StringRenderer: fix bug with .substring() vs code points...

RFC 6570 section 2.1 defines a character as a Unicode character; which means
code points outside the BMP are possible.

In which case the prefix modifier must return the matching number of code
points, not the number of Java chars. .substring() is therefore not suitable,
neither is .length().

Author: fge

src/main/java/com/github/fge/uritemplate/render/StringRenderer.java

@@ -61,11 +61,19 @@ private String doRender(final VariableSpec varspec, final String value)
             ret += '=';
         }
         // Account for a prefix, if any. Note: explode modifier is ignored.
-        final int len = value.length();
+        final int len = value.codePointCount(0, value.length());
         final int prefixLen = varspec.getPrefixLength();
         final String val = prefixLen == -1 ? value
-            : value.substring(0, Math.min(len, prefixLen));
+            : nFirstCodePoints(value, Math.min(len, prefixLen));
         ret += pctEncode(val);
         return ret;
     }
+
+    private static String nFirstCodePoints(final String s, final int n)
+    {
+        int realIndex = n;
+        while (s.codePointCount(0, realIndex) != n)
+            realIndex++;
+        return s.substring(0, realIndex);
+    }
 }

src/test/resources/expand/strings.json

@@ -2,9 +2,11 @@
     "vars": {
         "var": "value",
         "hello": "Hello World!",
-        "empty": ""
+        "empty": "",
+        "poo": "\ud83d\udca9 is a pile of poo"
     },
     "tests": {
+        "{poo:4}": "%F0%9F%92%A9%20is",
         "{var}": "value",
         "{+var}": "value",
         "{.var}": ".value",