Hi again, 2012/6/13 Gilles Sadowski <gil...@harfang.homelinux.org>: > On Wed, Jun 13, 2012 at 12:34:51PM +0200, Benedikt Ritter wrote: >> Hi, >> >> 2012/6/13 Gilles Sadowski <gil...@harfang.homelinux.org>: >> > Hi. >> > >> >> > >> >> > Is there a way for an application to depend on the classes defined in >> >> > the >> >> > "test" subtree of the code repository? >> >> >> >> If the component releases the test jar, then it should be possible. >> > >> > It seems that it's not released as a JAR (probably only in the zip/tar >> > and source archive). >> > >> >> >> >> But if the test code is more generally useful, perhaps it should be >> >> moved to the java subtree. >> > >> > I'd like to make it easy to use "PerfTestUtils" outside of the CM >> > development environment. >> > >> > Is it OK to move it to "main"? Before doing so, I also expected that more >> > people would have a look at it, and agree on the "methodology" (or suggest >> > ways to improve it) and API. >> > >> > When I move it, should it go to the top-level package "o.a.c.m"? Or maybe >> > into "util"? Or even a new "util/perf" subpackage (to separate utilities >> > used by the CM code from utilities used to test the code's performance? >> > >> >> I don't know, to me it feels rather strange to add a TestPerfUtil to >> the public API of a mathematical library. > > I agree. > [That's why I put the class where it is.] > >> OTOH TestPerfUtils looks to >> specific to be used outside the context of CM (it uses classes defined >> in CM). > > I don't understand that. It's a (mini-)benchmarking tool; it can be used to > time any code, not just CM routines. >
PerfTestUtils defines imports: import org.apache.commons.math3.stat.descriptive.SummaryStatistics; import org.apache.commons.math3.stat.descriptive.StatisticalSummary; import org.apache.commons.math3.exception.MathIllegalStateException; it can not be used without having CM on the classpath. That's what i meant ;-) >> I see two solutions: >> 1. if you need it just for one project, you could simply copy the >> class to that project in the test tree > > No. The purpose is to provide a tool to anyone who claims that "changing > X in the CM code will improve performance", so that the claim is backed by > actual numbers. > In this case one will have to check out the hole CM anyway, I guess? >> 2. if you need it for several projects and there are additional useful >> classes in the test tree that are concerned with tests in context of >> CM, a new component could be created that defines a dependency to CM. >> This dependency (commons-math-testutils ?!) could then be added to >> what ever project in the test scope. > > A clean solution, but I don't think that there is enough code to be worth > the burden of a separate project. > > A third solution is to indeed publish a JAR composed of the classes in the > "test" subtree of the repository, as hinted at by Sebb. > If it's OK, could someone please make the necessary changes (both for > producing official releases and for daily snapshots)? > I agree. regards, Benedikt > > Thanks, > Gilles > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > For additional commands, e-mail: dev-h...@commons.apache.org > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
