Vertex(int number_of_vertex = -1, VertexDataType vertex_data = VertexDataType())
: number_of_vertex_(number_of_vertex),
vertex_data_(vertex_data) {}
std::vector<vertex_value<VertexDataType> > get_top_vertices(std::string basefilename, int ntop, vid_t from=0, vid_t to=0) {
typedef vertex_value<VertexDataType> vv_t;
/* Initialize striped IO manager */
metrics m("toplist");
stripedio * iomgr = new stripedio(m);
std::string filename = filename_vertex_data<VertexDataType>(basefilename);
int f = iomgr->open_session(filename, true);
size_t sz = get_filesize(filename);
/* Setup buffer sizes */
int nverts = (int) (sz / sizeof(VertexDataType));
ntop = std::min(ntop, nverts);
size_t bufsize = 1024 * 1024; // Read one megabyte a time
int nbuf = (int) (bufsize / sizeof(VertexDataType));
bufsize = sizeof(VertexDataType) * nbuf; // Ensure that data type size divides bufsize
assert(bufsize % sizeof(VertexDataType) == 0);
/* Initialize buffer */
VertexDataType * buffer = (VertexDataType*) calloc(nbuf, sizeof(VertexDataType));
vv_t * buffer_idxs = (vv_t*) calloc(nbuf, sizeof(vv_t));
vv_t * topbuf = (vv_t*) calloc(ntop, sizeof(vv_t));
vv_t * mergearr = (vv_t*) calloc(ntop * 2, sizeof(vv_t));
/* Read the data and maintain the top-list */
size_t nread = 0;
size_t offset = from * sizeof(VertexDataType);
size_t endoff = (to > 0 ? to * sizeof(VertexDataType) : sz);
size_t idx = 0;
int count = 0;
while (offset + nread < endoff) {
size_t len = std::min(endoff - (offset + nread), bufsize);
iomgr->preada_now(f, buffer, len, offset + nread);
nread += len;
int nt = (int) (len / sizeof(VertexDataType));
int k = 0;
VertexDataType minima = VertexDataType();
if (count > 0) {
minima = topbuf[ntop - 1].value; // Minimum value that should be even considered
}
for(int j=0; j < nt; j++) {
if (count == 0 || (buffer[j] > minima)) {
buffer_idxs[k] = vv_t((vid_t)idx + from, buffer[j]);
k++;
}
idx++;
}
nt = k; /* How many were actually included */
/* Sort buffer-idxs */
quickSort(buffer_idxs, nt, vertex_value_greater<VertexDataType>);
/* Merge the top with the current top */
if (count == 0) {
/* Nothing to merge, just copy */
memcpy(topbuf, buffer_idxs, ntop * sizeof(vv_t));
} else {
// void merge(ET* S1, int l1, ET* S2, int l2, ET* R, F f) {
merge<vv_t>(topbuf, ntop, buffer_idxs, std::min(ntop, nt), mergearr, vertex_value_greater<VertexDataType>);
memcpy(topbuf, mergearr, ntop * sizeof(vv_t));
}
count++;
}
/* Return */
std::vector< vv_t > ret;
for(int i=0; i < ntop; i++) {
ret.push_back(topbuf[i]);
}
free(buffer);
free(buffer_idxs);
free(mergearr);
free(topbuf);
iomgr->close_session(f);
delete iomgr;
return ret;
}
