You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@zookeeper.apache.org by "Charlie Helin (JIRA)" <ji...@apache.org> on 2014/11/07 18:13:33 UTC
[jira] [Commented] (ZOOKEEPER-1716) jute/Utils.fromCSVBuffer cannot
parse data returnd by toCSVBuffer
[ https://issues.apache.org/jira/browse/ZOOKEEPER-1716?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14202280#comment-14202280 ]
Charlie Helin commented on ZOOKEEPER-1716:
------------------------------------------
Why not just?
{code}
static String toCSVBuffer(byte array[]) {
if (array == null || array.length == 0) {
return "";
}
StringBuilder builder = new StringBuilder(array.length * 2 + 1);
builder.append('#');
for (byte item : array) {
builder.append(String.format("%02x", item));
}
return builder.toString();
}
{code}
(As a side note, why does Utils implements it's own Arrays.equals?, The vm will intrinsitize the implementation rendering meaning that it will utilize SSE cmp if available on the platform etc, doing single byte compare will be orders of magnitude slower)
> jute/Utils.fromCSVBuffer cannot parse data returnd by toCSVBuffer
> ------------------------------------------------------------------
>
> Key: ZOOKEEPER-1716
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1716
> Project: ZooKeeper
> Issue Type: Bug
> Components: jute
> Affects Versions: 3.4.5
> Reporter: Robert Joseph Evans
>
> I was trying to use org.apache.zookeeper.server.LogFormatter to analyze the access pattern of a particular application. As part of this I wanted to get the size of the data that was being written into ZK.
> I ran into an issue where in some cases the hex data had an odd length. I looked into it and found that the buffer is being written out using Integer.toHexString(barr[idx])
> Looking at the javadoce for toHexString it indicates that it does not pad the bits at all, and will output the twos compliment of the number if it is negative. I then looked at how the data was being parsed and it assumed that every byte consisted of exactly two characters, which is not true.
> {code}
> Utils.toCSVBuffer(new byte[] {0xff}) returns "#ffffffff"
> Utils.toCSVBuffer(new byte[] {0x01}) returns "#1"
> If I combine those
> Utils.fromCSVBuffer(Utils.toCSVBuffer(new byte[] {0xff, 0x01, 0xff})) will return {0xff, 0xff, 0xff, 0xff, 0x1f, 0xff, 0xff, 0xff}
> {code}
> I think what we want is something like
> {code}
> static final char[] NIBBLE_TO_HEX = {
> '0', '1', '2', '3', '4', '5', '6', '7',
> '8', '9', 'a', 'b', 'c', 'd', 'e', 'f'
> };
> static String toCSVBuffer(byte barr[]) {
> if (barr == null || barr.length == 0) {
> return "";
> }
> StringBuilder sb = new StringBuilder(barr.length + 1);
> sb.append('#');
> for(int idx = 0; idx < barr.length; idx++) {
> byte b = barr[idx];
> sb.append(NIBBLE_TO_HEX[b&0x0f]);
> sb.append(NIBBLE_TO_HEX[(b&0xf0)>>4]);
> }
> return sb.toString();
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)