LOG4J2-3672 Avoid invoking DateFormatSymbols.getZoneStrings() in FastDateParser

Author: tristantarrant

log4j-core/src/main/java/org/apache/logging/log4j/core/util/datetime/FastDateParser.java

@@ -19,7 +19,6 @@
 import java.io.IOException;
 import java.io.ObjectInputStream;
 import java.io.Serializable;
-import java.text.DateFormatSymbols;
 import java.text.ParseException;
 import java.text.ParsePosition;
 import java.util.ArrayList;
@@ -40,6 +39,7 @@
 import java.util.regex.Pattern;
 
 import org.apache.logging.log4j.core.util.Integers;
+import org.apache.logging.log4j.util.PropertiesUtil;
 
 import static org.apache.logging.log4j.util.Strings.toRootUpperCase;
 
@@ -79,6 +79,11 @@
  * @see FastDatePrinter
  */
 public class FastDateParser implements DateParser, Serializable {
+    /**
+     * Enables the parsing of "friendly" timezone names. Since building a list of all available names incurs an additional heap cost (~3MB on x86_64),
+     * in internal JDK structures which cannot be garbage collected, it is disabled by default.
+     */
+    public static final String LOG4J2_ENABLE_FRIENDLY_TIMEZONE_NAMES = "log4j2.enableFriendlyTimezoneNames";
 
     /**
      * Required for serialization support.
@@ -102,7 +107,7 @@ public class FastDateParser implements DateParser, Serializable {
     // comparator used to sort regex alternatives
     // alternatives should be ordered longer first, and shorter last. ('february' before 'feb')
     // all entries must be lowercase by locale.
-    private static final Comparator<String> LONGER_FIRST_LOWERCASE = (left, right) -> right.compareTo(left);
+    private static final Comparator<String> LONGER_FIRST_LOWERCASE = Comparator.reverseOrder();
 
     /**
      * <p>Constructs a new FastDateParser.</p>
@@ -551,7 +556,7 @@ boolean parse(final FastDateParser parser, final Calendar calendar, final String
 
     /**
      * Obtain a Strategy given a field from a SimpleDateFormat pattern
-     * @param formatField A sub-sequence of the SimpleDateFormat pattern
+     * @param f A sub-sequence of the SimpleDateFormat pattern
      * @param definingCalendar The calendar to obtain the short and long values
      * @return The Strategy that will handle parsing for the field
      */
@@ -719,7 +724,7 @@ private static class CaseInsensitiveTextStrategy extends PatternStrategy {
         @Override
         void setCalendar(final FastDateParser parser, final Calendar cal, final String value) {
             final Integer iVal = lKeyValues.get(value.toLowerCase(locale));
-            cal.set(field, iVal.intValue());
+            cal.set(field, iVal);
         }
     }
 
@@ -843,44 +848,31 @@ private static class TzInfo {
             final StringBuilder sb = new StringBuilder();
             sb.append("((?iu)" + RFC_822_TIME_ZONE + "|" + GMT_OPTION );
 
-            final Set<String> sorted = new TreeSet<>(LONGER_FIRST_LOWERCASE);
+            if (PropertiesUtil.getProperties().getBooleanProperty(LOG4J2_ENABLE_FRIENDLY_TIMEZONE_NAMES)) {
+                final Set<String> sorted = new TreeSet<>(LONGER_FIRST_LOWERCASE);
 
-            final String[][] zones = DateFormatSymbols.getInstance(locale).getZoneStrings();
-            for (final String[] zoneNames : zones) {
-                // offset 0 is the time zone ID and is not localized
-                final String tzId = zoneNames[ID];
-                if (tzId.equalsIgnoreCase("GMT")) {
-                    continue;
-                }
-                final TimeZone tz = TimeZone.getTimeZone(tzId);
-                // offset 1 is long standard name
-                // offset 2 is short standard name
-                final TzInfo standard = new TzInfo(tz, false);
-                TzInfo tzInfo = standard;
-                for (int i = 1; i < zoneNames.length; ++i) {
-                    switch (i) {
-                    case 3: // offset 3 is long daylight savings (or summertime) name
-                            // offset 4 is the short summertime name
-                        tzInfo = new TzInfo(tz, true);
-                        break;
-                    case 5: // offset 5 starts additional names, probably standard time
-                        tzInfo = standard;
-                        break;
+                for (final String tzId : TimeZone.getAvailableIDs()) {
+                    if (tzId.equalsIgnoreCase("GMT")) {
+                        continue;
                     }
-                    if (zoneNames[i] != null) {
-                        final String key = zoneNames[i].toLowerCase(locale);
-                        // ignore the data associated with duplicates supplied in
-                        // the additional names
-                        if (sorted.add(key)) {
-                            tzNames.put(key, tzInfo);
+                    final TimeZone tz = TimeZone.getTimeZone(tzId);
+                    final TzInfo standard = new TzInfo(tz, false);
+                    final TzInfo dst = new TzInfo(tz, true);
+
+                    for(boolean daylight : new Boolean[] {false, true}) {
+                        for(int style : new Integer[]{TimeZone.SHORT, TimeZone.LONG}) {
+                            String name = tz.getDisplayName(daylight, style, locale).toLowerCase(locale);
+                            if (sorted.add(name)) {
+                                tzNames.put(name, daylight ? dst : standard);
+                            }
                         }
                     }
                 }
-            }
-            // order the regex alternatives with longer strings first, greedy
-            // match will ensure longest string will be consumed
-            for (final String zoneName : sorted) {
-                simpleQuote(sb.append('|'), zoneName);
+                // order the regex alternatives with longer strings first, greedy
+                // match will ensure longest string will be consumed
+                for (final String zoneName : sorted) {
+                    simpleQuote(sb.append('|'), zoneName);
+                }
             }
             sb.append(")");
             createPattern(sb);

src/site/xdoc/manual/configuration.xml.vm

@@ -2978,12 +2978,21 @@ public class AwesomeTest {
       that advertises the same language(s) in order for scripting to be enabled. If no languages are specified, which is
       the default, the ScriptManager will not be installed.</td>
   </tr>
-    <tr>
+  <tr>
     <td><a name="log4j2.disableCloudConfigLoggingSystem"/>log4j2.disableCloudConfigLoggingSystem</td>
     <td></td>
     <td></td>
     <td>Disables the usage of the Spring Boot <tt>Log4j2CloudConfigLoggingSystem</tt>. Defaults to <tt>false</tt>.</td>
   </tr>
+  <tr>
+      <td><a name="log4j2.enableFriendlyTimezoneNames"/>log4j2.enableFriendlyTimezoneNames</td>
+      <td></td>
+      <td></td>
+      <td>Enables the parsing of "friendly" timezone names in the <a href="layouts.html#PatternDate">Pattern Layout
+        date</a>. Since building a list of all available names incurs an additional heap cost (~3MB on x86_64) in
+        internal JDK structures which cannot be garbage collected, it is disabled by default.
+        Defaults to <tt>false</tt>.</td>
+    </tr>
 </table>
 
         </subsection>