mihaibudiu commented on code in PR #3481:
URL: https://github.com/apache/calcite/pull/3481#discussion_r1375644960


##########
linq4j/src/main/java/org/apache/calcite/linq4j/tree/Primitive.java:
##########
@@ -367,28 +369,67 @@ public static List<Double> asList(double[] elements) {
   }
 
   /**
-   * Converts a number into a value of the type
-   * specified by this primitive.
+   * Check if a value after rounding falls within a specified range.
+   *
+   * @param value  Value to compare.
+   * @param min    Minimum value allowed.
+   * @param max    Maximum value allowed.
+   */
+  static void checkRoundedRange(Number value, double min, double max) {
+    double dbl = value.doubleValue();
+    // The equivalent of DOWN rounding for BigDecimal
+    dbl = dbl > 0 ? Math.floor(dbl) : Math.ceil(dbl);
+    if (dbl < min || dbl > max) {
+      throw new ArithmeticException("Value " + value + " out of range");
+    }
+  }
+
+  /**
+   * Converts a number into a value of the type specified by this primitive
+   * using the SQL CAST rules.  If the value conversion causes loss of 
significant digits,
+   * an exception is thrown.
    *
    * @param value  Value to convert.
-   * @return       The converted value, or null if the
-   *               conversion cannot be performed.
+   * @return       The converted value, or null if the type of the result is 
not a number.
    */
   public @Nullable Object numberValue(Number value) {
     switch (this) {
     case BYTE:
+      checkRoundedRange(value, Byte.MIN_VALUE, Byte.MAX_VALUE);
       return value.byteValue();
     case CHAR:
+      // No overflow checks for char values

Review Comment:
   This is a good question. You see, there is actually no written specification 
of what should happen in this case, that's why I filed 
https://issues.apache.org/jira/projects/CALCITE/issues/CALCITE-6073.
   I checked Postgres and it doesn't actually report an error when casting to 
CHAR, it look like it only truncates the value.
   But in the absence of a spec for what CASTs are supposed to do there is no 
good answer.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to