On 12/22/2011 10:43 PM, Puneet Kishor wrote:
On Dec 22, 2011, at 9:29 PM, Maggie X wrote:
Thank you for trying it out. Glad it worked! But I'm not sure why
just printing out the numbers would make it work. Was there
anything else that you have tried to make it work?
No... just added the `print` statements. That was it. Maybe the
`print` output resolved to `true` and that was it. You might want to
alter that test to account for such things.
Many thanks for your help. Now I can have a lot of fun wasting time
trying to redo stuff in PDL::Stats that I can already do in
Postgres/R ;-)
On a tangential note -- seriously, it would be a great boon if PDL
could ingest results of DBI queries as piddles, kinda like
(theoretical code ahead)
$sth->execute; my $piddle = $sth->fetchrow_arrayref; my $result =
$piddle->do_magic; return $result->to_json;
Two things: it appears that your email app has problems with
correctly encoding line endings (it seems to be a mac only
issue), and I seem to recall a module for reading database
info as piddles. Try checking the ones in the list of
"other" PDL modules I posted in the past week. Sorry, I
can't recall and I'm not using DBI stuff and the moment
myself....
Cheers,
Chris
piano, piano...
Best, Maggie
On Thu, Dec 22, 2011 at 10:13 PM, Puneet
Kishor<[email protected]> wrote:
Here ya go... I did as you said, and now it says all tests
passed. My guess is that the darn thing would work well if I just
installed it without testing it, but `cpanm` stops as soon as the
first test croaks.
punkish@mumbai ~/Projects/PDL-Stats-0.5.5$make test
PERL_DL_NONLAZY=1 /opt/local/bin/perl "-MExtUtils::Command::MM"
"-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/stats_basic.t ..... ok t/stats_glm.t ....... ok
t/stats_kmeans.t .... 1/18 R2 0.786191536748329 centroid [ [
1.5 5.5] [ 8.5 12] [15.5 19] ]
cluster [ [1 1 1 1 0 0 0] [0 0 0 0 1 1 1] ]
ms [ [ 1.25 0.25] [ 1.25 0.66666667] [ 1.25
0.66666667] ]
n [4 3] -6.66666570836583e-08R2 0.786191536748329 centroid
[ [ 1.5 5.5] [ 8.5 12] [15.5 19] ]
cluster [ [1 1 1 1 0 0 0] [0 0 0 0 1 1 1] ]
ms [ [ 1.25 0.25] [ 1.25 0.66666667] [ 1.25
0.66666667] ]
n [4 3] t/stats_kmeans.t .... ok t/stats_ols_rptd.t .. ok
t/stats_ts.t ........ ok All tests successful. Files=5,
Tests=147, 1 wallclock secs ( 0.03 usr 0.01 sys + 0.93 cusr
0.03 csys = 1.00 CPU) Result: PASS No tests defined for
PDL::Stats::Basic extension. PERL_DL_NONLAZY=1
/opt/local/bin/perl "-MExtUtils::Command::MM"
"-e" "test_harness(0, '../blib/lib', '../blib/arch')" t/*.t
t/stats_distr.t .. ok All tests successful. Files=1, Tests=43, 0
wallclock secs ( 0.02 usr 0.00 sys + 0.13 cusr 0.00 csys =
0.15 CPU) Result: PASS No tests defined for PDL::Stats::GLM
extension. No tests defined for PDL::Stats::Kmeans extension.
PERL_DL_NONLAZY=1 /opt/local/bin/perl "-MExtUtils::Command::MM"
"-e" "test_harness(0, '../blib/lib', '../blib/arch')" t/*.t
t/cdf.t .. ok All tests successful. Files=1, Tests=4, 0
wallclock secs ( 0.02 usr 0.00 sys + 0.06 cusr 0.00 csys =
0.08 CPU) Result: PASS No tests defined for PDL::Stats::TS
extension. punkish@mumbai ~/Projects/PDL-Stats-0.5.5$
On Dec 22, 2011, at 9:08 PM, Maggie X wrote:
I've seen cases where the test fails simply because of the
precision
differ
on different platforms. That's why I wanted to see what you get
when you
do
print $m{ms}->sumover(). But I guess make test doesn't actually
print out the values to stdout... How about this?
