为什么此更新不会失败？ 这是预期的行为吗？

Question

I'm trying to understand why this update doesn't fail, and instead updates the table even though the sub-query contains an incorrect column name that doesn't exist in the table it is selecting from. If you run the sub-query by itself it returns a syntax error. Shouldn't the update fail and give an error? How is this being interpreted? a Co-worker suggested it may be interpreting this as a cross-apply, is that a possiblity?

Create table MyTable(column1 nvarchar(max), column2 nvarchar(max) , exclude nvarchar(max))
Create table MyTable2(c1 varchar)

INSERT INTO MyTable (column1, column2, exclude)
VALUES
    ('Fred', 'Smith',0),
    ('John', 'Smith',0),
    ('Michael', 'Smith',0),
    ('Robert', 'Smith',0);

INSERT INTO MyTable2
  ( c1)
VALUES
  ('x' ), 
  ('y'), 
  ('w'),
  ('n')

  Update [MyTable]  set [exclude] = 1 where [column1] in (select [column1] from [MyTable2])

  drop table MyTable
  drop table MyTable2

In the Sub-query for the update, Column1 does not exist. Despite this, the update runs on every row in the table. If you run the sub-query by itself, it returns a syntax error.

The expected behavior is that the update would fail, and the transaction would be rolled back

Answer 1

Yes. Expected. [column1] is in scope in the subquery because it exists in the outer query.

Eg, for lots of queries like:

 Update [MyTable]  
 set [exclude] = 1 
 where exists (select * from [MyTable2] where c1 = [column1])

It's both legal and necessary to reference columns from the outer query in the subquery.

As @IłyaBursov points out, this is one of the many reasons to always prefix your column references with table aliases.

eg:

 Update [MyTable] t  
 set t.[exclude] = 1 
 where t.[column1] in (select t2.[column1] from [MyTable2] t2)