def query_issue(scan_output, max_depth=4):
"""
根据 已经排序过的app相似度降序列表 scan_output 搜索所有可能 issue
~1分钟得出结果
:param scan_output: 格式参考 descript（）函数的输出
:param max_depth: 限制搜索深度，取最相似的前几个
:return: 所有查询
"""
# TODO 查询的 key 哪里出来的？
logger = logging.getLogger("StreamLogger")
rdb = issuedb.ISSuedb()
sql = """select issue_num, comments, state, title, body, commit_id, labels from {}
order by length(body) desc"""
overall_table = {}
# 所有相关app和item
for i in range(min(len(scan_output), max_depth)):
one_dict = {}
app = scan_output[i][0]
one_dict['sim'] = scan_output[i][1]
tab_name = table2tsv.file2table(app)
one_dict['data'] = []
one_dict['keys'] = []
score_list = scan_output[i][2]
keys_sea = _filter_search_keys(score_list, threshold=0.7)
logger.debug(f"{app}\t{tab_name}\tsimilar keys length: {len(keys_sea)}")
output = rdb.db_retrieve(sql.format(tab_name))
head = ["issue_num", "comments", "state", "title", "body", "commit_id", "labels"]
f_output = issuedb.retrieve_formatter(head, output)
title_list = util.get_col(output, head.index('title'))
body_list = util.get_col(output, head.index('body'))
label_list = util.get_col(output, head.index('labels'))
reply_list = util.get_col(output, head.index('issue_num'))
pre_calc_val = _pre_calc(title_list=title_list,
body_list=body_list,
label_list=label_list,
reply_list=reply_list,
keys_sea=keys_sea)
for k in keys_sea:
keys = []
for i in k:
keys.append(" ".join(i))
keys = " ".join(keys)
ess_keys = nlp_util.stem_sentence(keys)
tmp = search_rank.sort_candidate_seq(f_output, ess_keys, pre_calc_val)
leng = min(3, len(tmp))
one_dict['keys'].extend([ess_keys] * leng)
one_dict['data'].extend(tmp[:leng])
overall_table[tab_name] = one_dict
logger.debug(pp.pformat(overall_table))
logger.debug("#" * 50)
return overall_table
def descript(query_decp, source_category, except_files=None,extend=False, pool_size=32):
"""
生成描述文件
~1分钟得出结果
:param query_decp: 描述文件矩阵
example line: xml_file_name, class_name, element_name
:param except_files: 排除文件关键词，接受字符串或字符串数组
:param pool_size: 并行池大小
:return: a tuple. 得到src app与 数据库每个app的总相似度，按照相似度降序排列. 用作 搜索 app
"""
query_decp = nlp_util.process_xsv(query_decp)
if extend :
src_dir = work_path.in_project('./model/data/description_extend_all')
else:
src_dir = work_path.in_project('./model/data/description')
print("PATH!!!! {}".format(src_dir))
logger = logging.getLogger("StreamLogger")
file_list = os.listdir(src_dir)
file_list = [os.path.join(src_dir, f) for f in file_list]
if except_files is not None:
tmp = []
rms = []
if type(except_files) == str:
for i in file_list:
if except_files not in i:
tmp.append(i)
else:
rms.append(i)
elif type(except_files) == list or type(except_files) == set:
except_files = set(except_files)
for i in file_list:
flag = False
for j in except_files:
if j in i:
flag = True
break
if not flag:
tmp.append(i)
else:
rms.append(i)
logger.debug(pp.pformat(rms))
file_list = tmp
logger.debug(pp.pformat(file_list))
scan_output = _scan_match(source_category, query_decp, file_list, match_name.ngram_compare, [1, 0.5, 0.5],
threshold=0.7,
pool_size=pool_size)
# 得到src app与 数据库每个app的总相似度，按照相似度降序排列。
# tuple(
# str "参考APP描述文件名",
# float "APP相似度",
# list "参考APP的组件相似度" [(请求app组件, 参考app组件，组件相似度)]
# )
logger.debug(pp.pformat(util.get_col(scan_output, [0, 1])))
return scan_output
def test():
# import search_rank
# import matplotlib.pyplot as plt
# import numpy as np
#
# bins = np.linspace(0, 1, 100)
# plt.hist(util.get_col(list_score, 2), bins, density=True, histtype='step', cumulative=-1, label='Empirical')
# plt.show()
path = "tsv/"
filelist = os.listdir(path)
filelist.sort(key=lambda x: x.lower())
out_dict = dict()
count = 0
for file in filelist:
count += 1
full_path = os.path.join(path, file)
print(full_path)
tmp_out = util.read_tsv(full_path)
out_dict[file] = nlp_util.process_tsv(tmp_out)
print("file count", count)
count = 0
