myui / btree4j Goto Github PK

Hi,

This is pretty easy to produce: Modify the BtreeIndexTest.testBtreeIndex test. Increase the repeat in the invokeTest method (multiply by 10)
final int repeat = 10000000;

=>

java.lang.IllegalStateException: range scan failed... bug?

at btree4j.BTree.scanRange(BTree.java:365)
at btree4j.BTree.search(BTree.java:333)
at btree4j.BTreeIndex.search(BTreeIndex.java:136)
at btree4j.BTreeIndexTest.invokeTest(BTreeIndexTest.java:181)
at btree4j.BTreeIndexTest.testBTreeIndex(BTreeIndexTest.java:147)
...

By the way, thanks for the great work 👍.

why need convert to byte ?

since 129 convert to -127. how to identify 129 and -127 ?

• Setup travis CI.

Hey MyUi,

less an issue, but more a question: How ACID compliant are the methods on the tree (especially atomicity)?

If for example the JVM shuts down (planned or as a fault) during the following method call

btree.addValue(k, v);

would then there be a risk that the whole tree might (a) or parts of the tree (b) might be corrupt or is there no such risk (c)? Any hints regarding the ACID level compliance in the description would be greatly appreciated.

Kind regards,
Borbarad13

    for (int i = 0; i < 1000; i++) {
        Value k = new Value(i);
        btree.addValue(k, i);

}

    IndexConditionBW bw = new IndexConditionBW(new Value(100), new Value(128));

btree.search(bw, callback) will throw exception

            throw new IllegalStateException("range scan failed... bug?");

Why don't you remove unnecessary nodes when they become under-occupied? How are you going to deal with a large number of useless branch nodes in tree?

I have tried the following code:

      BTreeIndex btree = new BTreeIndex(indexFile, false);
      btree.init(false);

      Random rand = new Random();
      for (int i = 0; i < 10000000; i++) {
        long nt = System.nanoTime(), val = rand.nextLong();
        Value key = new Value(String.valueOf(nt) + val);
        Value value = new Value(val);
        if (i % 10000 == 0) {
          System.out.println(i + " " + key + " " + btree.addValue(key, value));
        } else {
          btree.addValue(key, value);
        }
      }

Then in another method I open index file and try to search for keys:

       BTreeIndex btree = new BTreeIndex(indexFile, false);
       Value val = btree.getValue(new Value("205388120935395-497341292157541218"));
       System.out.println(Primitives.getLong(val.getData()));

On first few million, it works, but then throws NullPointerException inside Value class. Am I doing something wrong?

• fix pom.xml
• sync to Sonatype

BTreeNode.write and read operation should be atomic for a better Durability (sudden shutdown etc).

To do that, undo log is required.

Since log4j is deprecated, use SLF4j or java.util.logging for the logging scheme.

if there are duplicate keys then btree.getValue(dupKey) only gives one value, how to get all?

Change the package structure as

• top-level
  • core
  • memcache

Based on my past work
https://github.com/myui/gridool/tree/master/gridool/src/memcached/src/gridool/memcached

or
https://github.com/netty/netty/tree/master/codec-memcache/src/main/java/io/netty/handler/codec/memcache
https://github.com/couchbase/couchbase-jvm-core/tree/master/src/main/java/com/couchbase/client/deps/io/netty/handler/codec/memcache