Fixed: Support non-breaking spaces in numeric strings (OFBIZ-13168) (#…

…847)

In forms, numeric fields are represented by an input with type "text", which allows user to enter/paste all characters, including non-breaking spaces ('\u00A0', '\u202F', '\u2007'), like "29 000" (the space is \u202F). More specifically, a user can copy/paste a string from an external tool which uses non-breaking spaces as a thousands separator, and expect that a visually correct string will be correctly interpreted by OFBiz.

OFBiz uses java.text.NumberFormat::parse method, which does not support non-breaking spaces : characters after this kind of spaces are simply ignored, and "29 000" becomes "29".

This PR only illustrates the problem and a way of fixing it, I'm not sure that this would be a good solution (maybe handle submitted strings elsewhere ? maybe more upstream ?).

Thanks: Nereide team

framework/base/src/main/java/org/apache/ofbiz/base/util/ObjectType.java

@@ -20,6 +20,7 @@
 
 import java.io.Serializable;
 import java.lang.reflect.InvocationTargetException;
+import java.math.BigDecimal;
 import java.util.Collection;
 import java.util.HashMap;
 import java.util.List;
@@ -330,6 +331,15 @@ public static Object simpleTypeOrObjectConvert(Object obj, String type, String f
         }
 
         if (converter != null) {
+            // numeric types : replace non-breaking spaces (which break parsing) by normal spaces
+            List<?> numericClasses = UtilMisc.toList(BigDecimal.class, Double.class, Float.class);
+            if (obj instanceof String && numericClasses.contains(targetClass)) {
+                final String[] tmp = {String.valueOf(obj)};
+                List<Character> nonBreakingWhitespaces = UtilMisc.toList('\u00A0', '\u202F', '\u2007');
+                nonBreakingWhitespaces.forEach(character -> tmp[0] = tmp[0].replace(character, ' '));
+                obj = tmp[0];
+            }
+
             if (converter instanceof LocalizedConverter) {
                 LocalizedConverter<Object, Object> localizedConverter = UtilGenerics.cast(converter);
                 if (timeZone == null) {
@@ -345,6 +355,7 @@ public static Object simpleTypeOrObjectConvert(Object obj, String type, String f
                     throw new GeneralException(e.getMessage(), e);
                 }
             }
+
             try {
                 return converter.convert(obj);
             } catch (ConversionException e) {

framework/base/src/test/java/org/apache/ofbiz/base/util/ObjectTypeTests.java

@@ -48,6 +48,7 @@ public class ObjectTypeTests {
     // These numbers are all based on 1 / 128, which is a binary decimal
     // that can be represented by both float and double
     private final BigDecimal dcml = new BigDecimal("781.25");
+    private final BigDecimal largeBigDecimal = new BigDecimal("29000");
     private final Double dbl = Double.valueOf("7.8125E2");
     private final Float flt = Float.valueOf("7.8125E2");
     private final Long lng = Long.valueOf("781");
@@ -385,6 +386,12 @@ public void testString() throws GeneralException, Exception {
                 new String[] {"TimeDuration", "org.apache.ofbiz.base.util.TimeDuration"}, duration);
         simpleTypeOrObjectConvertTestError("String->error-TimeDuration", "o",
                 new String[] {"TimeDuration", "org.apache.ofbiz.base.util.TimeDuration"});
+
+        // usual pattern assumes that the String->BigDecimal conversion will break with bad timezone/locale
+        // which is not the case for this particular test
+        assertEquals("String->BigDecimal supports NBSP",
+                simpleTypeOrObjectConvert("29 000", "BigDecimal", null, LOCALE_DATA.goodTimeZone,
+                        LOCALE_DATA.goodLocale, false), largeBigDecimal);
     }
 
     @Test