score_distribute_list = []
for i in range(len(filelist)):
i_file = filelist[i]
for j in range(len(filelist)):
if i == j:
print("Ignore same file", i_file)
continue
count += 1
j_file = filelist[j]
name = "{}^{}".format(i_file, j_file)
if len(out_dict[i_file]) == 0 or len(out_dict[j_file]) == 0:
print("EMPTY", name)
list_score = []
continue
else:
list_score = weight_compare_list(out_dict[i_file],
out_dict[j_file],
ngram_compare)
score_col = util.get_col(list_score, 2)
score_distribute_list.append((name, copy.deepcopy(score_col)))
print("ADD", count, name)
util.save_json(score_distribute_list, "score_distribute_list.json")
def query_issue(scan_output, max_depth=4):
"""
根据 已经排序过的app相似度降序列表 scan_output 搜索所有可能 issue
~1分钟得出结果
:param scan_output: 格式参考 descript（）函数的输出
:param max_depth: 限制搜索深度，取最相似的前几个
:return: 所有查询
"""
# TODO 查询的 key 哪里出来的？
logger = logging.getLogger("StreamLogger")
rdb = issuedb.ISSuedb()
# sql = """select review_id,content,bold,star_num,helpful_num,reply_content from {}
# order by length(content) desc"""
sql = """select review_id,content,star_num from {}
order by length(content) desc"""
overall_table = {}
# 所有相关app和item
for i in range(min(len(scan_output), max_depth)):
one_dict = {}
app = scan_output[i][0]
one_dict['sim'] = scan_output[i][1] # similarity_score
# print(app)
tab_name = table2tsv.file2table(app) # suppose running
print("@@@@@@@@@@@@")
print(tab_name)
one_dict['data'] = []
one_dict['keys'] = []
score_list = scan_output[i][2]
keys_sea = _filter_search_keys(score_list, threshold=0.7)
logger.debug(f"{app}\t{tab_name}\tsimilar keys length: {len(keys_sea)}")
output = rdb.db_retrieve(sql.format(tab_name)) # output is list of tuple
# head = ["review_id", "content", "bold", "star_num", "helpful_num", "reply_content"]
head = ["review_id", "content", "star_num"]
f_output = issuedb.retrieve_formatter(head, output)
# title_list = util.get_col(output, head.index('title'))
body_list = util.get_col(output, head.index('content'))
star_list = util.get_col(output, head.index('star_num'))
# reply_list = util.get_col(output, head.index('reply_content'))
# bold_list = util.get_col(output, head.index('bold'))
# label_list = util.get_col(output, head.index('labels'))
# reply_list = util.get_col(output, head.index('issue_num'))
pre_calc_val = _pre_calc(body_list=body_list, keys_sea=keys_sea) # suppose running
for k in keys_sea:
keys = []
for i in k:
keys.append(" ".join(i))
keys = " ".join(keys)
ess_keys = nlp_util.stem_sentence(keys)
tmp = search_rank.sort_candidate_seq(f_output, ess_keys, pre_calc_val)
leng = min(3, len(tmp))
one_dict['keys'].extend([ess_keys] * leng)
one_dict['data'].extend(tmp[:leng])
overall_table[tab_name] = one_dict
logger.debug(pp.pformat(overall_table))
logger.debug("#" * 50)
return overall_table
for k in count_dict:
score += count_dict[k]
return score
if __name__ == '__main__':
eee = "it.feio.android.omninotes"
s = time.strftime("%Y-%m-%d-%H:%M:%S", time.localtime(time.time()))
test = util.read_csv("model/data/description_extend_all/"+eee+".csv")
print("begin search similar apps")
scan_output = descript(test, source_category="Productivity",
except_files=eee,extend=True, pool_size=32) # get similar app
print("begin rank reviews")
rank_result = rank_review(scan_output)
print(util.get_col(scan_output, [0, 1]))
now = time.strftime("%Y-%m-%d-%H_%M_%S", time.localtime(time.time()))
# 1. 创建文件对象
z = open(csv_path +eee+ now + ".csv", 'w', encoding='utf-8', newline='')
# # 2. 基于文件对象构建 csv写入对象
csv_writer = csv.writer(z)
# # 3. 构建列表头
csv_writer.writerow(["app_id", "score", "star_num", "helpful_num", "review_content"])
for i in rank_result:
# # 写入文件
csv_writer.writerow([i[0], i[1], i[2].star_num, i[2].helpful_num, i[2].content])
# 5. 关闭文件
z.close()
print("end.")
print(s)
print(time.strftime("%Y-%m-%d-%H:%M:%S", time.localtime(time.time())))