00:29 < marcosirc> lozhnikov: Hi

00:30 < marcosirc> I think there are some errors in the copy constructor of rectangle trees.

00:31 < marcosirc> when creating the children nodes. We should update the parent pointer

00:31 < marcosirc> something like: children[i]->parent = this;

00:33 < marcosirc> Also, in the initialization of the dataset pointer:

00:33 < marcosirc> dataset(deepCopy ? new MatType(*other.dataset) : &other.Dataset()),

00:34 < marcosirc> This means that when deepCopy == true, it will create a new copy of the dataset for each node in the tree. Is this intended behaviour?

00:35 < marcosirc> I mean, all the nodes in the tree will have a different pointer in dataset, they won't share the same instance of the dataset.

00:35 < marcosirc> I am not sure if this is a problem, because I don't know how rectangle trees work.

00:38 < marcosirc> Also, it would be great if we could add a move constructor.

05:21 < lozhnikov> marcosirc: Hi, you are right. If I am not mistaken, we have discussed that with Ryan and we have decided that we should rewrite that later. Right now, this constructor is used as a move constructor (i.e. deepCopy = false) in different split types (e.g. RStarTreeSplit::SplitLeafNode()). The only purpose of this constructor is to increase the tree in depth

05:21 < lozhnikov> (i.e. copy the root and insert the new node to the root instead of its children). If deepCopy = true, the dataset is copied scores of times, but it is not used anywhere.

14:10 < marcosirc> lozhnikov: Ok, thanks for your comments.

14:11 < marcosirc> I am working on a PR to use rvalues when passing a reference tree to NeighborSearch class. Because of that I need a move constructor to the Rectangle Tree.

14:12 < marcosirc> If have enough time today I will add it :)

14:14 < lozhnikov> marcosirc: I saw that PR. I started working on the copy-constructor. That requires some refactoring of the Hilbert R tree.

14:15 < marcosirc> lozhnikov: Ok, great! Thanks!

16:28 < travis-ci> mlpack/mlpack#1449 (master - 20f4eac : Ryan Curtin): The build was fixed.

16:28 < travis-ci> Change view : https://github.com/mlpack/mlpack/compare/31fe55d4c7d7...20f4eacd1b08

16:28 < travis-ci> Build details : https://travis-ci.org/mlpack/mlpack/builds/153596879

16:46 < sumedhghaisas_> @marcosirc Hey Marcos...

16:46 < marcosirc> Hi

16:47 < sumedhghaisas_> I guess with the Spill Trees merge your project is complete :)

16:49 < sumedhghaisas_> Sorry I could not help you out in the last PR

16:50 < marcosirc> Yeah, finally it was merged! :)

16:51 < marcosirc> I finishing other PR now!

16:51 < sumedhghaisas_> yeah... lot of them :)

16:51 < sumedhghaisas_> You are doing a great job ... :)

16:55 < marcosirc> Thanks! :)

16:59 < sumedhghaisas_> I think you and Ryan changed a lot of code in these days... very hard to follow what is happening :P

17:00 < sumedhghaisas_> The defeatist search traversal idea is really good

17:00 < sumedhghaisas_> I liked it...

17:04 < marcosirc> Yeah, it was Ryan's idea.

17:07 < sumedhghaisas_> @rcurtin Hey Ryan.. I am not able to connect to big machine for ssh... Somehow a '/repo/' is getting added to every file in the coverage report...

17:11 < rcurtin> what do you mean, it looks like you hve logged in via ssh ok

17:11 < sumedhghaisas_> And some tests are faling... but I think thats because code is not compiled with optimizations...

17:11 < rcurtin> my internet connection is very bad right now, so sorry for the laggy responses :)

17:11 < sumedhghaisas_> huh... wait...

17:12 < sumedhghaisas_> ohh sorry... now its responding properly... yesterday it just hanged all the time...

17:13 < rcurtin> the machine isn't on a great connection,s o I think it can be flaky sometimes

17:13 < rcurtin> the new build server will come online soon, but I am still fighting with IT

17:13 < rcurtin> we have an IP now, but the firewall team is not opening the ports...

17:15 < rcurtin> there was a really good GSoC applicant we had this year who wanted to overhaul our build system, and the project would have included getting everything set up on the new build server and so forth

17:15 < rcurtin> he was very good, but I'm glad we didn't accept him because the system he would have been using is still not online!

17:15 < rcurtin> it would have been a very boring summer for him I guess...

17:16 < rcurtin> internally at Symantec we did a little better with our actual intern this year... he only had to wait until the last four weeks of his internship to start on his project, because we couldn't get the data he needed to do his work

17:16 < rcurtin> too much bureaucracy in both cases

18:59 < rcurtin> marcosirc: I think I looked at your PR too early in the morning, I have a couple more comments to make :)

19:00 < marcosirc> rcurtin: Ok, they are welcome

19:06 < marcosirc> rcurtin: Sorry, I have to leave now for some hours. I will check that comments as soon as I can.

19:07 < rcurtin> no worries, I am going too :)

19:07 < rcurtin> other than those two comments I agree it is ready to merge

19:14 < marcosirc> Ok, great.

19:15 < marcosirc> Also, I think the PR for move semantics is ready (except for fixing copy constructor of rectangle tree and adding a move constructor)

19:28 < lozhnikov> marcosirc: I think I've finished (see PR #767).

19:44 < marcosirc> lozhnikov: Great, thanks!