Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2014-08-30

| Channels | #pdl index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
06:23 sivoais ok, it's the weekend (past midnight here)! I'm going to start looking at these PDL tests now. I've already run test coverage and profiling. I need to analyse that to see which functions I should look at first.
06:23 mohawk nice
06:24 sivoais but I think the best use of my time now is find which files do not use the standard Test modules
06:24 sivoais easy enough with ack
06:25 sivoais ack -L 'Test::' t/
06:25 sivoais and convert them
06:26 sivoais also, saw this in #news just now: <http://cjfields.wordpress.com/2​014/08/30/whither-biofeatureio/>
06:35 sivoais I would eventually like to see Test::PDL functionality rolled into PDL.
06:44 mohawk interesting point about branches
06:45 mohawk i'm fighting dmake (i'm now the official dmake maintainer, btw)
06:45 mohawk it WILL learn to handle strings with spaces!
07:05 sivoais woah, what. dmake. That's awesome. I used to use amake on Windows since it was easy to install, but that doesn't look like it's changed in a while <http://amake.sourceforge.net/>.
07:07 sivoais ah, you imported it from SVN <https://github.com/mohawk2/dmake>
07:10 mohawk i took it from mercurial
07:15 sivoais it's been through a lot. Initial commit: Sep 22, 2000.
07:19 mohawk that was version 4.3 (or so)
07:20 mohawk so been through a lot more ;-)
14:41 jberger joined #pdl
14:51 chm joined #pdl
14:52 chm sivoais: See the TODO file in the PDL distribution
14:52 chm sivoais: Yes, I think Test::PDL needs to be part of PDL.
14:53 mohawk this reminds me of EUMM's tests
14:53 mohawk there's some good library functions in the test suite i moved into core EUMM
14:54 mohawk (like, 2 - not loads)
14:54 mohawk chm, is sivoais going to have commit access to PDL-code so he can work in a branch?
14:57 chm mohawk, sivoais:  I'm looking forward to the general pdl-porters discussion at http://mailman.jach.hawaii.edu/pipermai​l//pdl-porters/2014-August/006362.html
14:57 chm as I'
14:57 chm I think it is important to have the discussion for the record with all PDL developers
14:58 chm As I've mentioned before (and before...and before), the pdl-porters list is the place for this discussion.
14:58 chm The nuances and content don't work in single line "tweets" of information sans context
15:00 chm mohawk: Sheesh!  For the record I suggested a PDL devel release on Apr 14, 2014.  Again the week of Aug 20-ish, 2014.  Then within 3 days of said git repo on-line, the devel release was pushed.
15:00 chm I've already mentioned the context for why the sf.net git repo was not in sync but you seem more interested
15:01 chm in stirring the pot than in helping to synchronize things and move forward.
15:02 chm Also, awarding git repo is something all developers should be aware of and concur (generally).
15:02 chm Since I don't know either of you personally I can't just vouch for you myself.
15:03 chm To do so I would need a "warm fuzzy feeling" that things won't break if xxx comes on board.
15:03 chm To be honest, help would be welcome and sorting out the work flow and git branch structure
15:04 chm was not much of an issue when I was pretty much the only PDL developer doing main
15:04 chm distribution development.  With more activity, things need cleaning up and I will support that process.
