https://issues.apache.org/jira/browse/TINKERPOP3-670 created for tail step. Let me know if the scenarios described there are sufficient.
On Thu, May 7, 2015 at 9:57 AM, Marko Rodriguez <[email protected]> wrote: > Hey Matt, > > Yea, "tail(long)" would be good. I don't like limit(-1). If you want to > implement TailGlobalStep (as there should be a local too -- but that is > trivial to write), it should be pretty straightforward. Look at RangeStep > as you will have to be smart about traverser.bulk(). And it will look like > RangeStep, save you have a queue of Traversers you build up based on the > bulk(). And when the !starts.hasNext(), you emit the queued Traversers. > > HTH, > Marko. > > http://markorodriguez.com > > On May 7, 2015, at 10:50 AM, Matt Frantz <[email protected]> > wrote: > > > One clarification inline. > > > > On Thu, May 7, 2015 at 5:23 AM, Marko Rodriguez <[email protected]> > > wrote: > > > >> Hello Matt, > >> > >>> - 639 <https://issues.apache.org/jira/browse/TINKERPOP3-639> (select > >> and > >>> selectList). There is already PR#57 > >> > >> I'm still not sure this is the path that we want to go down. I just see > >> rippling requirements of xxxList or xxxMap falling from such a choice. > >> > > > > What about the discussion on tail/head or last/first or allowing negative > > indices in range/limit? Spin that into its own JIRA ticket? > > > > > >> > >>> - 619 <https://issues.apache.org/jira/browse/TINKERPOP3-619> (select > >>> should not throw). There was some progress on choosing the desired > >>> behavior, but it sounds like a final decision is pending. > >> > >> This is doable. I suspect a no-result for when nothing binds --- like > how > >> filter() works (i.e. while(true) { if(hasSomething) return something;}) > >> > >>> - 652 <https://issues.apache.org/jira/browse/TINKERPOP3-652> (select > >>> from Map explicitly). Is it a good idea to require Scope.local to do > >>> select from Map? Or is selectKeys more appealing? > >> > >> This is a good one. I just don't like the select(local,"a","b") in this > >> context as it makes MatchStep look ugly. However, it is probably the > way to > >> go and only MatchStep usage would be effected. SelectGlobalStep, > >> SelectOneGlobalStep, SelectLocalStep, and SelectOneLocalStep. eek. Lots > of > >> "stuff" but yea, the Scope concept should propagate to here. > >> > >> What would be crazy is g.V.out.select(local,"name","age")…….. gnarly. I > >> always wanted Element to implement Map, but there were reason (I forget) > >> why it wouldn't work. If it did, then has() would work on Maps. > >> > >> hashMapStep.has("a").has("a",eq(32)) > >> > >> Instead, for such situations, we have where(). > >> > >> hasMapStep.where("a",eq(32)) > >> > >> We don't have where("a"). Be easy to add. > >> > >>> Not trying to rush anyone, but I have cycles to spend on this over the > >> next > >>> couple of weeks. > >> > >> HTH, > >> Marko. > >> > >> http://markorodriguez.com > >
