This looks like the best argument I've yet seen for breaking up the PDL distribution -- perhaps making the stuff in "Basic" the actual PDL module, and putting Graphics, IO, and Lib into their own module trees.
On Oct 30, 2009, at 9:27 AM, David Mertens wrote: > Ilya Z, author of Numeric::LL_Array, is a long time Perl developer, > a rather brilliant one, and currently teaches/researches at the Cal > Math dept. I feel a bit vindicated in that I was not the only one > experiencing difficulty installing PDL. If Ilya felt it was > difficult, well, then, it must have been difficult. Either that, or > he had some other itch to scratch which caused him to develop his > module instead of sticking it out with PDL. > > See, this is what I find so surprising. I don't think it's really > all that difficult to get PDL to install. I've put instructions on > use.perl and, over the course of writing the instructions I realized > that it's really pretty simple if you know what you're doing or have > somebody to hold your hand. A Gentoo-quality wiki for installing > PDL seems like an appropriate stop-gap measure until we can get > better external dependency handling (in which case a cpan install > would truly be one-click). > > I don't think the installation was too dificult for him. Rather, I > think he has a philosophical gripe about it's apparent inability to > install correctly. Version 2.4.2 has an attrocious pass rate at > CPAN Testers. If PDL is supposed to be THE Perl numerical data > module, it should easily install on any Perl system and pass all its > tests. I believe that when he saw that, as well as all of the > dependencies, he decided to start fresh with his own solution. > > Ilya and I have discussed his failed attempts to get PDL installed, > and he made a very interesting observation. PDL 2.4.5 was released > on October 24 and Numeric::LL_Array v 0.12 was released on October > 26. Both distributions have passed on all the machines that have > tested them on CPAN Testers, but his distribution has run on 57 > machines while PDL has only been tested on four machines. We both > thought tthat these tests were automated, so why does > Numeric::LL_Array have over 14 times as many reported tests? My > hunch is that the Perl OpenGL module doesn't install on a lot of > machines so PDL can't even be attempted because its dependencies > cannot be satisfied. I conjecture that if PDL's dependencies can't > be met that the installation silently fails. But I'm not sure about > how all that stuff is reported. > > Ilya's take-away point is that authors of other modules cannot rely > on listing PDL as a dependency because the chances of a successful > automated install of PDL are low. Splitting up the current PDL into > a collection of modules, and especially having a lean core that > focused only on the data manipulatioin and left out the file > handling and visualization, would solve this problem. (As far as I > know, Numeric::LL_Array may simply be Ilya's vehicle for motivating > us to actually do that partitioning... and an opportunity for him to > have fun with XS code.) > > David > _______________________________________________ > Perldl mailing list > [email protected] > http://mailman.jach.hawaii.edu/mailman/listinfo/perldl _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