15:09 mohawk ok
15:10 mohawk well, i'm not trying to sound-bite this
15:10 mohawk i'm subscribed to the pdl-porters list
15:10 mohawk the perl dev team and the perl toolchain guys seem to do ok with wider push access
15:10 mohawk by all means let's air this on pdl-p
15:11 mohawk i'd love to help with tidying the repo
15:11 mohawk on PTG's EUMM repo, it was a mess with 20 branches
15:11 mohawk i've got it down to 4 or so
15:12 mohawk (the others all got merged previously, or i brought them up to date, got consensus, and got them merged)
15:43 mohawk since this is on "the record", let me just respond to a couple of chm's points:
15:48 mohawk 1) i infer from your point about suggesting dev release Apr 14, that i had been a source of delay; in fact, i'd submitted a patch Sat, 12 Apr (https://sourceforge.net/p/pdl/patches/74/) which is identical to one i later put on my github repo when asked, and was (obviously) in favour of a dev release. if there's anything i should have done in terms of process, i'd like to know about it so i can be more helpful going forward
15:51 mohawk 2) as i referred to above, i've been on the pdl-p list for some time, and partaken in discussions - i've advocated the evolutionary approach for PDL2; it seems that what we have right now is a period of "calm" and there are changes we can usefully put out in a PDL2.008 once the smoke-testers have shown a dev release to be good
15:53 mohawk 3) as for a process for bringing in new git committers, similar to my proposal on Alien-Base - why not have someone email the list with a proposal of someone (with a little statement from them), and if in 7 days there are no objections, add them
15:54 mohawk further to (3) above, i'll propose myself on pdl-p right now
16:25 Mithaldu i'm all for more access, we're adults and anyone willing to invest time in PDL core is not going to be someone who flings shit at the walls
16:25 chm mohawk:  I'm not sure what took so long to communicate that a patch for a feature
16:25 chm is not what was needed.  I think perhaps because you don't communicate well/at all via email.
16:26 chm If you'll look through the pdl-porters logs and perldl you'll see that I've been working
16:26 chm in an evolutionary development cycle for *a* *long* *time*
16:30 chm As for (3), what has usually worked is that the developer participates
16:30 Mithaldu chm: idle curiosity and trying to untangle something a bit
16:31 chm in PDL development, contributes something, it then makes sense to directly push/pull as a developer
16:31 Mithaldu do i understand this right that he made that issue on SF, and you answered as to the necessity of that patch on the mailing list?
16:32 chm mithaldu: I've been trying to untangle this myself which is the difficulty.
16:33 chm The ExtUtils::Depends support patch was nice but when asked to
16:33 chm discuss for pdl-porters as to use/benefit/trade off there was no real follow through
16:34 chm When he understood that the offer was a devel release to demo the capability mohawk
16:34 chm made a git repo available which was promptly stood up.  At this point I would like to see a
16:35 chm use case for the new feature and how it helps and doesn't hurt from folks in the know.
16:35 chm Arguably, this could be a case of "duh" for someone in the know
16:35 Mithaldu chm: see, my sticking point here is:
16:35 Mithaldu i see no record of this in the issue on sf
16:36 chm I wanted someone else to have a discussion on usability/validity but
16:36 Mithaldu and i can't tell if that is because both parties mutually agreed to discuss elsewhere
16:36 chm since that discussion was never had, nothing went farther.
16:36 chm That is the problem in that I don't really use #pdl effectively and it is not
16:37 chm the support mechanism for PDL but I've tried to look at logs and backfill.
16:37 chm That seems to have backfired in the case of mohawk since he never followed up
16:37 Mithaldu chm: put more simply
16:37 chm with discussion on pdl-porter and usually only made constructive (to me) points on #pdl
16:38 chm so we're kind of at dysfunctional operating purposes.  Very frustrating to me!
16:38 Mithaldu aha
16:38 Mithaldu note for future: it is generally good to keep a record in a ticket and copypaste discussions to it when necessary
16:39 chm Yes, again, I think that is a *great* idea but don't have the time to scoop xxxx from other peoples
16:39 chm discussions when they could scoop their own, as it were.
16:39 Mithaldu understandable, but easily solved by asking them to do it :)
16:39 chm I've asked that tickets be opened.
16:40 Mithaldu not what i mean
16:40 chm I've mentioned that #pdl doesn't really lend it self to logging and reference
16:40 chm What do you mean?
16:40 Mithaldu i mean, when someone tells you something here, and you consider it valuable to an existing ticket, ask them to paste the discussion log in the ticket
16:41 Mithaldu we've had good success in creating reliable records like that with quite a few toolchain projects
16:41 chm I definitelydo that.  The catch is new blood has new/good ideas that need their own new tickets
16:42 Mithaldu but he made a ticket, didn't he?
