def get_blame_test():
git_repo = "/Users/aniketp/Workspace/Capstone/Source/eclipse/eclipse.platform.runtime"
sample_file = "bundles/org.eclipse.core.jobs/src/org/eclipse/core/internal/jobs/JobManager.java"
exec_dir = os.getcwd()
repo = Repository(os.path.join(git_repo, git_folder))
os.chdir(git_repo)
blameObj = repo.blame(sample_file, min_line=1, max_line=20)
def map(
self,
repo: Repository,
file_name: str,
line: int,
commit: Commit) -> str:
blamer = repo.blame(file_name, newest_commit=commit.parent_ids[0])
hunk = blamer.for_line(line)
return str(hunk.orig_commit_id)
class CollectGit(object):
"""
Small Helper class for small repositories.
This does not scale because we hold a lot of data in memory.
"""
_regex_comment = re.compile(
r"(//[^\"\n\r]*(?:\"[^\"\n\r]*\"[^\"\n\r]*)*[\r\n]|/\*([^*]|\*(?!/))*?\*/)(?=[^\"]*(?:\"[^\"]*\"[^\"]*)*$)"
)
_regex_jdoc_line = re.compile(r"(- |\+)\s*(\*|/\*).*")
def __init__(self, path):
if not path.endswith('.git'):
if not path.endswith('/'):
path += '/'
path += '.git'
self._log = logging.getLogger(self.__class__.__name__)
self._path = path
self._repo = Repository(self._path)
self._hunks = {}
self._file_actions = {}
self._bugfix = {}
self._msgs = {}
self._days = {}
self._cdays = {}
self._branches = {}
self._tags = {}
self._dopts = GIT_DIFF_FIND_RENAMES | GIT_DIFF_FIND_COPIES
self._SIMILARITY_THRESHOLD = 50
self._graph = nx.DiGraph()
@classmethod
def clone_repo(cls, uri, local_path):
project_name = uri.split('/')[-1].split('.git')[0]
repo_path = local_path + '/' + project_name + '/'
if os.path.isdir(repo_path):
c = subprocess.run(['git', 'fetch'],
cwd=repo_path,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
if c.returncode != 0:
err = 'Error pulling repository {} to {}'.format(
uri, repo_path)
raise Exception(err)
else:
os.mkdir(repo_path)
c = subprocess.run(['git', 'clone', uri, repo_path],
cwd=repo_path,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
if c.returncode != 0:
err = 'Error cloning repository {} to {}'.format(
uri, repo_path)
raise Exception(err)
return repo_path
def _changed_lines(self, hunk):
added_lines = []
deleted_lines = []
del_line = hunk['old_start']
add_line = hunk['new_start']
for line in hunk['content'].split('\n'):
tmp = line[1:].strip()
# is_comment = tmp.startswith('//') or tmp.startswith('/*') or tmp.startswith('*')
if line.startswith('+'):
added_lines.append((add_line, tmp))
del_line -= 1
if line.startswith('-'):
deleted_lines.append((del_line, tmp))
add_line -= 1
del_line += 1
add_line += 1
return added_lines, deleted_lines
def _comment_only_change(self, content):
content = content + '\n' # required for regex to drop comments
content = re.sub(self._regex_comment, "", content)
removed = ''
added = ''
for line in content.split('\n'):
line = re.sub(
r"\s+", " ", line, flags=re.UNICODE
) # replace all kinds of whitespaces (also multiple) with sińgle whitespace
if not re.match(self._regex_jdoc_line, line):
if line.startswith('-'):
removed += line[1:].strip()
elif line.startswith('+'):
added += line[1:].strip()
return removed == added
def _blame_lines(self,
revision_hash,
filepath,
strategy,
ignore_lines=False,
validated_bugfix_lines=False):
"""We want to find changed lines for one file in one commit (from the previous commit).
For this we are iterating over the diff and counting the lines that are deleted (changed) from the original file.
