Merge pull request #173 from typesafehub/wip/havocp-out-of-range-bytes

Throw an exception if size-in-bytes values are out of Long range

HOCON.md

@@ -1265,6 +1265,13 @@ spec copies that. You can certainly find examples of mapping these
 to powers of ten, though. If you don't like ambiguity, don't use
 the single-letter abbreviations.
 
+Note: any value in zetta/zebi or yotta/yobi will overflow a 64-bit
+integer, and of course large-enough values in any of the units may
+overflow. Most real-world APIs and apps will not support byte
+counts that overflow a 64-bit integer. The huge units are provided
+just to be complete but probably aren't useful in practice. At
+least not in 2014.
+
 ### Config object merging and file merging
 
 It may be useful to offer a method to merge two objects. If such a

config/src/main/java/com/typesafe/config/impl/SimpleConfig.java

@@ -5,6 +5,8 @@ package com.typesafe.config.impl;
 
 import java.io.ObjectStreamException;
 import java.io.Serializable;
+import java.math.BigDecimal;
+import java.math.BigInteger;
 import java.util.AbstractMap;
 import java.util.ArrayList;
 import java.util.HashMap;
@@ -575,19 +577,13 @@ final class SimpleConfig implements Config, MergeableValue, Serializable {
         final String prefix;
         final int powerOf;
         final int power;
-        final long bytes;
+        final BigInteger bytes;
 
         MemoryUnit(String prefix, int powerOf, int power) {
             this.prefix = prefix;
             this.powerOf = powerOf;
             this.power = power;
-            int i = power;
-            long bytes = 1;
-            while (i > 0) {
-                bytes *= powerOf;
-                --i;
-            }
-            this.bytes = bytes;
+            this.bytes = BigInteger.valueOf(powerOf).pow(power);
         }
 
         private static Map<String, MemoryUnit> makeUnitsMap() {
@@ -667,13 +663,20 @@ final class SimpleConfig implements Config, MergeableValue, Serializable {
         }
 
         try {
+            BigInteger result;
             // if the string is purely digits, parse as an integer to avoid
             // possible precision loss; otherwise as a double.
             if (numberString.matches("[0-9]+")) {
-                return Long.parseLong(numberString) * units.bytes;
+                result = units.bytes.multiply(new BigInteger(numberString));
             } else {
-                return (long) (Double.parseDouble(numberString) * units.bytes);
+                BigDecimal resultDecimal = (new BigDecimal(units.bytes)).multiply(new BigDecimal(numberString));
+                result = resultDecimal.toBigInteger();
             }
+            if (result.bitLength() < 64)
+                return result.longValue();
+            else
+                throw new ConfigException.BadValue(originForException, pathForException,
+                        "size-in-bytes value is out of range for a 64-bit long: '" + input + "'");
         } catch (NumberFormatException e) {
             throw new ConfigException.BadValue(originForException, pathForException,
                     "Could not parse size-in-bytes number '" + numberString + "'");

config/src/test/scala/com/typesafe/config/impl/UnitParserTest.scala

@@ -50,16 +50,19 @@ class UnitParserTest extends TestUtils {
 
         val conf = parseConfig("foo = 1d")
         assertEquals("could get 1d from conf as days",
-                     1L, conf.getDuration("foo", TimeUnit.DAYS))
+            1L, conf.getDuration("foo", TimeUnit.DAYS))
         assertEquals("could get 1d from conf as nanos",
-                     dayInNanos, conf.getNanoseconds("foo"))
+            dayInNanos, conf.getNanoseconds("foo"))
         assertEquals("could get 1d from conf as millis",
-                     TimeUnit.DAYS.toMillis(1), conf.getMilliseconds("foo"))
+            TimeUnit.DAYS.toMillis(1), conf.getMilliseconds("foo"))
     }
 
     @Test
     def parseMemorySizeInBytes(): Unit = {
-        def parseMem(s: String) = SimpleConfig.parseBytes(s, fakeOrigin(), "test")
+        def parseMem(s: String): Long = SimpleConfig.parseBytes(s, fakeOrigin(), "test")
+
+        assertEquals(Long.MaxValue, parseMem(s"${Long.MaxValue} bytes"))
+        assertEquals(Long.MinValue, parseMem(s"${Long.MinValue} bytes"))
 
         val oneMebis = List("1048576", "1048576b", "1048576bytes", "1048576byte",
             "1048576  b", "1048576  bytes",
@@ -88,7 +91,7 @@ class UnitParserTest extends TestUtils {
         }
 
         var result = 1024L * 1024 * 1024
-        for (unit <- Seq("tebi", "pebi", "exbi", "zebi", "yobi")) {
+        for (unit <- Seq("tebi", "pebi", "exbi")) {
             val first = unit.substring(0, 1).toUpperCase()
             result = result * 1024;
             assertEquals(result, parseMem("1" + first))
@@ -99,7 +102,7 @@ class UnitParserTest extends TestUtils {
         }
 
         result = 1000L * 1000 * 1000
-        for (unit <- Seq("tera", "peta", "exa", "zetta", "yotta")) {
+        for (unit <- Seq("tera", "peta", "exa")) {
             val first = unit.substring(0, 1).toUpperCase()
             result = result * 1000;
             assertEquals(result, parseMem("1" + first + "B"))
@@ -119,4 +122,45 @@ class UnitParserTest extends TestUtils {
         }
         assertTrue(e2.getMessage().contains("size-in-bytes number"))
     }
+
+    // later on we'll want to check this with BigInteger version of getBytes
+    @Test
+    def parseHugeMemorySizes(): Unit = {
+        def parseMem(s: String): Long = SimpleConfig.parseBytes(s, fakeOrigin(), "test")
+        def assertOutOfRange(s: String) = {
+            val fail = intercept[ConfigException.BadValue] {
+                parseMem(s)
+            }
+            assertTrue("number was too big", fail.getMessage.contains("out of range"))
+        }
+
+        import java.math.BigInteger
+        assertOutOfRange(s"${BigInteger.valueOf(Long.MaxValue).add(BigInteger.valueOf(1)).toString} bytes")
+        assertOutOfRange(s"${BigInteger.valueOf(Long.MinValue).subtract(BigInteger.valueOf(1)).toString} bytes")
+
+        var result = 1024L * 1024 * 1024
+        for (unit <- Seq("zebi", "yobi")) {
+            val first = unit.substring(0, 1).toUpperCase()
+            assertOutOfRange("1" + first)
+            assertOutOfRange("1" + first + "i")
+            assertOutOfRange("1" + first + "iB")
+            assertOutOfRange("1" + unit + "byte")
+            assertOutOfRange("1" + unit + "bytes")
+            assertOutOfRange("1.1" + first)
+            assertOutOfRange("-1" + first)
+        }
+
+        result = 1000L * 1000 * 1000
+        for (unit <- Seq("zetta", "yotta")) {
+            val first = unit.substring(0, 1).toUpperCase()
+            assertOutOfRange("1" + first + "B")
+            assertOutOfRange("1" + unit + "byte")
+            assertOutOfRange("1" + unit + "bytes")
+            assertOutOfRange("1.1" + first + "B")
+            assertOutOfRange("-1" + first + "B")
+        }
+
+        assertOutOfRange("1000 exabytes")
+        assertOutOfRange("10000000 petabytes")
+    }
 }