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

Merge pull request #189 from openvenues/fix_trie_search

Browse Source 
Reset to root node in trie search on partial failed matches before rolling back pointer
This commit is contained in:
Al Barrentine
2017-04-21 14:39:03 -04:00
committed by GitHub
parent 074b6ff802 278679b7fb
commit 04eb2d4539
1 changed files with 12 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/trie_search.c
View File

@@ -348,8 +348,19 @@ bool trie_search_tokens_from_index(trie_t *self, char *str, token_array *tokens,
node = last_node = trie_get_node(self, start_node_id);
continue;
} else if (last_state == SEARCH_STATE_PARTIAL_MATCH) {
log_debug("last_state == SEARCH_STATE_PARTIAL_MATCH\n");
// we're in the middle of a phrase that did not fully
// match the trie. Imagine a trie that contains "a b c" and "b"
// and our string is "a b d". When we get to "d", we've matched
// the prefix "a b" and then will fall off the trie. So we'll actually
// need to go back to token "b" and start searching from the root
// Note: i gets incremented with the next iteration of the for loop
i = phrase_start;
log_debug("last_state == SEARCH_STATE_PARTIAL_MATCH, i = %zu\n", i);
last_match_index = -1;
phrase_start = phrase_len = 0;
node_id = last_node_id = start_node_id;
node = last_node = trie_get_node(self, start_node_id);
continue;
} else {
phrase_start = phrase_len = 0;