[ImageJ-devel] [ImgLib2] moved revised DoG detection to algorithms-core.

[Johannes Schindelin]

Hi Tobias,

On Thu, 19 Sep 2013, Tobias Pietzsch wrote:

> I would like to put a revised version of the DoG detection
> implementation into imglib2-algorithms core.
>
> I pushed it in a branch "dog" https://github.com/imagej/imglib/commits/dog.

IMHO https://github.com/imagej/imglib/compare/dog shows it better ;-)

> I would have merged to master, but I'm having slight reservations
> because this requires the addition of two dependencies.

Actually, I am happy to see that the reluctance to add dependencies has
waned a bit ;-)

> First a dependency on imglib2-meta, this is for convenience to use
> calibration of ImgPlus and could be possibly removed.

Note that imglib2-meta *will* depend on scijava-common soon, to avoid
wheel reinvention. That is not a big problem IMO because scijava-common is
developed very carefully, avoiding at high maintenance cost to introduce
backwards-incompatibilities or other possibly breaking changes. So I'm all
for it.

> Second is a dependency on JAMA.

Steffi identified a hard crash (IIRC) with some JVMs that we fixed in
Fiji. We used to call that "version 1.0.2" but that is wrong, of course.
We should probably call it "1.0.3-fiji1" or some such; question is whether
we want to care about those JVM crashes at all? After all, those JVMs have
been superseded by later JVMs that apparently do not expose those crashes.

> What is your opinion? Is it okay to add these dependencies? Or should I
> rather move this stuff to some other place? If so, where?

As I said, I like to avoid reinventing the wheel and add dependencies
instead. The whole point of using Maven is to manage dependencies, after
all.

So: thumbs up from my side!

Ciao,
Dscho