++) {
+            Assert.assertEquals(expectedPageLabels[page], pageLabels[page]);
+        }
+
+        reader.close();
+    }
+}