GH-1300: Use Locale.ROOT when building C Data Interface format strings

Format.asString built the FixedSizeList, FixedSizeBinary and Decimal
format strings with String.format and no explicit Locale, so JVMs whose
default locale uses digits outside ASCII exported strings such as "+w:٨"
that other Arrow implementations cannot parse.

Closes #1300.

Author: lawofcycles

c/src/main/java/org/apache/arrow/c/Format.java

@@ -63,10 +63,10 @@ static String asString(ArrowType arrowType) {
         {
           ArrowType.Decimal type = (ArrowType.Decimal) arrowType;
           if (type.getBitWidth() == 128) {
-            return String.format("d:%d,%d", type.getPrecision(), type.getScale());
+            return String.format(Locale.ROOT, "d:%d,%d", type.getPrecision(), type.getScale());
           }
           return String.format(
-              "d:%d,%d,%d", type.getPrecision(), type.getScale(), type.getBitWidth());
+              Locale.ROOT, "d:%d,%d,%d", type.getPrecision(), type.getScale(), type.getBitWidth());
         }
       case Duration:
         {
@@ -88,12 +88,12 @@ static String asString(ArrowType arrowType) {
       case FixedSizeBinary:
         {
           ArrowType.FixedSizeBinary type = (ArrowType.FixedSizeBinary) arrowType;
-          return String.format("w:%d", type.getByteWidth());
+          return String.format(Locale.ROOT, "w:%d", type.getByteWidth());
         }
       case FixedSizeList:
         {
           ArrowType.FixedSizeList type = (ArrowType.FixedSizeList) arrowType;
-          return String.format("+w:%d", type.getListSize());
+          return String.format(Locale.ROOT, "+w:%d", type.getListSize());
         }
       case FloatingPoint:
         {

c/src/test/java/org/apache/arrow/c/FormatTest.java

@@ -20,6 +20,7 @@
 import static org.junit.jupiter.api.Assertions.assertThrows;
 import static org.junit.jupiter.api.Assertions.assertTrue;
 
+import java.util.Locale;
 import org.apache.arrow.vector.types.DateUnit;
 import org.apache.arrow.vector.types.FloatingPointPrecision;
 import org.apache.arrow.vector.types.IntervalUnit;
@@ -156,4 +157,20 @@ public void testAsType()
     assertThrows(UnsupportedOperationException.class, () -> Format.asType(":", 0L));
     assertThrows(NumberFormatException.class, () -> Format.asType("w:1,2,3", 0L));
   }
+
+  @Test
+  public void testAsStringIgnoresDefaultLocale() {
+    // Locales that use digits other than 0-9 (Arabic-Indic, Bengali, Devanagari, ...)
+    // must not leak into the format string, which other implementations parse as ASCII.
+    Locale saved = Locale.getDefault();
+    try {
+      Locale.setDefault(Locale.forLanguageTag("ar-EG"));
+      assertEquals("d:10,2", Format.asString(new ArrowType.Decimal(10, 2, 128)));
+      assertEquals("d:10,2,256", Format.asString(new ArrowType.Decimal(10, 2, 256)));
+      assertEquals("w:16", Format.asString(new ArrowType.FixedSizeBinary(16)));
+      assertEquals("+w:8", Format.asString(new ArrowType.FixedSizeList(8)));
+    } finally {
+      Locale.setDefault(saved);
+    }
+  }
 }