Update utils.py
Browse files
utils.py
CHANGED
@@ -62,4 +62,30 @@ def rigorous_document_search(document: str, target: str):
|
|
62 |
start_index = document.find(reference)
|
63 |
end_index = start_index + len(reference)
|
64 |
|
65 |
-
return reference, start_index, end_index
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
start_index = document.find(reference)
|
63 |
end_index = start_index + len(reference)
|
64 |
|
65 |
+
return reference, start_index, end_index
|
66 |
+
|
67 |
+
def get_chunks_and_metadata(self, splitter):
|
68 |
+
# Warning: metadata will be incorrect if a chunk is repeated since we use .find() to find the start index. This isn't pratically an issue for chunks over 1000 characters.
|
69 |
+
documents = []
|
70 |
+
metadatas = []
|
71 |
+
for corpus_id in self.corpus_list:
|
72 |
+
corpus_path = corpus_id
|
73 |
+
if self.corpora_id_paths is not None:
|
74 |
+
corpus_path = self.corpora_id_paths[corpus_id]
|
75 |
+
|
76 |
+
with open(corpus_path, 'r') as file:
|
77 |
+
corpus = file.read()
|
78 |
+
|
79 |
+
current_documents = splitter.split_text(corpus)
|
80 |
+
current_metadatas = []
|
81 |
+
for document in current_documents:
|
82 |
+
try:
|
83 |
+
_, start_index, end_index = rigorous_document_search(corpus, document)
|
84 |
+
except:
|
85 |
+
print(f"Error in finding {document} in {corpus_id}")
|
86 |
+
raise Exception(f"Error in finding {document} in {corpus_id}")
|
87 |
+
# start_index, end_index = find_target_in_document(corpus, document)
|
88 |
+
current_metadatas.append({"start_index": start_index, "end_index": end_index, "corpus_id": corpus_id})
|
89 |
+
documents.extend(current_documents)
|
90 |
+
metadatas.extend(current_metadatas)
|
91 |
+
return documents, metadatas
|