16:42 chm Yes, and that has moved forward.  The final piece was to show how it worked and get
16:42 chm concurrence/agreement from someone who understands the ExtUtils::Depends and its
16:43 Mithaldu and of that exists no record in : http://sourceforge.net/p/pdl/patches/74/
16:43 chm benefits and usage.  From the dev release, it appears to be relatively minimal in changes
16:43 chm and could have some good benefits which I am still unclear on.
16:44 chm Well, that only happened very recently and with much of this impedent mismatch between
16:44 chm #pdl and pdl-porters discussion.  Never got to the point where it made sense to add to a ticket
16:45 Mithaldu anything makes sense to add to a ticket :)
16:45 chm without understanding by me, mohawk, and other developers
16:45 Mithaldu it should be the definite record of actions taken
16:45 Mithaldu even if the action is "explain this shit yo"
16:45 chm That speaks to a need to clarify or develop a better interaction and development model for PDL
16:46 Mithaldu also, on a more constructive vein: it looks like EU::Depends simply allows other dists to ask PDL for metadata about its XS bits
16:46 Mithaldu which mean other modules that need to build with PDL bits can simply have a single line there, instead of needing to track what PDL does and rereleasing when PDL changes things
16:47 chm Yes, lets be constructive:  a simple example of how EU::Depends would be used by another
16:47 Mithaldu which is a super great thing
16:47 chm PDL module developer would be great.
16:47 Mithaldu example is here: http://sourceforge.net/p/pdl/feature-requests/81/
16:48 Mithaldu instead of putting in the typemaps, inc, lbs and deps hash pairs manually
16:48 chm If you consider that an explanatory example then you are "in the know" and think "duh" is a tutorial.
16:48 Mithaldu eu::dp allows gimp-perl to simply call "get_makefile_vars" in Makefile.PL
16:48 Mithaldu no, i'm explaining as i go along
16:48 chm I'm on the other side of the fence looking for PDL::XXX used the old, hand rolled way.
16:49 chm Now with the EU::Depends support, it becomes this:  wonderfully better code here
16:49 chm Bonus points for an example from another developer than the original author/user
16:49 Mithaldu i understand your concerns, but can you please concentrate on telling me whether my explanation makes sense?
16:50 Mithaldu i can't change the world, but i can help you understand things by trying to explain them :)
16:50 chm I would like a non-gimp-perl example.  We have other PDL modules that need PDL XS (I believe).
16:50 Mithaldu name one?
16:50 chm Does the new feature help them?
16:51 Mithaldu name one and i can look at it to see if it would help
16:51 chm Again, I don't know enough about the EUD details to know what would qualify there.
16:51 chm This seems to be a chicken and egg problem.  I *don't* know what the big win would be
16:52 chm and a good feature should support more than one use case to be worth it.  Presumably
16:52 Mithaldu it seems like pdlmaker does some of this already, if not all
16:52 chm mohawk would understand who/what other modules could use EUD support.
16:53 Mithaldu do you know where i can find the code for pdlmaker?
16:53 chm What is pdlmaker?
16:53 Mithaldu ah, not part of pdl
16:53 Mithaldu https://metacpan.org/source/MOO​COW/PDL-CCS-1.19.1/pdlmaker.plm
16:53 Mithaldu https://metacpan.org/source/MOO​COW/PDL-CCS-1.19.1/Makefile.PL
16:54 Mithaldu i suspect using eu::dp would make pdlmaker.plm unnecessary there
16:54 chm That would be _precisely_ the example/demonstration I was looking for.
16:54 Mithaldu https://metacpan.org/source/CHM/​PDL-2.007/Basic/Core/Dev.pm#L442
16:55 Mithaldu it could possibly also obviate some of that
16:55 Mithaldu mohawk: have you looked at these files before?
