python-3.x - 为什么 pandas.read_sql 返回一个空的 DataFrame？

Question

我正在尝试从数据库中检索数据并保存到 pandas.DataFrame 中。这是我的 Python 脚本，

conn = pyodbc.connect(sql_server)
query = '''SELECT a1, a2, a3
FROM '''  + dbschema + '''.SomeResults
WHERE FactorName = \' ''' + FactorName + ''' \' AND parametername = 'Param1' ORDER BY Factor1 '''
df = pd.read_sql(query, conn)
print(df)

然而，它返回，

Empty DataFrame
Columns: [a1, a2, a3]
Index: []

我很确定这不是 SQL 问题，因为我可以使用 conn.cursor() 从数据库中检索数据。

Answer 1

the reason is the way of generating that SQL:

In [307]: dbschema = 'db'

In [308]: FactorName = 'Factor1'

In [309]: query = '''SELECT a1, a2, a3
     ...: FROM '''  + dbschema + '''.SomeResults
     ...: WHERE FactorName = \' ''' + FactorName + ''' \' AND parametername = 'Param1' ORDER BY Factor1 '''

In [310]: print(query)
SELECT a1, a2, a3
FROM db.SomeResults
WHERE FactorName = ' Factor1 ' AND parametername = 'Param1' ORDER BY Factor1

# NOTE: spaces      ^       ^

You should not generate SQL this way, as it might be dangerous (read about SQL injections).

This would be a proper way:

query = """
SELECT a1, a2, a3
FROM {}.SomeResults
WHERE FactorName = ? AND parametername = 'Param1'
ORDER BY Factor1
"""

df = pd.read_sql(query.format(dbschema), conn, params=(FactorName,))

NOTE: only literals can be parameterized. I.e. we can NOT parameterized schema names, table names, column, names, etc.

Here is a funny example of a SQL injection: