Simplified output of PicaMultiscriptRemodeler

The original version of `PicaMultiscriptRemodeler` used the original
script names for the intermediate layer of entities. This makes
processing of the output unnecessary complex because of the large number
of possible entity names. As currently only three types of scripts need
to be distinguished, the module was changed to map the script names to
one of the three script types which are then used as entity names.

Author: cboehme

src/main/java/org/culturegraph/mf/stream/pipe/PicaMultiscriptRemodeler.java

@@ -38,8 +38,13 @@
  * This module scans the input stream for Pica multiscript fields and remodels
  * them by merging all fields belong to the same multiscript group into one
  * entity. This entity has the original field name. Within this entity a new
- * entity for each of the original fields is created. These entities carry the
- * name of the script used in the field.
+ * entity for each of the original fields is created. These entities are named
+ * depending on the type of script used. Three scripts are distinguished:
+ * <ul>
+ *   <li>Latin</<li>
+ *   <li>NonLatinLR</li>
+ *   <li>NonLatinRL</li>
+ * </ul>
  *
  * The following example shows how the input
  *
@@ -56,16 +61,16 @@
  *
  * <pre>
  *    021A {
- *        Latn { T: 01, U: Latn, a: Title }
- *        Grek { T: 01, U: Grek, a: Greek title }
+ *        Latin { T: 01, U: Latn, a: Title }
+ *        NonLatinLR { T: 01, U: Grek, a: Greek title }
  *    }
  *    021C {
- *        Latn { T: 01, U: Latn, a: Subseries A }
- *        Grek { T: 01, U: Grek, a: Greek subseries A }
+ *        Latin { T: 01, U: Latn, a: Subseries A }
+ *        NonLatinLR { T: 01, U: Grek, a: Greek subseries A }
  *    }
  *    021C {
- *        Latn { T: 02, U: Latn, a: Subseries B }
- *        Grek { T: 02, U: Grek, a: Greek subseries B}
+ *        Latin { T: 02, U: Latn, a: Subseries B }
+ *        NonLatinLR { T: 02, U: Grek, a: Greek subseries B}
  *    }
  * </pre>
  *
@@ -97,11 +102,19 @@
 public final class PicaMultiscriptRemodeler extends
 		DefaultStreamPipe<StreamReceiver> {
 
+	public static final String ENTITY_NAME_FOR_LATIN = "Latin";
+	public static final String ENTITY_NAME_FOR_NON_LATIN_LR = "NonLatinLR";
+	public static final String ENTITY_NAME_FOR_NON_LATIN_RL = "NonLatinRL";
+
 	private static final BufferedField BEFORE_FIRST_FIELD = new BufferedField("", null);
 
 	private static final String GROUP_SUBFIELD = "T";
 	private static final String SCRIPT_SUBFIELD = "U";
 
+	private static final String LATIN_SCRIPT = "Latn";
+	private static final String ARABIC_SCRIPT = "Arab";
+	private static final String HEBREW_SCRIPT = "Hebr";
+
 	private BufferedField currentField;
 	private BufferedField lastField;
 
@@ -178,11 +191,11 @@ private void emitNonMultiscriptField() {
 	private void emitRemodeledMultiscriptField(final BufferedField firstField, final BufferedField secondField) {
 		getReceiver().startEntity(firstField.name);
 
-		getReceiver().startEntity(firstField.script);
+		getReceiver().startEntity(mapScriptToEntityName(firstField.script));
 		firstField.stream.replay();
 		getReceiver().endEntity();
 
-		getReceiver().startEntity(secondField.script);
+		getReceiver().startEntity(mapScriptToEntityName(secondField.script));
 		secondField.stream.replay();
 		getReceiver().endEntity();
 
@@ -198,6 +211,16 @@ private void emitAsSingleMultiscriptFields(final Map<?, BufferedField> fields) {
 		fields.clear();
 	}
 
+	private String mapScriptToEntityName(final String script) {
+		if (LATIN_SCRIPT.equals(script)) {
+			return ENTITY_NAME_FOR_LATIN;
+		} else if (ARABIC_SCRIPT.equals(script)
+				|| HEBREW_SCRIPT.equals(script)) {
+			return ENTITY_NAME_FOR_NON_LATIN_RL;
+		}
+		return ENTITY_NAME_FOR_NON_LATIN_LR;
+	}
+
 	private static class BufferedField {
 
 		public String group;

src/test/java/org/culturegraph/mf/stream/pipe/PicaMultiscriptRemodelerTest.java

@@ -41,12 +41,16 @@ public class PicaMultiscriptRemodelerTest {
 	private static final String FIELD_021C = "021C";
 	private static final String SCRIPT_LATIN = "Latn";
 	private static final String SCRIPT_GREEK = "Grek";
+	private static final String SCRIPT_ARABIC = "Arab";
+	private static final String SCRIPT_HEBREW = "Hebr";
 
 	private static final String VALUE_1 = "Subfield 1";
 	private static final String VALUE_2 = "Subfield 2";
 	private static final String VALUE_3 = "Subfield 3";
 	private static final String VALUE_1_GREEK = "ĸµ 1";
 	private static final String VALUE_2_GREEK = "ĸµ 2";
+	private static final String VALUE_1_ARABIC = "Subfield/Arabic 1";
+	private static final String VALUE_1_HEBREW = "Subfield/Hebrew 1";
 
 	private PicaMultiscriptRemodeler remodeler;
 
@@ -215,6 +219,34 @@ public void shouldPassThroughIncompleteMultiscriptFields() {
 		ordered.verify(receiver).endRecord();
 	}
 
+	@Test
+	public void shouldLabelArabicAsNonLatinRightToLeftScript() {
+		remodeler.startRecord(RECORD_ID);
+		emitMultscriptField(FIELD_021A, "01", SCRIPT_LATIN, VALUE_1);
+		emitMultscriptField(FIELD_021A, "01", SCRIPT_ARABIC, VALUE_1_ARABIC);
+		remodeler.endRecord();
+
+		final InOrder ordered = inOrder(receiver);
+		ordered.verify(receiver).startRecord(RECORD_ID);
+		verifyMultiscriptField(ordered, FIELD_021A, "01", SCRIPT_LATIN,
+				VALUE_1, SCRIPT_ARABIC, VALUE_1_ARABIC);
+		ordered.verify(receiver).endRecord();
+	}
+
+	@Test
+	public void shouldLabelHebrewAsNonLatinRightToLeftScript() {
+		remodeler.startRecord(RECORD_ID);
+		emitMultscriptField(FIELD_021A, "01", SCRIPT_LATIN, VALUE_1);
+		emitMultscriptField(FIELD_021A, "01", SCRIPT_HEBREW, VALUE_1_HEBREW);
+		remodeler.endRecord();
+
+		final InOrder ordered = inOrder(receiver);
+		ordered.verify(receiver).startRecord(RECORD_ID);
+		verifyMultiscriptField(ordered, FIELD_021A, "01", SCRIPT_LATIN,
+				VALUE_1, SCRIPT_HEBREW, VALUE_1_HEBREW);
+		ordered.verify(receiver).endRecord();
+	}
+
 	@Test
 	public void shouldClearStateOnResetStream() {
 		remodeler.startRecord(RECORD_ID);
@@ -247,12 +279,12 @@ private void verifyMultiscriptField(final InOrder ordered,
 			final String value1, final String script2, final String value2) {
 
 		ordered.verify(receiver).startEntity(field);
-		ordered.verify(receiver).startEntity(script1);
+		ordered.verify(receiver).startEntity(mapScriptToEntityName(script1));
 		ordered.verify(receiver).literal("T", groupNumber);
 		ordered.verify(receiver).literal("U", script1);
 		ordered.verify(receiver).literal("a", value1);
 		ordered.verify(receiver).endEntity();
-		ordered.verify(receiver).startEntity(script2);
+		ordered.verify(receiver).startEntity(mapScriptToEntityName(script2));
 		ordered.verify(receiver).literal("T", groupNumber);
 		ordered.verify(receiver).literal("U", script2);
 		ordered.verify(receiver).literal("a", value2);
@@ -270,4 +302,14 @@ private void verifySingleMultiscriptField(final InOrder ordered,
 		ordered.verify(receiver).endEntity();
 	}
 
+	private String mapScriptToEntityName(final String script) {
+		if (SCRIPT_LATIN.equals(script)) {
+			return PicaMultiscriptRemodeler.ENTITY_NAME_FOR_LATIN;
+		} else if (SCRIPT_ARABIC.equals(script)
+				|| SCRIPT_HEBREW.equals(script)) {
+			return PicaMultiscriptRemodeler.ENTITY_NAME_FOR_NON_LATIN_RL;
+		}
+		return PicaMultiscriptRemodeler.ENTITY_NAME_FOR_NON_LATIN_LR;
+	}
+
 }