Hi Shaofeng, I think Inmem cubing code are so complicated, I wade through but get little.
1. In InMemCubeBuilder.java , line 276: CuboidResult newCuboid = buildCuboid(task.parent, task.childCuboidId); addChildTasks(newCuboid); Does this buildCuboid method output build results to ConcurrentDiskStore file ? If so, concurrent write to the same ConcurrentDiskStore file causes dirty records in Simultaneous jobs . And the blog http://kylin.apache.org/blog/2015/08/15/fast-cubing/ describing “ As “*” has no child, it will be outputted once be calculated, then comes “D”; After “C” be outputted, “CD” will be outputted as all of its children has finished; “ is not correct. it output parent cuboid then add child cuboids in above code . output order should be ABCD->BCD->CD->D->* 。 2. In BatchCubingJobBuilder2.java, line 122: final int maxLevel = CuboidUtil.getLongestDepth(seg.getCuboidScheduler().getAllCuboidIds()); Is this some more complicated? Why not use seg.getCuboidScheduler().getBuildLevel()); I see this snippet at line 71 in CuboidJob.java > On Oct 31, 2018, at 3:52 PM, ShaoFeng Shi <[email protected]> wrote: > > Thank you Jiatao! > > Dear users, if you see the document has a problem, please point it out with > email or JIRA, or fire a github PR to the "document" branch. We will continue > to refine the docs! > > JiaTao Tao <[email protected]> 于2018年10月31日周三 下午3:48写道： > You are welcome :). > > you Zhuang <[email protected]> 于2018年10月31日周三 上午10:11写道： > Much appreciation to JiaTao, bower install failed before , when I open the > proxy , all downloads succeed. Now everything goes well. > >> On Oct 31, 2018, at 9:46 AM, JiaTao Tao <[email protected]> wrote: >> >> bower --allow-root install >> > > > > -- > > Regards! > Aron Tao > > > -- > Best regards, > > Shaofeng Shi 史少锋 >
