这个程序,使用net.openhft:chronicle:3.5.3,说明了我对Chronicle的以下问题。
1 Vanilla Chronicle.size()返回疯狂值
2 ExcerptTailer.toStart不适用于Indexed Chronicle
3 ExcerptTailer.index(-1)不适用于Vanilla Chronicle
import net.openhft.chronicle.Chronicle;
import net.openhft.chronicle.ChronicleQueueBuilder;
import net.openhft.chronicle.ExcerptAppender;
import net.openhft.chronicle.ExcerptTailer;
import org.apache.commons.io.FileUtils;
import java.io.File;
/*
* These functions demonstrate a number of issues I've come across.
* */
public class Problems {
public static void main(String[] a) throws Exception {
sizeAndLastIndexWrittenProblem();
//indexedToStartDoesntWork();
//vanillaToIndexDoesntWork();
}
static void sizeAndLastIndexWrittenProblem() throws Exception {
File f = new File("cron");
FileUtils.deleteDirectory(f);
Chronicle chronicle = ChronicleQueueBuilder.vanilla(new File(f, "chronicle")).build();
ExcerptAppender ap = chronicle.createAppender();
// vanilla provides sensible values first of all
Assert(chronicle.size() == 0);
Assert(chronicle.lastWrittenIndex() == -1);
ap.startExcerpt();
ap.writeInt(1);
ap.finish();
// but after appending to a vanilla then these lines print insane values eg
// size = 18410222695481345
// lastWrittenIndex = 18410222695481344
System.out.println("size = " + chronicle.size());
System.out.println("lastWrittenIndex = " + chronicle.lastWrittenIndex());
Assert(chronicle.size() == 1); // FAILS HERE
Assert(chronicle.lastWrittenIndex() == 0);
// note that the same code using an indexed chronicla passes ok
}
static void indexedToStartDoesntWork() throws Exception {
File f = new File("cron");
FileUtils.deleteDirectory(f);
Chronicle chronicle = ChronicleQueueBuilder.indexed(new File(f,"chronicle")).build();
ExcerptAppender ap = chronicle.createAppender();
ExcerptTailer t = chronicle.createTailer();
ap.startExcerpt();
ap.writeInt(1);
ap.finish();
Assert(t.nextIndex()); // ok
Assert(t.readInt() == 1); // ok
t.finish();
ap.startExcerpt();
ap.writeInt(2);
ap.finish();
Assert(t.nextIndex()); // ok
Assert(t.readInt() == 2); // ok
t.finish();
/*
On an indexed chronicle toStart does not rewind to start so the next read fails.
I found that if using indexed then one must use index(-1) not toStart
*/
t.toStart(); // DOESN'T REWIND US TO START
Assert(t.nextIndex()); // FAILS HERE IF USING toStart BUT OK IF USING index(-1)
Assert(t.readInt() == 1);
}
static void vanillaToIndexDoesntWork() throws Exception {
File f = new File("cron");
FileUtils.deleteDirectory(f);
Chronicle chronicle = ChronicleQueueBuilder.vanilla(new File(f,"chronicle")).build();
ExcerptAppender ap = chronicle.createAppender();
ExcerptTailer t = chronicle.createTailer();
ap.startExcerpt();
ap.writeInt(1);
ap.finish();
Assert(t.nextIndex()); // ok
Assert(t.readInt() == 1); // ok
t.finish();
ap.startExcerpt();
ap.writeInt(2);
ap.finish();
Assert(t.nextIndex()); // ok
Assert(t.readInt() == 2); // ok
t.finish();
/*
On an vanilla chronicle index(-1) does not rewind to start so the next read fails.
I found that if using vanilla then one must use toStart not index(-1)
*/
t.index(-1); // DOESN'T REWIND US TO START
Assert(t.nextIndex()); // FAILS HERE IF USING index(-1) BUT OK IF USING toStart()
Assert(t.readInt() == 1);
}
static void Assert(boolean b) {
if (!b) throw new RuntimeException("bang");
}
}
答案 0 :(得分:0)
正如编年史小组所述,toStart问题看起来像一个错误,我需要检查。
如果你在GitHub上用失败的测试打开一个问题会很好。欢迎PR。
关于VanillaChronicle中的疯狂值请注意,索引不像IndexedChronicle那样工作,因为vanilla索引还包含有关您所处的循环的信息,编写摘录的线程等,因此您看到的值是正确的。