We ignore all added lines.
ignore_lines is already specific to all changed hunks of the file for which blame_lines is called
"""
c = self._repo.revparse_single('{}'.format(revision_hash))
self._hunks[revision_hash] = self._get_hunks(c)
changed_lines = []
if revision_hash not in self._hunks.keys(
) or not self._hunks[revision_hash]:
return changed_lines
for h in self._hunks[revision_hash]:
if h['new_file'] != filepath:
continue
# only whitespace or comment changes in the hunk, ignore
if strategy == 'code_only' and self._comment_only_change(
h['content']):
self._log.debug(
'detected whitepace or comment only change in {} for {}'.
format(revision_hash, filepath))
continue
added, deleted = self._changed_lines(h)
for dt in deleted:
if dt not in changed_lines and dt[1]:
if strategy == 'code_only' and dt[1].startswith(
('//', '/*', '*')):
continue
# we may only want validated lines
if validated_bugfix_lines is not False:
if dt[0] not in validated_bugfix_lines:
continue
# we may ignore lines, e.g., refactorings
if ignore_lines:
ignore = False
for start_line, end_line in ignore_lines:
if start_line <= dt[0] <= end_line:
ignore = True
break
# if we hit the line in our ignore list we continue to the next
if ignore:
# self._log.warn('ignore line {} in file {} in commit {} because of refactoring detection'.format(dt[0], filepath, revision_hash))
continue
changed_lines.append(dt)
return changed_lines
def blame(self,
revision_hash,
filepath,
strategy='code_only',
ignore_lines=False,
validated_bugfix_lines=False):
"""Collect a list of commits where the given revision and file were last changed.
Uses git blame.
:param str revision_hash: Commit for which we want to collect blame commits.
:param str filepath: File for which we want to collect blame commits.
:rtype: list
:returns: A list of tuples of blame commits and the original file for the given parameters.
"""
commits = []
# - ignore if commit is not in graph
if revision_hash not in self._graph:
return []
# # - ignore package-info.java
# if strategy == 'code_only' and filepath.lower().endswith('package-info.java'):
# self._log.debug('skipping blame on revision: {} for file {} because it is package-info.java'.format(revision_hash, filepath))
# return []
# # - ignore test/ /test/ example/ examples/
# if strategy == 'code_only' and re.match(self._regex_test_example, filepath):
# self._log.debug('skipping blame on revision: {} for file {} because it is a test or an example'.format(revision_hash, filepath))
# return []
# bail on multiple parents
parents = list(self._graph.predecessors(revision_hash))
if len(parents) > 1:
self._log.debug(
'skipping blame on revision: {} because it is a merge commit'.
format(revision_hash))
return []
changed_lines = self._blame_lines(revision_hash, filepath, strategy,
ignore_lines, validated_bugfix_lines)
parent_commit = self._repo.revparse_single('{}^'.format(revision_hash))
blame = self._repo.blame(filepath,
flags=GIT_BLAME_TRACK_COPIES_SAME_FILE,
newest_commit=parent_commit.hex)
for lineno, line in changed_lines:
# returns blamehunk for specific line
try:
bh = blame.for_line(lineno)
except IndexError as e:
# this happens when we have the wrong parent node
bla = 'tried to get file: {}, line: {}, revision: {}, blame commit: {}'.format(
filepath, lineno, revision_hash, str(bh.orig_commit_id))
self._log.error(bla)
raise # this is critical
inducing_commit = self._repo.revparse_single(str(
bh.orig_commit_id))
# start = bh.orig_start_line_number
# lines = bh.lines_in_hunk
# final_start = bh.final_start_line_number
# print(revision_hash, '->', inducing_commit.hex)
# print('original: {}: {}'.format(lineno, line))
# print('{},{}: {},{}'.format(start, lines, final_start, lines))
# blame_lines = []
# for hunk in self._hunks[inducing_commit.hex]:
# if hunk['new_file'] != bh.orig_path:
# continue
# ls = final_start
# for i, blame_line in enumerate(hunk['content'].split('\n')):
# if blame_line[1:].strip() and line[1:].strip() and blame_line[1:] == line[1:]:
# print('blame: {}:{}'.format(ls, blame_line))
# ls += 1
commits.append((inducing_commit.hex, bh.orig_path))
