erdavila/m-tree

A data structure for efficient nearest-neighbor queries.

m-tree's Issues

number of distance calculations

Hi,

this M-Tree Java implementation works as it outputs the correct nearest neighbors in the correct order. Thank you for open sourcing this! But when I count the number of distance calculations I always get a larger number than there are items in the M-Tree :/ Am I doing something wrong?

flo

error: ISO C++17 does not allow dynamic exception specifications

I am trying to run this project with new GCC version 11.2.0 and facing an issue with dynamic exception specifications error:

mtree.h:669:85: error: ISO C++17 does not allow dynamic exception specifications
669 | void addData(const Data& data, double distance, const mtree* mtree) throw(SplitNodeReplacement) {
| ^~~~~
mtree.h:723:98: error: ISO C++17 does not allow dynamic exception specifications
723 | virtual void doRemoveData(const Data& data, double distance, const mtree* mtree) throw (DataNotFound) = 0;
...

Removing the throw specification completely will resolve the issue but is there any other way to fix this problem? I have tried to change the default configuration settings(i.e. "cppStandard": "c++17" to "cppStandard": "c++11") but still getting the same issue.

Splitting realization with throw-catch

As I can see splitting the nodes in a bottom-up fashion is realized by throwing and catching some data. But it's said that throwing-catching should only be applied for exceptional situation, errors, not in the usual program execution, because that mechanism is very slow and just breakes the rule of good programming style.
So why not add in every node a reference to it's parent? Is it the way to avoid the stack overflow due to a recursion?