x = null 与 x IS NULL 之间的区别
Difference between x = null vs. x IS NULL
在 Snowflake 中，条件表达式中的 x = NULL 和 x IS NULL 有什么区别？根据经验，当我想查找某些列为空白的行时， x IS NULL 就是我想要的。我问是因为 x = NULL 被视为有效语法，我很好奇这个表达式是否有不同的应用程序。
与大多数 SQL 语言一样，比较 NULL = NULL 不会 return TRUE。在 SnowFlake 中，它 returns NULL，与 NULL 值的任何比较也是如此。其原因与 SQL 的错综复杂的历史有关，并且一直在争论这是否是一个好功能。无论如何，这就是我们所拥有的。
因此，当您比较两个可能为 NULL 的值时，这里有一些您通常可以使用的不同解决方案。
-- NVL will return the second value if the first value is NULL
-- So if both of your values are NULL, then an NVL around each of them will
-- return a value so that they are both equal.
-- This only works if you know that your values will never be equal to -1 for example
SELECT ...
WHERE NVL(x, -1) = NVL(y, -1)
-- A little messier, especially among more complicated filters,
-- but guaranteed to work regardless of values
SELECT ...
WHERE x = y OR (x is null and y is null)
-- My new favorite which works in SnowFlake (thanks to @waldente)
SELECT x IS NOT DISTINCT FROM y;
-- For most SQL languages, this is a neat way to take advantage of how
-- INTERSECT compares values which does treat NULLs as equal
SELECT ...
WHERE exists (select x intersect select y)
what is the difference between x = NULL and x IS NULL
在 Snowflake 中就像在其他 RDBMS 中一样，Nothing 等于 NULL（甚至 NULL 本身），因此条件 x = NULL（这是有效的 SQL 语法) 将始终评估为 false（好吧，实际上，在大多数 RDBMS 中它评估为 NULL，这是不正确的）。请注意，对于非相等比较也是如此：即 NULL <> NULL 也是错误的。
检查变量是否为 NULL 的典型方法是使用 x IS NULL 结构，如果 x 为 NULL，则其计算结果为真。您也可以使用 x IS NOT NULL。此语法是为 NULL 保留的，因此 x IS y 是语法错误。
这里是a small demo:
select
case when 1 = null then 1 else 0 end 1_equal_null,
case when 1 <> null then 1 else 0 end 1_not_equal_null,
case when null is null then 1 else 0 end null_is_null,
case when 1 is not null then 1 else 0 end 1_is_not_null
1_equal_null | 1_not_equal_null | null_is_null | 1_is_not_null
-----------: | ---------------: | -----------: | ------------:
0 | 0 | 1 | 1
这种特殊情况在 Snowflake 的文档中有详细描述：
Compares whether two expressions are equal. The function is NULL-safe, meaning it treats NULLs as known values for comparing equality. Note that this is different from the EQUAL comparison operator (=), which treats NULLs as unknown values.
+------+------+--------------------------------+------------------------------------------+----------------------------+--------------------------------------+
| X1_I | X2_I | X1.I IS NOT DISTINCT FROM X2.I | SELECT IF X1.I IS NOT DISTINCT FROM X2.I | X1.I IS DISTINCT FROM X2.I | SELECT IF X1.I IS DISTINCT FROM X2.I |
|------+------+--------------------------------+------------------------------------------+----------------------------+--------------------------------------|
| 1 | 1 | True | Selected | False | Not |
| 1 | 2 | False | Not | True | Selected |
| 1 | NULL | False | Not | True | Selected |
| 2 | 1 | False | Not | True | Selected |
| 2 | 2 | True | Selected | False | Not |
| 2 | NULL | False | Not | True | Selected |
| NULL | 1 | False | Not | True | Selected |
| NULL | 2 | False | Not | True | Selected |
| NULL | NULL | True | Selected | False | Not |
+------+------+--------------------------------+------------------------------------------+----------------------------+--------------------------------------+
在 Snowflake 中，条件表达式中的 x = NULL 和 x IS NULL 有什么区别？根据经验，当我想查找某些列为空白的行时， x IS NULL 就是我想要的。我问是因为 x = NULL 被视为有效语法，我很好奇这个表达式是否有不同的应用程序。
与大多数 SQL 语言一样，比较 NULL = NULL 不会 return TRUE。在 SnowFlake 中，它 returns NULL，与 NULL 值的任何比较也是如此。其原因与 SQL 的错综复杂的历史有关，并且一直在争论这是否是一个好功能。无论如何，这就是我们所拥有的。
因此，当您比较两个可能为 NULL 的值时，这里有一些您通常可以使用的不同解决方案。
-- NVL will return the second value if the first value is NULL
-- So if both of your values are NULL, then an NVL around each of them will
-- return a value so that they are both equal.