16:56 chm BTW, there are two parts to this discussion as I see it:
16:56 chm (1) Clean up of XS dependencies handling which is critical to get right/better for PDL3
16:57 chm (2) Adding in new capabilities to the stable PDL-2.x chain in an evolutionary way.
16:57 chm New features are nice but they shouldn't break old code.
16:57 chm It would also be nice not to set a standard for something that will *have* to change to get completely "right" for PDL3
16:58 Mithaldu that is beyond my knowledge and scope right now
16:58 chm Those were the considerations.  Difficult to make progress without the
16:58 Mithaldu but i hope mohawk will be able to deal with this now that the dialogue has progressed :)
16:58 chm specific EUD understanding and demo of capability.
16:59 chm Me too!  :-)
17:00 Mithaldu great :)
17:00 Mithaldu that means i can go shopping now
17:00 Mithaldu and if nothing else happens while i'm gone, i'll paste this log to the ticket
17:00 chm It would be great to have MOOCOW concur as well.
17:00 chm Thanks for the patience and dialog with mibbit-#pdl-er
17:01 Mithaldu np, at least you have command of the english language :)
17:01 chm Have to go myself for now... o/
17:01 Mithaldu o/
17:28 mohawk Mithaldu, thanks for the well-made points there
17:40 mohawk i've emailed MOOCOW, we'll see if anything comes of it
17:48 mohawk sivoais, unless you can see a reason not to, why not submit a proposal statement that you gain PDL git access? maybe along the lines of this, but not as long-winded and better ;-) http://mailman.jach.hawaii.edu/piperma​il/pdl-porters/2014-August/006368.html
17:55 sivoais mohawk: I will. I only want to be able to create branches and get code review. I guess the conventions for that have been informal so far. I'll note all this in my e-mail.
17:55 sivoais thanks mohawk, Mithaldu, and chm for having this discussion.
17:58 Mithaldu cheers, hope you guys can draw use from it :)
18:01 mohawk sivoais, you should draw inspiration from the current perl guide on this
18:01 mohawk it may be in "perlgit"
18:07 Mithaldu mohawk: can you please update your tickets accordingly too?
18:09 mohawk https://metacpan.org/pod/dist​ribution/perl/pod/perlgit.pod
18:09 mohawk Mithaldu, which tickets?
18:10 Mithaldu mohawk: the two i linked
18:10 Mithaldu see: above :)
18:10 sivoais mohawk: thanks
18:10 mohawk welcome
18:11 mohawk Mithaldu, i see my patch, and my feature-request
18:12 Mithaldu yes
18:12 Mithaldu add record of events to those :)
18:12 mohawk which one? both?
18:12 mohawk that seems redundant
18:12 Mithaldu do it in one andmake other refer to it
18:47 mohawk i'm looking at the SF pdl-code repo page
18:47 mohawk i literally didn't know SF had the capability to do forks, though it seems obvious in retrospect
18:48 mohawk there's no v2.007_04 tag, that needs sorting
18:48 mohawk there's SOOO many branches that could use tidying
19:29 jberger I have made a move, essentially by fiat and then covered my tracks with policy post-facto. I want to see AB continue forward. All those pending are in since I haven't heard anyone say anything against
19:30 jberger open-source is driven by the people with the enthusiasm to get things done
19:30 jberger and I don't want to waste that
19:35 mohawk jberger++
19:35 sivoais jberger: thanks for the Alien::Base add. I hope to add to the discussion in the coming months as time permits. :-)
19:41 sivoais jberger: you might find what I'm doing here with PDL + Moo interesting <https://github.com/zmughal/data-frame-​experiment/blob/master/t/pdl-factor.t>. I based it off your code from a year ago.
19:48 mohawk https://sourceforge.net/p/pdl/feature-requests/81/ updated
19:49 mohawk sivoais, feel any urge to include XS (possibly via Inline) in your experiment?
19:53 sivoais quick question: is there a way to subscribe via e-mail to all the issue notifications on the PDL sourceforge?
