accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From ctubbsii <...@git.apache.org>
Subject [GitHub] accumulo pull request #292: ACCUMULO-4681 Created Table & Namespace WeakRefe...
Date Sat, 12 Aug 2017 19:34:40 GMT
Github user ctubbsii commented on a diff in the pull request:

    https://github.com/apache/accumulo/pull/292#discussion_r132825236
  
    --- Diff: core/src/main/java/org/apache/accumulo/core/client/impl/Table.java ---
    @@ -16,24 +16,59 @@
      */
     package org.apache.accumulo.core.client.impl;
     
    +import java.lang.ref.WeakReference;
    +import java.util.WeakHashMap;
    +
     import org.apache.accumulo.core.client.Instance;
     
     public class Table {
     
       /**
        * Object representing an internal table ID. This class was created to help with type
safety. For help obtaining the value of a table ID from Zookeeper, see
        * {@link Tables#getTableId(Instance, String)}
    +   *
    +   * Uses an internal WeakHashMap and private constructor for storing a WeakReference
of every Table.ID. Therefore, a Table.ID can't be instantiated outside
    +   * this class and is accessed by calling Table.ID.{@link #of(String)}.
        */
       public static class ID extends AbstractId {
         private static final long serialVersionUID = 7399913185860577809L;
    +    static final WeakHashMap<String,WeakReference<Table.ID>> tableIds = new
WeakHashMap<>();
     
    -    public static final ID METADATA = new ID("!0");
    -    public static final ID REPLICATION = new ID("+rep");
    -    public static final ID ROOT = new ID("+r");
    +    public static final ID METADATA = of("!0");
    +    public static final ID REPLICATION = of("+rep");
    +    public static final ID ROOT = of("+r");
     
    -    public ID(final String canonical) {
    +    private ID(final String canonical) {
           super(canonical);
         }
    +
    +    /**
    +     * Get a Table.ID object for the provided canonical string.
    +     *
    +     * @param canonical
    +     *          table ID string
    +     * @return Table.ID object
    +     */
    +    public static Table.ID of(final String canonical) {
    +      return dedupeTableId(canonical);
    +    }
    +
    +    private static Table.ID dedupeTableId(String tableIdString) {
    +      Table.ID tableId;
    +      synchronized (tableIds) {
    +        WeakReference<Table.ID> tableIdRef = tableIds.get(tableIdString);
    +        if (tableIdRef != null) {
    +          tableId = tableIdRef.get();
    +          if (tableId != null) {
    +            return tableId;
    +          }
    +        }
    +
    +        tableId = new ID(tableIdString);
    +        tableIds.put(tableIdString, new WeakReference<>(tableId));
    --- End diff --
    
    I was thinking through the logic of this put method (and it's "compute" alternatives),
and realized that there's no way to get a strong reference to the key itself from the map.
You can only do comparisons. So, there's a chance that when you do a "put" here, it will reuse
the old key, but the new value (in the case where the map entry exists, but the WeakReference
is empty). The result is that the entry could disappear from the map even though the object
in the WeakReference is still being used. In this case, we lose deduplication. There does
not appear to be a solution to this (short of writing our own map class, or iterating over
the keyset and comparing keys ourselves), and our deduplication is "best effort", so it's
not a big deal... but it's worth keeping in mind, so we don't even try to use this paradigm
for guaranteed deduplication.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message