在 Oracle 插入期间将“无”字符串转换为空

Question

我有一个具有以下值的数据框 (df)。注意下面的项目代码是字符串“None”

   YEAR    QUARTER    ID          AMOUNT       DATE       COMMENTS  PROJECT_CODE
0  2020        3     ABC            300.0   2020-10-07      N/A       ABC_0001
0  2020        3     ABC            500.0   2020-10-07      N/A          None

我想在使用以下查询插入 oracle 表之前将“无”转换为无：

con3 = cx_Oracle.connect(connstr)
cur3 = con3.cursor()
rows3 = [tuple(x) for x in df.values]
cur3.executemany('''merge into O_TABLE
                   using dual
                   on (YEAR = :1 and QUARTER = :2 and ID = :3 and AMOUNT = :4 and DATE = :5 and COMMENTS = :6 and PROJECT_CODE = :7)
                   when not matched then insert values (:1, :2, :3, :4, :5, :6, :7)
                   where :3 IS NOT NULL''',rows)
con3.commit()
cur3.close()
con3.close()

我想在 Oracle 表上插入一个空值来代替“无”。我尝试在插入之前将“无”转换为Nan，但这似乎会导致错误：

df['PROJECT_CODE'] = df['PROJECT_CODE'].replace('None', np.nan)

Error:
Traceback (most recent call last):
  File "./test_program.py", line 266, in <module>
    where :3 IS NOT NULL''',rows3)
TypeError: expecting string or bytes object

如何为我的 df 中的所有“无”值在 oracle 表中插入一个真正的空值？我不想使用 to_sql，因为这限制了我编写 sql 查询的方式。谢谢。

Answer 1

Oracle 很乐意将空字符串用作 null，并且当您将来进行比较时，空字符串等同于 null。

您可以只传入一个空字符串，但是您还需要使用“PROJECT_CODE”进行比较，或者使用

and NVL(PROJECT_CODE, '''None''') = :7

或

and ((:7 is null and PROJECT_CODE is null) or PROJECT_CODE = :7)

例如

con3 = cx_Oracle.connect(connstr)
cur3 = con3.cursor()
rows3 = [tuple(x) for x in df.values]
cur3.executemany('''merge into O_TABLE
                   using dual
                   on (YEAR = :1
                       and QUARTER = :2
                       and ID = :3
                       and AMOUNT = :4
                       and DATE = :5 
                       and COMMENTS = :6 
                       and ((:7 is null and PROJECT_CODE is null) or PROJECT_CODE = :7)
                   )
                   when not matched then insert values (:1, :2, :3, :4, :5, :6, :7)
                   where :3 IS NOT NULL''',rows)
con3.commit()
cur3.close()
con3.close()