libkd documentation =================== C API ----- .. highlight:: c .. c:function:: kdtree_t* kdtree_build(kdtree_t* kd, void *data, int N, int D, int Nleaf, int treetype, unsigned int options); Build a tree from an array of data, of size N*D*sizeof(data_item). *kd*: NULL to allocate a new *kdtree_t* structure, or the address of the structure in which to store the result. *data*: your N x D-dimensional data, stored in N-major direction: data[n*D + d] is the address of data item "n", dimension "d". If 3-dimensional data, eg, order is x0,y0,z0,x1,y1,z1,x2,y2,z2. *N*: number of vectors *D*: dimensionality of vectors *Nleaf*: number of element in a kd-tree leaf node. Typical value would be about 32. *treetype*: * if your data are doubles, *KDTT_DOUBLE* * if your data are floats, *KDTT_FLOAT* For fancier options, see *kd_tree_types*. *options*: bitfield of *kd_build_options* values. Specify one of: * *KD_BUILD_BBOX*: keep a full bounding-box at each node; * *KD_BUILD_SPLIT*: just keep the split dimension and value at each node. see *kd_build_options* for additional fancy stuff. NOTE that this function will *permute* the contents of the *data* array! When you're done with your tree, be sure to *kdtree_free()* it. Example: .. code-block:: c double mydata[] = { 1,1, 2,2, 3,3, 4,4, 5,5, 6,6, 7,7, 8,8 }; int D = 2; int N = sizeof(mydata) / (D * sizeof(double)); kdtree_t* kd = kdtree_build(NULL, mydata, N, D, 4, KDTT_DOUBLE, KD_BUILD_BBOX); kdtree_print(kd); kdtree_free(kd); .. c:function:: void kdtree_free(kdtree_t *kd); Frees the given *kd*. By default, the *kd->data* is NOT freed. Set *kd->free_data = 1* to free the data when *kdtree_free()* is called. Python API ---------- Code Internals --------------