Re: parallel iterative deconvolution

Posted by dscho on Sep 10, 2013; 6:38pm
URL: http://imagej.273.s1.nabble.com/parallel-iterative-deconvolution-tp5004620p5004744.html

Hi Aryeh,

On Sat, 7 Sep 2013, Aryeh Weiss wrote:

> On 9/4/13 9:17 PM, Johannes Schindelin wrote:
> >
> > On Sun, 1 Sep 2013, Aryeh Weiss wrote:
> >
> > > java.lang.NoSuchMethodError:
> > > cern.colt.Sorting.parallelQuickSort([IIILcern/colt/function/tint/IntComparator;)V
> > >  at cern.colt.matrix.tfloat.algo.FloatSorting.runSort(Unknown Source)
> > >  at cern.colt.matrix.tfloat.algo.FloatSorting.sortIndex(Unknown Source)
> > >  at ... [more stuff that I will send if needed]
> > >
> > > This type of error was discussed in the past, most recently (which I
> > > found):
> > > https://list.nih.gov/cgi-bin/wa.exe?A2=ind1105&L=IMAGEJ&P=R35667&1=IMAGEJ&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4
> > >
> > > The problem then was outdated parallelcolt in the load path. So I
> > > purged my system of all of the old parallelcolt jars, and made sure
> > > I have only parallelcolt-0.9.4, which I believe is the latest.
> >
> > The problem is that parallelcolt is developed in a slightly
> > backwards-incompatible manner. That, combined with the fact that
> > parallel iterative convolution dependended on a different (and
> > incompatible) version of parallelcolt than superresolution prevented
> > the inclusion of those otherwise very interesting plugins into Fiji.
> >
> > In your case, it might also be one of those creative .jar files that
> > bundle other projects which leads to all kinds of unpleasant
> > surprises.
> >
> > To find out what .jar file contains the offending Sorting class, use
> > Fiji's 'Find Jar For Class' script via the command launcher and
> > provide the full class name: cern.colt.Sorting
> >
>
> I found the class in colocalization-pipeline and in colt-1.2.0.jar
>
> I uninstalled colocalization-pipeline and disabled (for now)
> colt-1.2.0.jar (I renamed it to colt-1.2.0.jar-dist).
>
> This allows parallel iterative deconvolution to run, but I do not think
> that it is a good long term solution, because the updater "wants" to
> restore colt, and I prefer to keep the standard distribution of
> automatically updated files.

I agree, but I can only repeat my suggestion to Piotr to make sure that
parallelcolt stays backwards-compatible and if that is not possible, at
least to make all of the plugins depending on parallelcolt compatible with
the newest version of the latter.

> Also, I dont know what (if anything) will break by disabling
> colt-1.2.0.jar.

Theoretically, colt should be completely superseded by parallelcolt. But
as I said, parallelcolt is not completely backwards-compatible to itself,
so I imagine that it might even have broken backwards-compatibility to
colt, too. Piotr is really the best person to ask: Piotr, any insights?

Ciao,
Johannes

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html