[
https://issues.apache.org/jira/browse/LUCENE-2570?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12892711#action_12892711
]
Shai Erera commented on LUCENE-2570:
------------------------------------
I don't think that removing the class is a good idea. Some methods like rmDir
can exist on the base test classes, however the class also contains large
arrays blockStarts/Ends and it will just clatter the base test class, I think.
What do you think?
bq. Another option, is if _TestUtil has methods that need a random, make it a
required argument?
I had thought about it, but that method is called from several places in the
code, and the randomness seems very insignificant. It just adds a random number
to the directory name the caller requested. If it is in order to prevent
collisions, then the code isn't safe because there is a chance the same number
will be generated twice ...
> Some improvements to _TestUtil and its usage
> --------------------------------------------
>
> Key: LUCENE-2570
> URL: https://issues.apache.org/jira/browse/LUCENE-2570
> Project: Lucene - Java
> Issue Type: Test
> Reporter: Shai Erera
> Assignee: Shai Erera
> Priority: Minor
> Fix For: 4.0
>
>
> I've started this issue because I've noticed that
> _TestUtil.getRandomMultiplier() is called from many loops' condition check,
> sometimes hundreds and thousands of times. Each time it does Integer.parseInt
> after calling System.getProperty. This really can become a constant IMO,
> either in LuceneTestCase(J4) or _TestUtil, as it's not expected to change
> while tests are running ...
> I then reviewed the class and spotted some more things that I think can be
> fixed/improved:
> # getTestCodec() can become a constant as well
> # arrayToString is marked deprecated. I've checked an no one calls them, so
> I'll delete them. This is a 4.0 code branch + a test-only class. No need to
> deprecate anything.
> # getTempDir calls new Random(), instead of newRandom() in LuceneTestCaseJ4,
> which means that if something fails, we won't know the random seed used ...
> #* In that regard, we might want to output all the classes that obtained a
> static seed in reportAdditionalFailures(), instead of just the class that ran
> the test.
> # rmDir(String) can be removed IMO, and leave only rmDir(File)
> # I suggest we include some recursion in rmDir(File) to handle the deletion
> of nested directories.
> #* Also, it does not check whether the dir deletion itself succeeds (but it
> does so for the files). This can bite us on Windows, if some test did not
> close things properly.
> I'll work out a patch.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