# make unique
return list(set(commits))
def commit_information(self, revision_hash):
obj = self._repo.get(revision_hash)
return {
'author_name':
obj.author.name,
'author_email':
obj.author.email,
'committer_name':
obj.committer.name,
'committer_email':
obj.committer.email,
'committer_date_utc':
datetime.fromtimestamp(obj.commit_time, tz=timezone.utc),
'committer_date':
obj.commit_time,
'committer_date_offset':
obj.commit_time_offset,
'message':
obj.message,
'file_actions':
self._file_actions[revision_hash]
}
def file_actions(self, revision_hash):
return self._file_actions[revision_hash]
def all_files(self, revision_hash):
# 1. checkout repo
self._checkout_revision(revision_hash)
# 2. list files
return self._list_files()
def first_occurence(self, filename):
# file rename tracking is not possible currently in libgit, see:
# https://github.com/libgit2/libgit2/issues/3041
# find first occurence of file with git cli
# git log --follow --diff-filter=A --find-renames=40% foo.js
path = self._path.replace('.git', '')
c = subprocess.run([
'git', 'log', '--all', '--pretty=tformat:"%H %ci"', '--follow',
'--diff-filter=A', '--find-renames=80%', '--', filename
],
cwd=path,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
if c.returncode != 0:
err = 'Error finding first occurrence of file: {}'.format(filename)
self._log.error(err)
self._log.error(c.stderr)
raise Exception(err)
full = c.stdout.decode('utf-8')
try:
first_line = full.split('\n')[-2]
except IndexError:
if not full:
print('no git log for file {}'.format(filename))
print(full)
raise
first_date = ' '.join(first_line.split(' ')[1:]).replace('"', '')
dt = datetime.strptime(
first_date, '%Y-%m-%d %H:%M:%S %z'
) # we can do this here because we control the input format, %z does not cover +01:00 just +100 (at least in 3.6)
return dt
def tags(self):
regex = re.compile('^refs/tags')
ret = []
for tagref in filter(lambda r: regex.match(r),
self._repo.listall_references()):
tag = self._repo.lookup_reference(tagref)
target = self._repo.lookup_reference(tagref).peel()
ret.append({
'name': tag.name.replace('refs/tags/', ''),
'revision_hash': target.id
})
return ret
def _checkout_revision(self, revision):
"""Checkout via shell, we ignore stdout output."""
path = self._path.replace('.git', '')
c = subprocess.run(['git', 'checkout', '-q', '-f', revision],
cwd=path,
stdout=subprocess.PIPE)
return c.returncode == 0
def _list_files(self):
"""The slower list_files"""
path = self._path.replace('.git', '')
ret = []
for root, dirs, files in os.walk(path):
for file in files:
filepath = os.path.join(root, file)
relative_filepath = filepath.replace(path, '')
ret.append(relative_filepath)
return ret
def _list_files2(self):
"""The faster list_files (relies on find command)"""
path = self._path.replace('.git', '')
lines = subprocess.check_output(['find', '.', '-iname', '*.java'],
cwd=path)
files = []
for f in lines.decode('utf-8').split('\n'):
if f.lower().endswith('.java'):
files.append(f.replace('./', ''))
return files
def _get_hunks(self, commit):
diffs = []
hunks = []
# for initial commit (or orphan commits) pygit2 needs some special attention
initial = False
if not commit.parents:
initial = True
diffs.append((None,
commit.tree.diff_to_tree(context_lines=0,
interhunk_lines=1)))
# we may have multiple parents (merge commit)
for parent in commit.parents:
# we need all information from each parent because in a merge each parent may add different files
tmp = self._repo.diff(parent,
commit,
context_lines=0,
interhunk_lines=1)
tmp.find_similar(self._dopts, self._SIMILARITY_THRESHOLD,
self._SIMILARITY_THRESHOLD)
diffs.append((parent.hex, tmp))
for parent, diff in diffs:
checked_paths = set()
for patch in diff:
if patch.delta.new_file.path in checked_paths:
self._log.warn('already have {} in checked_paths'.format(
patch.delta.new_file.path))
continue
mode = 'X'
if patch.delta.status == 1:
mode = 'A'
elif patch.delta.status == 2:
mode = 'D'
elif patch.delta.status == 3:
mode = 'M'
elif patch.delta.status == 4:
mode = 'R'
elif patch.delta.status == 5:
mode = 'C'
elif patch.delta.status == 6:
mode = 'I'
elif patch.delta.status == 7:
