Testcase for Issue 72: LayerNodeIndex:add() performance

[jonathanwin]

No description provided.

[peterneubauer]

@jonathanwin great work Jonathan! In order to pull this in, could you please send in the CLA, http://docs.neo4j.org/chunked/snapshot/cla.html ?

[jonathanwin]

---------- Forwarded message ----------
From: Jonathan Winterflood jonathan.winterflood@gmail.com
Date: Thu, Nov 15, 2012 at 3:15 PM
Subject: CLA
To: cla@neotechnology.com

Hi. My name is Jonathan Winterflood (jonathan.winterflood@gmail.com).
I agree to the terms in the attached Neo4j Contributor License Agreement.

Jonathan

[peterneubauer]

Great, will try to merge today if I get the time!

[craigtaverner]

I merged in the test case, and then disabled it until we can actually fix the performance bug. This should be done in issue 72, so we can consider issue 73 completed (pull request accepted).

[peterneubauer]

Awesome.
Thanks Craig!

On Tue, Jan 1, 2013 at 11:22 PM, Craig Taverner notifications@github.comwrote:

I merged in the test case, and then disabled it until we can actually fix
the performance bug. This should be done in issue 72, so we can consider
issue 73 completed (pull request accepted).

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/73#issuecomment-11794286.

[jonathanwin]

Constant is most certainly impossible (finding the position of an element
in a set is surely dependent on the size of the set, barring massive memory
trade-offs)

I imagine an r-tree is something like o(log(n)) (ignoring node overflows).
And as far as I know, the spatial r-tree itself does not have a performance
problem.
The problem here is that in front of the r-tree there is a o(n) test, and a
o(n) insert into a plain linked list of all indexed node ids.

This could probably be replaced by a second tree on the indexed ids,
yielding an overall o(log(n)). Or maybe something else...

Jonathan

On Fri, Jan 4, 2013 at 2:23 PM, amorgner notifications@github.com wrote:

What is the expected curve of insert performance? Constant rate?

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/73#issuecomment-11882589.

[amorgner]

I'm currently working on a fix. Seems much, much easier, with some hints I got from Craig here:
https://groups.google.com/forum/#!topic/neo4j/o07iBeHI8TU

[amorgner]

As this issue is closed, we should continue in #72

[ehx-v1]

you mean #82