-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17502/
-----------------------------------------------------------
Review request for Sqoop.
Bugs: SQOOP-1273
https://issues.apache.org/jira/browse/SQOOP-1273
Repository: sqoop-trunk
Description
-------
I've tweaked the algorithm described on JIRA a bit to lower the probability of
collision:
* I've increased the time precision from milliseconds to nanoseconds.
* I've changed semantics of tableName to be general salt and rather then
passing null in query based import, I'm using query's hash code.
* Per Venkat's suggestion I've added JVM name that at least on linux uses
Process id and hostname. Not sure what exactly will be there on other platforms
such as Windows.
So with those addition to get collision, user would have to submit two Sqoop
jobs at the same millisecond for exactly the same table or query on different
machines that do have the same hostname and process id. Seems significantly
less likely :-)
Diffs
-----
src/java/org/apache/sqoop/tool/ImportTool.java
59279b6909f4a74e44385ba7a75321c7cf1eab13
src/java/org/apache/sqoop/util/AppendUtils.java
d7cd6c53054c01226346ee2ebe85043cc718323a
src/test/com/cloudera/sqoop/TestAppendUtils.java
ed66b4aeb58a7ec245126b077101d93544221cfc
Diff: https://reviews.apache.org/r/17502/diff/
Testing
-------
Added AppendUtils test that will exercise --query based import.
Thanks,
Jarek Cecho
