Update pipeline
Browse filesFix a bug when getting metadata license and topics, use list of dictionaries to store the results instead of a single dictionary by pushing repo name inwards, add repo stars to the result
- pipeline.py +32 -22
pipeline.py
CHANGED
@@ -19,16 +19,12 @@ def extract_code_and_docs(text: str):
|
|
19 |
tuple: A tuple of two sets, the first is the code set, and the second is the docs set,
|
20 |
each set contains unique code string or docstring, respectively.
|
21 |
"""
|
22 |
-
root = ast.parse(text)
|
23 |
-
def_nodes = [
|
24 |
-
node
|
25 |
-
for node in ast.walk(root)
|
26 |
-
if isinstance(node, (AsyncFunctionDef, FunctionDef, ClassDef, Module))
|
27 |
-
]
|
28 |
-
|
29 |
code_set = set()
|
30 |
docs_set = set()
|
31 |
-
|
|
|
|
|
|
|
32 |
docs = ast.get_docstring(node)
|
33 |
node_without_docs = node
|
34 |
if docs is not None:
|
@@ -55,16 +51,22 @@ def get_metadata(repo_name, headers=None):
|
|
55 |
|
56 |
|
57 |
def download_and_extract(repos, headers=None):
|
58 |
-
|
59 |
for repo_name in tqdm(repos, disable=len(repos) <= 1):
|
60 |
# Get metadata
|
61 |
metadata = get_metadata(repo_name, headers=headers)
|
62 |
-
|
|
|
63 |
"funcs": set(),
|
64 |
"docs": set(),
|
65 |
-
"topics":
|
66 |
-
"license":
|
|
|
67 |
}
|
|
|
|
|
|
|
|
|
68 |
|
69 |
# Download repo tarball bytes
|
70 |
download_url = f"https://api.github.com/repos/{repo_name}/tarball"
|
@@ -86,8 +88,8 @@ def download_and_extract(repos, headers=None):
|
|
86 |
file_content = tar.extractfile(member).read().decode("utf-8")
|
87 |
code_set, docs_set = extract_code_and_docs(file_content)
|
88 |
|
89 |
-
|
90 |
-
|
91 |
except UnicodeDecodeError as e:
|
92 |
tqdm.write(
|
93 |
f"[-] UnicodeDecodeError in {member.name}, skipping: \n{e}"
|
@@ -95,7 +97,9 @@ def download_and_extract(repos, headers=None):
|
|
95 |
except SyntaxError as e:
|
96 |
tqdm.write(f"[-] SyntaxError in {member.name}, skipping: \n{e}")
|
97 |
|
98 |
-
|
|
|
|
|
99 |
|
100 |
|
101 |
class RepoEmbeddingPipeline(Pipeline):
|
@@ -140,6 +144,7 @@ class RepoEmbeddingPipeline(Pipeline):
|
|
140 |
|
141 |
if self.st_messager:
|
142 |
self.st_messager.info("[*] Downloading and extracting repos...")
|
|
|
143 |
extracted_infos = download_and_extract(inputs, headers=self.API_HEADERS)
|
144 |
|
145 |
return extracted_infos
|
@@ -171,14 +176,19 @@ class RepoEmbeddingPipeline(Pipeline):
|
|
171 |
return sentence_embeddings
|
172 |
|
173 |
def _forward(self, extracted_infos, max_length=512, st_progress=None):
|
174 |
-
repo_dataset =
|
175 |
-
num_texts = sum(
|
176 |
-
len(x["funcs"]) + len(x["docs"]) for x in extracted_infos.values()
|
177 |
-
)
|
178 |
with tqdm(total=num_texts) as pbar:
|
179 |
-
for
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
180 |
pbar.set_description(f"Processing {repo_name}")
|
181 |
-
entry = {"topics": repo_info["topics"], "license": repo_info["license"]}
|
182 |
|
183 |
message = f"[*] Generating embeddings for {repo_name}"
|
184 |
tqdm.write(message)
|
@@ -219,7 +229,7 @@ class RepoEmbeddingPipeline(Pipeline):
|
|
219 |
else None
|
220 |
)
|
221 |
|
222 |
-
repo_dataset
|
223 |
|
224 |
return repo_dataset
|
225 |
|
|
|
19 |
tuple: A tuple of two sets, the first is the code set, and the second is the docs set,
|
20 |
each set contains unique code string or docstring, respectively.
