(query, candiate, model, type = 'cal')
| 123 | |
| 124 | |
| 125 | def alignment_process(query, candiate, model, type = 'cal'): |
| 126 | #query和candiate分别是两个表，然后需要提出他们的column当作节点 |
| 127 | def compute_score(args): |
| 128 | s_name, t_name, s, t, s_sem, t_sem, m = args |
| 129 | logging.info(f"Processing: {s_name, t_name}") |
| 130 | score = compute_ensemble_score(s, t, s_sem, t_sem, m) |
| 131 | return s_name, t_name, score |
| 132 | |
| 133 | df_query = pd.read_csv('benchmark/' + query) |
| 134 | df_cand = pd.read_csv('benchmark/' + candiate) |
| 135 | df_query_sem = pd.read_csv('UsemLshTest/' + query) |
| 136 | df_cand_sem = pd.read_csv('UsemLshTest/'+ candiate) |
| 137 | G = nx.Graph() |
| 138 | query_columns = list(df_query.columns) |
| 139 | query_columns = [query+' '+l for l in query_columns] |
| 140 | cand_columns = list(df_cand.columns) |
| 141 | cand_columns = [candiate+' '+l for l in cand_columns] |
| 142 | G.add_nodes_from(query_columns, bipartite=0) |
| 143 | G.add_nodes_from(cand_columns, bipartite=1) |
| 144 | max_c = min(len(query_columns), len(cand_columns))#最大的c |
| 145 | |
| 146 | task_list = [] |
| 147 | for attr_s in query_columns: |
| 148 | s_name = attr_s.split(maxsplit=1)[1] |
| 149 | s = df_query[s_name] |
| 150 | s_sem = df_query_sem[s_name] |
| 151 | for attr_t in cand_columns: |
| 152 | t_name = attr_t.split(maxsplit=1)[1] |
| 153 | t = df_cand[t_name] |
| 154 | t_sem = df_cand_sem[t_name] |
| 155 | |
| 156 | task_list.append((attr_s, attr_t, s, t, s_sem, t_sem, model)) |
| 157 | |
| 158 | with ProcessPoolExecutor(max_workers = 4) as executor: |
| 159 | results = executor.map(compute_score, task_list) |
| 160 | |
| 161 | for s_name, t_name, score in results: |
| 162 | G.add_edge(s_name, t_name, weight=score) |
| 163 | |
| 164 | matching = {} |
| 165 | edges = sorted(G.edges(data=True), key=lambda x: x[2]['weight'], reverse=True) |
| 166 | max_c_scores = {} |
| 167 | c = 1 |
| 168 | mul = 1 |
| 169 | for u, v, d in edges: |
| 170 | if u in matching or v in matching.values(): |
| 171 | continue |
| 172 | matching[u] = [v, d['weight']] |
| 173 | mul = mul * d['weight'] |
| 174 | max_c_scores[c] = mul |
| 175 | c += 1 |
| 176 | if len(matching) == max_c: |
| 177 | break |
| 178 | if type == 'distribution': |
| 179 | return max_c_scores |
| 180 | |
| 181 | goodness_score = 0 |
| 182 | best_c = 1 |
no test coverage detected