19:54 sivoais mohawk: quite likely. My next move with it is to make data frames work with R's data frames
19:55 mohawk that sounds hard, good luck
19:57 sivoais worth trying :-) PDL doesn't have the number of stats tools developers R has, so that would be a very nice route to share the work with the R community
20:05 jberger sivoais: I'm not really sure what I'm looking at here
20:09 sivoais jberger: I have made a PDL subclass that supports enums (like R's factors for categorical variables)
20:10 sivoais I have also made one that holds arbitrary Perl scalars (like a Math::BigInt)
20:10 sivoais all just using Moo and pure Perl
20:15 jberger interesting
20:16 jberger I know run4flat had some PDL string handling (I think that was it) at some point
20:18 sivoais string output or input? I'd be interested in output, since I'm trying to generalise stringification using roles (all so I can get tabular layout in my data frames).
20:18 mohawk this is some high-octane thinking, wow
20:20 jberger IIRC it was a PDL that had the ability to contain strings in the values
20:20 jberger perhaps it was just pointers to strings, I can't remember
20:20 jberger run4flat is really one whole brain smarter than me in some of that area
20:20 sivoais lol, hence the -experiment part. That allows me to test out crazy things and not worry about people using them. :-D
20:23 sivoais I'll e-mail this to the list later to get feedback. I would like to try pointers to SVs later to save space. I currently index into a Perl array. I think that helps with the refcount.
20:24 mohawk pointers to SVs - now that's some XS stuff
20:24 mohawk Inline that shiz!
20:25 jberger PDL will never be free of XS (in Perl5 at least)
20:59 mohawk not saying it should be free of XS!
20:59 mohawk just pointing out the opportunity to do a little Inline magic ;-)
20:59 mohawk Inline - because directly writing XS sucks
21:04 sivoais jberger: is there a voting process for merging PRs?
21:04 sivoais in Alien::Base
21:05 sivoais I guess that falls under a feature :-/
21:08 mohawk i suggest the way should be to get someone else to merge your own (although i just didn't do that with my relatively trivial test update)
21:08 mohawk i used the PR mechanism there for accountability reasons
21:08 mohawk i would suggest we not have votes or delay - if a core person likes a PR, merge it
21:08 mohawk fix any problems in post-production
21:08 mohawk what do people think?
21:09 jberger use the PR mechanism, ask for opinions, the more contentious something is likely to be, be more careful (even if you believe in the change)
21:10 sivoais I'll err on the side of caution with PR 30.
21:13 mohawk jberger, any reason not to go forward on https://github.com/Perl5-Alien/Alien-Base/pull/30 ?
21:23 mohawk silence = assent
21:23 mohawk let's not have a voting process - with source control we can always back changes out, or modify them?
21:24 mohawk now, when it comes to releases, that's when the formal stuff is essential
22:26 mohawk what do we think of the idea of having a good stab at all the current AB issues/PRs that are straightforward, then making a dev release?
22:28 jberger what silence?
22:29 jberger this has been the most activity day AB has had (outside of my own self) ever
22:30 jberger and I'm not kidding mohawk, the point of the system is to be something that people can rely on for functionality that is many installation layers deeper than apps
22:30 jberger I want enthusiasm, but not recklessly
22:47 mohawk ok
22:47 mohawk i withdraw my loud demand to immediately release a::b v0.5
22:52 mohawk i will remind you of how many modules depend on EUMM
22:53 mohawk since you seem to think of me as reckless, which is nice
23:01 jberger true :-)
23:01 jberger maybe I shouldn't say reckless, but I am feeling a bit overwhelmed atm
23:01 jberger also, I just saw this: https://github.com/jberger/Galileo/issues/30
23:02 jberger so I'm even distracted on top of overwhelmed
23:03 mohawk if i can help, let me know
23:03 mohawk you're a man in great demand ;-)

| Channels | #pdl index | Today | | Search | Google Search | Plain-Text | summary