mode = 'U'
elif patch.delta.status == 8:
mode = 'T'
# diff to tree gives D for inital commit otherwise
if initial:
mode = 'A'
# we may have hunks to add
if patch.hunks and commit.hex not in self._hunks.keys():
self._hunks[commit.hex] = []
# add hunks
for hunk in patch.hunks:
# initial is special case
if initial:
content = ''.join(
['+' + l.content for l in hunk.lines])
hunks.append({
'header': hunk.header,
'new_file': patch.delta.new_file.path,
'new_start': hunk.old_start,
'new_lines': hunk.old_lines,
'old_start': hunk.new_start,
'old_lines': hunk.new_lines,
'content': content
})
else:
content = ''.join(
[l.origin + l.content for l in hunk.lines])
hunks.append({
'header': hunk.header,
'new_file': patch.delta.new_file.path,
'new_start': hunk.new_start,
'new_lines': hunk.new_lines,
'old_start': hunk.old_start,
'old_lines': hunk.old_lines,
'content': content
})
return hunks
def _changed_files(self, commit):
changed_files = []
diffs = []
# for initial commit (or orphan commits) pygit2 needs some special attention
initial = False
if not commit.parents:
initial = True
diffs.append((None,
commit.tree.diff_to_tree(context_lines=0,
interhunk_lines=1)))
# we may have multiple parents (merge commit)
for parent in commit.parents:
# we need all information from each parent because in a merge each parent may add different files
tmp = self._repo.diff(parent,
commit,
context_lines=0,
interhunk_lines=1)
tmp.find_similar(self._dopts, self._SIMILARITY_THRESHOLD,
self._SIMILARITY_THRESHOLD)
diffs.append((parent.hex, tmp))
for parent, diff in diffs:
checked_paths = set()
for patch in diff:
if patch.delta.new_file.path in checked_paths:
self._log.warn('already have {} in checked_paths'.format(
patch.delta.new_file.path))
continue
mode = 'X'
if patch.delta.status == 1:
mode = 'A'
elif patch.delta.status == 2:
mode = 'D'
elif patch.delta.status == 3:
mode = 'M'
elif patch.delta.status == 4:
mode = 'R'
elif patch.delta.status == 5:
mode = 'C'
elif patch.delta.status == 6:
mode = 'I'
elif patch.delta.status == 7:
mode = 'U'
elif patch.delta.status == 8:
mode = 'T'
# diff to tree gives D for inital commit otherwise
if initial:
mode = 'A'
# we may have hunks to add
if patch.hunks and commit.hex not in self._hunks.keys():
self._hunks[commit.hex] = []
# add hunks
for hunk in patch.hunks:
# initial is special case
if initial:
content = ''.join(
['+' + l.content for l in hunk.lines])
self._hunks[commit.hex].append({
'header':
hunk.header,
'new_file':
patch.delta.new_file.path,
'new_start':
hunk.old_start,
'new_lines':
hunk.old_lines,
'old_start':
hunk.new_start,
'old_lines':
hunk.new_lines,
'content':
content
})
else:
content = ''.join(
[l.origin + l.content for l in hunk.lines])
self._hunks[commit.hex].append({
'header':
hunk.header,
'new_file':
patch.delta.new_file.path,
'new_start':
hunk.new_start,
'new_lines':
hunk.new_lines,
'old_start':
hunk.old_start,
'old_lines':
hunk.old_lines,
'content':
content
})
# collect line stats
if initial:
fa = {
'lines_added': patch.line_stats[2],
'lines_deleted': patch.line_stats[1],
'changeset_size': len(diff),
'parent': None
}
else:
fa = {
'lines_added': patch.line_stats[1],
'lines_deleted': patch.line_stats[2],
'changeset_size': len(diff),
'parent': parent
}
#if mode == 'R':
# print('R {} -> {}, sim: {}'.format(patch.delta.old_file.path, patch.delta.new_file.path, patch.delta.similarity))
if mode in ['C', 'R']:
changed_file = [
mode, patch.delta.new_file.path,
patch.delta.old_file.path, fa
]
else:
changed_file = [mode, patch.delta.new_file.path, None, fa]
checked_paths.add(patch.delta.new_file.path)
changed_files.append(changed_file)
return changed_files
def collect(self):
# list all branches
for branch in list(self._repo.branches):
self._collect_branch(branch)
# list all tags
for obj in self._repo:
tag = self._repo[obj]
if tag.type == GIT_OBJ_TAG:
self._collect_branch(tag, is_tag=True)
return self._graph
def _collect_branch(self, branch, is_tag=False):
if type(branch) == str:
branch = self._repo.branches[branch]
# add nodes to graph
try:
for c in self._repo.walk(branch.target):
self._graph.add_node(c.hex)
