Datasets:

Languages:
English
ArXiv:
License:
dp-bench / evaluation_code /dp_bench_classes.py
faisalchow's picture
Evaluation codes
116cc62 verified
from dataclasses import dataclass
import copy
from typing import *
@dataclass
class Table:
name: str
columns: Set[str]
column_types: Dict[str, str] = None
@dataclass
class DerivedColumn:
name: str
provenance_sql: str
provenance_tables: Dict[str, Table]
@dataclass
class DataBase:
name: str
tables: Dict[str, Table]
derived_columns: List[DerivedColumn] = None
def get_column_count(self, include_derived_columns: bool = False):
if include_derived_columns:
temp_self = copy.deepcopy(self)
if temp_self.derived_columns is not None:
for dc in temp_self.derived_columns:
for tabname, table in dc.provenance_tables.items():
if tabname not in temp_self.tables:
temp_self.tables[tabname] = Table(name=tabname, columns=set())
temp_self.tables[tabname].columns |= table.columns
else:
temp_self = self
return sum([len(tab.columns) for _, tab in temp_self.tables.items()])
def compare_column_overlap(self, other_db, include_derived_columns=False):
count = 0
if include_derived_columns:
temp_self = copy.deepcopy(self)
temp_other = copy.deepcopy(other_db)
if temp_self.derived_columns is not None:
for dc in temp_self.derived_columns:
for tabname, table in dc.provenance_tables.items():
if tabname not in temp_self.tables:
temp_self.tables[tabname] = Table(name=tabname, columns=set())
temp_self.tables[tabname].columns |= table.columns
if temp_other.derived_columns is not None:
for dc in temp_other.derived_columns:
for tabname, table in dc.provenance_tables.items():
if tabname not in temp_other.tables:
temp_other.tables[tabname] = Table(name=tabname, columns=set())
temp_other.tables[tabname].columns |= table.columns
else:
temp_self = self
temp_other = other_db
for tabname, table in temp_self.tables.items():
if tabname not in temp_other.tables:
continue
if tabname in other_db.tables:
count += len(table.columns.intersection(other_db.tables[tabname].columns))
return count
def get_schema_for_sqlglot(self):
return {
tabname: {
col: coltype
for col, coltype in table.column_types.items()
}
for tabname, table in self.tables.items()
}
def filter_real_table_columns(real_db: DataBase, filter_db: DataBase, debug=False):
bad_tables = []
for tabname, tab in filter_db.tables.items():
if tabname not in real_db.tables:
bad_tables.append(tabname)
else:
if debug and tab.columns-real_db.tables[tabname].columns:
print(f"bad cols ", tab.columns-real_db.tables[tabname].columns)
tab.columns = tab.columns.intersection(real_db.tables[tabname].columns)
if filter_db.derived_columns:
for dc in filter_db.derived_columns:
dc_bad_tables = []
for tabname, table in dc.provenance_tables.items():
if tabname not in real_db.tables:
dc_bad_tables.append(tabname)
else:
cols = table.columns
if debug:
if cols.intersection(real_db.tables[tabname].columns) != cols:
print(real_db.name, tabname, cols - real_db.tables[tabname].columns)
dc.provenance_tables[tabname].columns = cols.intersection(real_db.tables[tabname].columns)
for tab in dc_bad_tables:
dc.provenance_tables.pop(tab)
for tab in bad_tables:
if debug:
print(f"bad table ", tab)
filter_db.tables.pop(tab, None)
return filter_db