-- This only works if you know that your values will never be equal to -1 for example
SELECT ...
WHERE NVL(x, -1) = NVL(y, -1)
-- A little messier, especially among more complicated filters,
-- but guaranteed to work regardless of values
SELECT ...
WHERE x = y OR (x is null and y is null)
-- My new favorite which works in SnowFlake (thanks to @waldente)
SELECT x IS NOT DISTINCT FROM y;
-- For most SQL languages, this is a neat way to take advantage of how
-- INTERSECT compares values which does treat NULLs as equal
SELECT ...
WHERE exists (select x intersect select y)
what is the difference between x = NULL and x IS NULL
在 Snowflake 中就像在其他 RDBMS 中一样，Nothing 等于 NULL（甚至 NULL 本身），因此条件 x = NULL（这是有效的 SQL 语法) 将始终评估为 false（好吧，实际上，在大多数 RDBMS 中它评估为 NULL，这是不正确的）。请注意，对于非相等比较也是如此：即 NULL <> NULL 也是错误的。
检查变量是否为 NULL 的典型方法是使用 x IS NULL 结构，如果 x 为 NULL，则其计算结果为真。您也可以使用 x IS NOT NULL。此语法是为 NULL 保留的，因此 x IS y 是语法错误。
这里是a small demo:
select
case when 1 = null then 1 else 0 end 1_equal_null,
case when 1 <> null then 1 else 0 end 1_not_equal_null,
case when null is null then 1 else 0 end null_is_null,
case when 1 is not null then 1 else 0 end 1_is_not_null
1_equal_null | 1_not_equal_null | null_is_null | 1_is_not_null
-----------: | ---------------: | -----------: | ------------:
0 | 0 | 1 | 1
这种特殊情况在 Snowflake 的文档中有详细描述：
Compares whether two expressions are equal. The function is NULL-safe, meaning it treats NULLs as known values for comparing equality. Note that this is different from the EQUAL comparison operator (=), which treats NULLs as unknown values.
+------+------+--------------------------------+------------------------------------------+----------------------------+--------------------------------------+
| X1_I | X2_I | X1.I IS NOT DISTINCT FROM X2.I | SELECT IF X1.I IS NOT DISTINCT FROM X2.I | X1.I IS DISTINCT FROM X2.I | SELECT IF X1.I IS DISTINCT FROM X2.I |
|------+------+--------------------------------+------------------------------------------+----------------------------+--------------------------------------|
| 1 | 1 | True | Selected | False | Not |
| 1 | 2 | False | Not | True | Selected |
| 1 | NULL | False | Not | True | Selected |
| 2 | 1 | False | Not | True | Selected |
| 2 | 2 | True | Selected | False | Not |
| 2 | NULL | False | Not | True | Selected |
| NULL | 1 | False | Not | True | Selected |
| NULL | 2 | False | Not | True | Selected |
| NULL | NULL | True | Selected | False | Not |
+------+------+--------------------------------+------------------------------------------+----------------------------+--------------------------------------+