dict

squall.py

@@ -166,7 +166,12 @@ class Squall(datasets.GeneratorBasedBuilder):
                     tmp.append(sample[sample_key][j][k])
                 cols[keys[k]] = tmp
             return cols
-        
+
+        columns_keys = ["raw_header", "tokenized_header", "column_suffixes", "column_dtype", "example"]
+        sql_keys = ["sql_type", "value", "span_indices"]
+        align_keys = ["nl_indices", "sql_indices"]
+        ralign_keys = ["aligned_sql_token_type", "aligned_sql_token_info"]
+
         if split_key != 'test':
             with open(squall_full, encoding="utf-8") as f:
                 squall_full_data = json.load(f)
@@ -256,17 +261,13 @@ class Squall(datasets.GeneratorBasedBuilder):
             idx = 0
             for sample in set:
                 # transform columns
-                keys = ["raw_header", "tokenized_header", "column_suffixes", "column_dtype", "example"]
-                cols = transform(sample, "columns", keys)
+                cols = transform(sample, "columns", columns_keys)
                 # transform sql
-                keys = ["sql_type", "value", "span_indices"]
-                sqls = transform(sample, "sql", keys)
+                sqls = transform(sample, "sql", sql_keys)
                 # transform align
-                keys = ["nl_indices", "sql_indices"]
-                aligns = transform(sample, "align", keys)
+                aligns = transform(sample, "align", align_keys)
                 # transform ralign
-                keys = ["aligned_sql_token_type", "aligned_sql_token_info"]
-                raligns = transform(sample, "nl_ralign", keys)
+                raligns = transform(sample, "nl_ralign", ralign_keys)
                 yield idx, {
                     "nt": sample["nt"],
                     "tbl": sample["tbl"],
@@ -298,12 +299,12 @@ class Squall(datasets.GeneratorBasedBuilder):
                     "nl": sample["nl"],
                     "nl_pos": sample["nl_pos"],
                     "nl_ner": sample["nl_ner"],
-                    "nl_ralign": {},
+                    "nl_ralign": {k:[] for k in ralign_keys},
                     "nl_incolumns": sample["nl_incolumns"],
                     "nl_incells": sample["nl_incells"],
                     "columns_innl": sample["columns_innl"],
                     "tgt": '',
-                    "sql": {},
-                    "align": {}
+                    "sql": {k:[] for k in sql_keys},
+                    "align": {k:[] for k in align_keys}
                 }
                 idx += 1