[seek-kr-sms] Re: [seek-dev] full garp pipeline

[Shawn Bowers]

Bertram Ludaescher wrote:
>>>>>>"CB" == Chad Berkley <berkley at nceas.ucsb.edu> writes:
>>>
>>>I guess the fact that there is no actor is on purpose (this seems to
>>>be a design workflow, not an executable one).
> 
> CB> 
> CB> If you mean director instead of actor, then yes.
> CB> 
> 
> yes, that's what I meant.
> 
> 
>>>In the long run, we may actually have a "design director" -- if
>>>nothing else, it could do type checking and other static analysis!
>>>
> 
> CB> 
> CB> that's a good idea.  I'd thought about that a couple times before, 
> CB> although I'm not sure if there is a need.  the type system works without 
> CB>   a director.
> 
> Right. But how about the semantic type checking -- well, ideally it
> should become part of the "regular" type checking.
> Another try: what about scheduling? Now this needs a
> director. But we can probably reuse the existing directors, if you
> implement your generic design/blank-slate actor (and its instantiated
> versions) in such a way that they output some (fixed or generated)
> gobbledygook, just for the purpose of simulating the actual execution
> ... 
> 
> 

One thing that we might consider adding to a "design director" are 
access to simple planning rules and an interface for exploiting these 
rules.  As an example, a scientist may wish to connect two actors that 
do not have compatible types, but by adding intermediate conversions, 
they can be made to be compatible.  The director would search for and 
suggest appropriate conversions. Another related use would be an "actor 
completion" functionality.  Based on an existing actor, we might ask the 
director to find matching actors that could take the existing actor's 
output as input, perhaps based on "keywords" acting as filters.

Ultimately (but maybe not part of SEEK), it would be nice to have a 
button that "cleans up" the designed workflow -- applying rules for 
simplifying loops, checks that the workflow does not have erros, and so 
on -- where this button is implemented by or part of the design director.

Shawn


>>>Also: why do you call this "dynamic source"?
>>>Is this because you want to plug in the code into the "design actors"
>>>later?
> 
> CB> 
> CB> it's just called dynamic source because that's what I called it when I 
> CB> build the new actor wizard.  it's probably a bad name at this point and 
> CB> should be changed.
> 
> ok. How about "design actor" or "stub actor" or "blank actor" or... ?
> 
> I think design is good since it indicates what it is used for.
> 
> 
>>>Finally, 1 or 2 months for the installer sound good -- Kepler-2-Go;
>>>here we come =B-)
> 
> CB> 
> CB> coded by Bertram :)
> 
> almost. But I only make the suggestions here ;-)
> 
> Bertram
> 
> PS But I'm working on the shipping and handling (SHA) algebra right
> now. Here is what a few lines of Prolog output:
> 
> 
> y at c = f at a(x at b) -->
> [x at b -> a, @a: y = f(x), y at a -> c]
> [f at a => b, @b: y = f(x), y at b -> c]
> [x at b -> c, f at a => c, @c: y = f(x)]
> 
> This is for Grid stuff: it says that for executing the code f,
> residing at a on dataset x residing at b, resulting in dataset y at c, 
> you can use 3 different plans: 
> 1. ship data x to a, execute f, then ship the result to c
> 2. ship code f to b, execute f, then ship result to c
> 3. ship x and code f to c,  execute f.
> 
> the "->" is data shipping, the "=>" is code shipping. 
> Of course optimizations are possible, based on whether additional
> copies of x,y,z, or f are around. 
> 
> Also, this becomes an interesting optimization problem when longer
> plans are created... thinking ... ;-)
> 
> 
> CB> 
> 
>>>Bertram
>>>
>>>
>>>
>>>
>>>>>>>>"CB" == Chad Berkley <berkley at nceas.ucsb.edu> writes:
>>>>>>>
> CB> 
> CB> Hi,
> CB> I've put the GARP pipeline we developed last week in SEV up on my web 
> CB> site for anyone to download.  Note that this pipeline does not run, but 
> CB> is a prototype of the entire GARP process.  There are two files you 
> CB> need.  First get http://trestles.nceas.ucsb.edu/full-garp.xml.  You can 
> CB> save this file anywhere...I recommend in your kepler directory.  Then 
> CB> get http://trestles.nceas.ucsb.edu/dynsrc.tar.gz.  This file will need 
> CB> to be unzipped into your kepler directory.  The directory structure 
> CB> after you unzip it should be 
> CB> kepler/dynsrc/org/ecoinformatics/seek/workflow/.  All of the needed 
> CB> actors should be in that directory.  Let me know if you can't get this 
> CB> to work.
> CB> 
> CB> In case any of you want to fix the bug we discovered where the garp 
> CB> pipeline window gets created off the screen, edit the file 
> CB> kepler/lib/garpModel.xml.  Edit the file and find the line that says 
> CB> something like
> CB> <property name="_windowProperties"...  on that line there should be an 
> CB> xml attribute that looks like
> CB> value="{bounds={1300, 2700, 975,590}}".  Change the first two values 
> CB> (they may not be 1300 and 2700 but they will be rather large) to 10,10. 
> CB> Then execute 'ant clean run'.  the next time you open the garp model 
> CB> it should be on the screen in front of you instead of off in 
> CB> neverneverland.  let me know if you have problems.
> CB> 
> CB> Thanks to everyone that was at SEV for the BEAM/AMS/KR meeting last 
> CB> week.  It was extremely productive and I think our software projects, 
> CB> specifically kepler, will benefit tremendously.  I'll let everyone know 
> CB> when I get various things working/fixed in kepler and hopefully we'll 
> CB> have an installer within the next month or two.
> CB> 
> CB> chad
> CB> 
> CB> _______________________________________________
> CB> seek-dev mailing list
> CB> seek-dev at ecoinformatics.org
> CB> http://www.ecoinformatics.org/mailman/listinfo/seek-dev
> CB> 
> CB> 
> CB> -- 
> CB> -----------------------
> CB> Chad Berkley
> CB> National Center for
> CB> Ecological Analysis
> CB> and Synthesis (NCEAS)
> CB> berkley at nceas.ucsb.edu
> CB> -----------------------
> _______________________________________________
> seek-kr-sms mailing list
> seek-kr-sms at ecoinformatics.org
> http://www.ecoinformatics.org/mailman/listinfo/seek-kr-sms