tommy/libpostal
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[parser] using a bipartite graph (indptr + indices) to represent postal code<=>admin relationships instead of a set of 64-bit ints. Requires |V(postal codes)| + |E| 32 bit ints instead of |E| 64 bit ints. Saves several hundred MB in file size and even more space in memory because of the hashtable overhead

Browse Source
This commit is contained in:
Al
2017-03-18 06:05:28 -04:00
parent cb112f0ea7
commit c67678087f
7 changed files with 94 additions and 86 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/graph_builder.c
View File

@@ -47,10 +47,9 @@ static graph_t *graph_builder_build_edges(graph_builder_t *self, bool remove_dup
graph_edge_t edge = self->edges->a[i];
if (edge.v1 > last_vertex) {
for (uint32_t row = last_vertex; row < edge.v1; row++) {
graph_finalize_vertex(graph);
// Sorting is done prior to this
graph_finalize_vertex_no_sort(graph);
}
}
if (!remove_duplicates || i == 0 || edge.v1 != last_vertex || edge.v2 != last_edge) {
@@ -60,7 +59,7 @@ static graph_t *graph_builder_build_edges(graph_builder_t *self, bool remove_dup
last_edge = edge.v2;
}
graph_finalize_vertex(graph);
graph_finalize_vertex_no_sort(graph);
return graph;
}