# branch stuff, used for traversing backwards for tags in svn->git conversions
# if c.hex not in self._branches.keys():
# self._branches[c.hex] = []
# what about tags which are also on branches?
# if is_tag:
# self._tags[c.hex] = branch.name
# else:
# self._branches[c.hex].append(branch.branch_name)
# add msg
# self._msgs[c.hex] = c.message
# add days, we use this later for lookup
# day = str(datetime.fromtimestamp(c.commit_time, tz=timezone.utc).date())
# if day not in self._days.keys():
# self._days[day] = []
# self._days[day].append(c.hex)
# add for convenience for OntdekBaanBfs
# self._cdays[c.hex] = day
# add changed files per node
# if c.hex not in self._file_actions.keys():
# self._file_actions[c.hex] = self._changed_files(c)
# still too expensive
# self._create_hunks(c)
# add edges to graph
for c in self._repo.walk(branch.target):
for p in c.parents:
self._graph.add_edge(p.hex, c.hex)
except ValueError as e:
pass
def __fetch_blamed_commits(
self, szz_hunks,
git_repo: pygit2.Repository) -> Dict[str, BlamedCommit]:
log.info("Process %d starts blame_commit_process", rank)
start = time.time()
blamed_commits = {}
contributors = {}
blame_counter = {}
for hunk in szz_hunks:
line_labels = hunk.line_labels
for bh in git_repo.blame(
hunk.patch.old_file,
newest_commit=hunk.patch.commit.sha_parent,
min_line=hunk.old_start,
max_line=hunk.old_start + hunk.old_lines - 1):
blamed_sha = str(bh.final_commit_id)
if blamed_sha not in blamed_commits:
try:
blamed_commit = CommitWrapper(
git_repo.revparse_single(blamed_sha))
blamed_parents = blamed_commit.parents
blamed_num_parents = len(blamed_parents)
if not blamed_num_parents:
ins = None
dels = None
num_files = None
else:
blamed_diff = blamed_commit.diff(git_repo)
ins = blamed_diff.stats.insertions
dels = blamed_diff.stats.deletions
num_files = blamed_diff.stats.files_changed
if num_files is None or num_files >= self.__max_num_files_changed:
continue
if ins and ins >= self.__max_new_lines:
continue
blamed_authored_datetime = blamed_commit.authored_date
(blamed_author_name,
blamed_author_email) = blamed_commit.author
(blamed_author_name_l,
blamed_author_email_l) = (blamed_author_name.lower(),
blamed_author_email.lower())
(blamed_committer_name,
blamed_committer_email) = blamed_commit.committer
(blamed_committer_name_l, blamed_committer_email_l) = (
blamed_committer_name.lower(),
blamed_committer_email.lower())
if (blamed_author_name_l,
blamed_author_email_l) not in contributors:
blamed_author_id = Szz.__hash_values(
self, blamed_author_name_l,
blamed_author_email_l)
contributors[blamed_author_id] = (
blamed_author_name_l, blamed_author_email_l)
if (blamed_committer_name_l,
blamed_committer_email_l) not in contributors:
blamed_committer_id = Szz.__hash_values(
self, blamed_committer_name_l,
blamed_committer_email_l)
contributors[blamed_committer_id] = (
blamed_committer_name_l,
blamed_committer_email_l)
blamed_message = blamed_commit.message
blamed_first_msg_line = blamed_message.split('\n')[0]
# get info about changes to src files in the new blamed commit
all_files, src_files, num_src_files_touched, src_loc_added, src_loc_deleted = \
CommitWrapper.get_src_changes(self.__basic_classifier,
blamed_commit.diff(git_repo))
blamed_commit = BlamedCommit(
blamed_sha, blamed_authored_datetime,
blamed_author_id, blamed_committer_id,
blamed_first_msg_line, blamed_num_parents, ins,
dels, num_files, all_files, src_loc_added,
src_loc_deleted, num_src_files_touched, src_files)
blamed_commits[blamed_sha] = (blamed_commit, hunk)
except Exception as e:
log.error(msg="{0}: revparse error {1}:\t{2}".format(
self.__repo_path, blamed_sha, e))
traceback.print_exc()
for line_num in range(
bh.final_start_line_number,
bh.final_start_line_number + bh.lines_in_hunk):
if line_labels[
line_num] == self.__basic_classifier.CG_CODE:
blame_counter.setdefault(blamed_sha, 0)
blame_counter[blamed_sha] += 1
blames = []
for blamed_sha, num_lines in blame_counter.items():
blamed_commit, hunk = blamed_commits[blamed_sha]
blames.append(
Blame(hunk.patch.commit.sha, hunk.patch.old_file,
hunk.patch.label, blamed_commit, num_lines))
result_contributors = [
SzzContributor(key, value[0], value[1])
for key, value in contributors.items()
]
log.info("Process %d give %d blames", rank, len(blames))
Szz.__log_processing_time(
self, "Process %d blame_commit elapsed time" % rank, start)
return blames, result_contributors