Add
print STDERR t_kmeans_bad();
168> is(tapprox( t_kmeans_bad(), 0 ), 1); 169> sub
t_kmeans_bad { 170> my $data = sequence 7, 3; 171>
$data = $data->setbadat(4,0); 172> my %m =
$data->kmeans({NCLUS=>2, NTRY=>10, V=>0});
print STDERR "$_\t$m{$_}\n" for (sort keys %m);
173> return sum( $m{ms}->sumover - pdl qw( 1.5 1.9166667
1.9166667
) );
174> }
Best, Maggie
On Thu, Dec 22, 2011 at 9:06 PM, Puneet
Kishor<[email protected]>
wrote:
I should have added my congratulations and gratitude for
taking 2.4.9
to a
point that it builds and installs beautifully, without any
heartburn.
Perhaps because most of the dependencies got installed when I
installed 2.4.6 via MacPorts.
I am doing a lot of stuff using R, so am curious to try out
PDL::Stats. For now, I am using R compiled inside Postgres
using PL/R... what a wonderful experience to get my data out
of Postgres and analyze it right within the database.
Wrt PDL::Stats, my guess is the test failure has nothing to
do with it being on a Mac/10.7. Perl 5.14.1 should be Perl
5.14.1 no matter what
platform, no?
On Dec 22, 2011, at 7:50 PM, chm wrote:
On 12/22/2011 7:48 PM, Puneet Kishor wrote:
Inspired by the lovely blurb below, I tried to install
PDL::Stats but
failed. Here is the relevant bit from the log of my failed
attempt --
---- PERL_DL_NONLAZY=1 /opt/local/bin/perl
"-MExtUtils::Command::MM" "-e"
"test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/stats_basic.t ..... ok t/stats_glm.t ....... ok
# Failed test at t/stats_kmeans.t line 168. #
got: '' # expected: '1' # Looks like you failed 1
test of 18. t/stats_kmeans.t .... Dubious, test returned
1 (wstat 256, 0x100) Failed 1/18 subtests
t/stats_ols_rptd.t .. ok t/stats_ts.t ........ ok
Test Summary Report ------------------- t/stats_kmeans.t
(Wstat: 256 Tests: 18 Failed: 1) Failed test: 13
Non-zero exit status: 1 Files=5, Tests=147, 1 wallclock
secs ( 0.03 usr 0.01 sys + 0.91
cusr
0.04 csys = 0.99 CPU)
Result: FAIL Failed 1/5 test programs. 1/147 subtests
failed. make: *** [test_dynamic] Error 255 -> FAIL
Installing PDL::Stats failed. See
/Volumes/roller/Users/punkish/.cpanm/build.log for details.
----
Here is line 168 (test #13) from t/stats_kmeans.t
168> is(tapprox( t_kmeans_bad(), 0 ), 1); 169> sub
t_kmeans_bad { 170> my $data = sequence 7, 3; 171>
$data = $data->setbadat(4,0); 172> my %m =
$data->kmeans({NCLUS=>2, NTRY=>10, V=>0}); 173>
return sum( $m{ms}->sumover - pdl qw( 1.5 1.9166667
1.9166667
) );
174> }
I am on Mac OS X Lion with Perl 5.14.1 and PDL 2.4.6
installed via
MacPorts
^^^^^ |||||
My guess it the problem is you are using PDL-2.4.6 and not
the current PDL-2.4.9 (or even better the current CPAN
Developers release version).
I know there were some fixes the are in PDL-2.4.9 that
helped Maggie X's stuff work. Maybe she can recall any
tricks to build her modules with a fairly old PDL (i.e.,
one not having the many fixes and features added in the
past couple of years).
--Chris
On Dec 22, 2011, at 3:26 PM, Maggie X wrote:
Hi Chris,
Thanks for including the info! Here's the blurb for
PDL::Stats.
-------------- This is a collection of statistics
modules in Perl Data Language,
with
a
quick-start guide for non-PDL people.
They make perldl--the simple shell for PDL--work like a
teenie weenie
R,
but with PDL threading--"the fast (and automagic)
vectorised
iteration
of
'elementary operations' over arbitrary slices of
multidimensional
data"--of
procedures including t-test, ordinary least squares
regression, and
k-means
clustering. ---------------
Best, Maggie
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
