simonstey

TOPIC: Scope and process for profiles work
NickC: For the profiles work, if we decide to abandon it, that decision would be made in the main group, not here.
NickC: So, I don't think we need to do much in this group unless things get serious. We can just discuss here.
NickC: Serious arguments should go to GitHub or the main group meeting.
AlexN: Thank you, my hands are off the keyboard then.
NickC: What I need in the next little while is feedback on scope and the discussions we're having about the nature of the profile we're trying to capture.
NickC: We're still in that formative stage and will be for some time. We're not up to making hard decisions yet.
NickC: We should come to a consensus on what we're doing there, and then we might move into a different mode.
NickC: Our charter's scope is about profiling. We don't have specific guidance on our scope, which is what we're still establishing.

simonstey

AndyS: Very simple implementation that really just goes round the loop and keeps iterating until nothing changes, which will get the right answers for a while.
AndyS: I like having a very simple implementation, because then you can compare it against the clever ones.
AndyS: Let's see if you get the same answers. It's the algorithm in the document, I think. There's one of two. You can do it this way or that way.
AndyS: Ways of writing out naive evaluation, almost slightly. Slight changes on efficiency, but they don't have any changes on the outcome.
DavidHabgooDavidHabgood: Okay.
AndyS: But as soon as negations involved... you don't have to... you have to have ways of saying no to negation.
AndyS: And stratification is a way to do that. It might be useful to write about stratification, and I hope that the well-formedness conditions.
AndyS: On the rules for a start. For that, because they relate to how rules relate to other rules, and it should give us some language to talk about.
AndyS: Stratification and rule

simonstey

you can try it out here -> http://shacl.org/playground/

simonstey

test.json was ~100MB (plain JSON; real-world data) 

for creating sample data you could use, e.g., http://www.json-generator.com 
    -> and then `curl -s 'http://www.json-generator.com/api/json/get/cvfsLVmKiG?indent=2' > test.json`
    
run it with `python shootout.py` assuming test.json is in the same folder as shootout.py