|
21 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
code_set = set()
|
23 |
docs_set = set()
|
24 |
+
root = ast.parse(text)
|
25 |
+
for node in ast.walk(root):
|
26 |
+
if not isinstance(node, (AsyncFunctionDef, FunctionDef, ClassDef, Module)):
|
27 |
+
continue
|
28 |
docs = ast.get_docstring(node)
|
29 |
node_without_docs = node
|
30 |
if docs is not None:
|
|
|
51 |
|
52 |
|
53 |
def download_and_extract(repos, headers=None):
|
54 |
+
extracted_infos = []
|
55 |
for repo_name in tqdm(repos, disable=len(repos) <= 1):
|
56 |
# Get metadata
|
57 |
metadata = get_metadata(repo_name, headers=headers)
|
58 |
+
repo_info = {
|
59 |
+
"name": repo_name,
|
60 |
"funcs": set(),
|
61 |
"docs": set(),
|
62 |
+
"topics": [],
|
63 |
+
"license": None,
|
64 |
+
"stars": metadata.get("stargazers_count"),
|
65 |
}
|
66 |
+
if metadata.get("topics"):
|
67 |
+
repo_info["topics"] = metadata["topics"]
|
68 |
+
if metadata.get("license"):
|
69 |
+
repo_info["license"] = metadata["license"]["spdx_id"]
|
70 |
|
71 |
# Download repo tarball bytes
|
72 |
download_url = f"https://api.github.com/repos/{repo_name}/tarball"
|
|
|
88 |
file_content = tar.extractfile(member).read().decode("utf-8")
|
89 |
code_set, docs_set = extract_code_and_docs(file_content)
|
90 |
|
91 |
+
repo_info["funcs"].update(code_set)
|
92 |
+
repo_info["docs"].update(docs_set)
|
93 |
except UnicodeDecodeError as e:
|
94 |
tqdm.write(
|
95 |
f"[-] UnicodeDecodeError in {member.name}, skipping: \n{e}"
|
|
|
97 |
except SyntaxError as e:
|
98 |
tqdm.write(f"[-] SyntaxError in {member.name}, skipping: \n{e}")
|
99 |
|
100 |
+
extracted_infos.append(repo_info)
|
101 |
+
|
102 |
+
return extracted_infos
|
103 |
|
104 |
|
105 |
class RepoEmbeddingPipeline(Pipeline):
|
|
|
144 |
|
145 |
if self.st_messager:
|
146 |
self.st_messager.info("[*] Downloading and extracting repos...")
|
147 |
+
|
148 |
extracted_infos = download_and_extract(inputs, headers=self.API_HEADERS)
|
149 |
|
150 |
return extracted_infos
|
|
|
176 |
return sentence_embeddings
|
177 |
|
178 |
def _forward(self, extracted_infos, max_length=512, st_progress=None):
|
179 |
+
repo_dataset = []
|
180 |
+
num_texts = sum(len(x["funcs"]) + len(x["docs"]) for x in extracted_infos)
|
|
|
|
|
181 |
with tqdm(total=num_texts) as pbar:
|
182 |
+
for repo_info in extracted_infos:
|
183 |
+
repo_name = repo_info["name"]
|
184 |
+
entry = {
|
185 |
+
"name": repo_name,
|
186 |
+
"topics": repo_info["topics"],
|
187 |
+
"license": repo_info["license"],
|
188 |
+
"stars": repo_info["stars"],
|
189 |
+
}
|
190 |
+
|
191 |
pbar.set_description(f"Processing {repo_name}")
|
|
|
192 |
|
193 |
message = f"[*] Generating embeddings for {repo_name}"
|
194 |
tqdm.write(message)
|
|
|
229 |
else None
|
230 |
)
|
231 |
|
232 |
+
repo_dataset.append(entry)
|
233 |
|
234 |
return repo_dataset
|
235 |
|