Function get_candidate

union/TUS/run_c_alignment.py:20–39

(table_name, lsh, folder_path, topk, Aunion, model, type='Uset', threshold=0.7)

from operator import itemgetter

def get_candidate(table_name, lsh, folder_path, topk, Aunion, model, type='Uset', threshold=0.7):#改调用query_lsh的方式
    file_path = os.path.join(folder_path, table_name)
    df = pd.read_csv(file_path)
    for column in df.columns:            
        values = df[column].to_list()
        if type == 'Uset' or type == 'Usem':
            results = query_lsh(values, lsh, type, n=128, folder_path=folder_path)
        else:
            results = query_lsh_nl(values, lsh, threshold=threshold)
        #对所有的LSH候选结果计算精确的结果
        set_col_dict = cal_precise_unionablity(type, values, results, folder_path, model)

        #把col_dict按照values的大小排序，倒序，把前几个放到集合A_union中
        sorted_col_dict = dict(sorted(set_col_dict.items(), key=itemgetter(1), reverse=True))

        #这里要不要筛选一遍？还是把所有的候选表都放进去，算表的可并性的时候再筛选
        if len(sorted_col_dict)>=topk:
            Aunion.update(list(sorted_col_dict.keys())[:topk])#这里是选top5个
        else:
            Aunion.update(list(sorted_col_dict.keys()))
#Usem的minhashLSH索引得改，改成最后一行插入的

def get_column_goodness(score, type = 'uset'):