[
https://issues.apache.org/jira/browse/DERBY-3355?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12563599#action_12563599
]
Bryan Pendleton commented on DERBY-3355:
----------------------------------------
ALTER TABLE conditionally checks the not null constraint. It first looks
to see if the table is empty or not, by calling getSemiRowCount, which
classifies the table into having 0, 1, or >1 rows. Then, if the table has
at least 1 row, it generates and executes the problematic SQL with
the unquoted column name.
So the first invocation in the repro script succeeds because the table is empty.
The code in question is around line 2420 of AlterTableConstantAction.java:
if (cd.getType().isNullable())
{
if (numRows > 0)
{
// already found a nullable column so add "AND"
if (foundNullable)
constraintText.append(" AND ");
constraintText.append("\"" + columnNames[colCtr] +
"\" IS NOT NULL ");
}
foundNullable = true;
nullCols[colCtr] = true;
}
The first time through the repro script, numRows is 0.
> Alter Column ... NULL ignores double quotes around column name
> --------------------------------------------------------------
>
> Key: DERBY-3355
> URL: https://issues.apache.org/jira/browse/DERBY-3355
> Project: Derby
> Issue Type: Bug
> Affects Versions: 10.3.1.4
> Environment: mac 0s x
> Reporter: geoff hendrey
> Assignee: Bryan Pendleton
> Attachments: patch.diff
>
>
> ' is not a column in the target table., SQL State: 42X04, Error Code: -1
> Hi:
> I think I have isolated a bug involving the use of double quotes to define a
> column name. Here s the SQL to reproduce the bug, followed by the error
> message generated by the final SQL statement. In order to make the bug go
> away, eliminate all use of double quotes in the SQL statements below. Note
> that the identical alter statement succeeds before the insert, and fail
> after. I have spent a long time trying to isolate this problem, so please
> take a look.
> CREATE TABLE Table2
> (
> "c" VARCHAR(32672)
> );
> alter table Table2 ALTER COLUMN "c" NULL;
> alter table Table2 ALTER COLUMN "c" NOT NULL;
> INSERT INTO Table2("c") VALUES('yo');
> alter table Table2 ALTER COLUMN "c" NULL;
> alter table Table2 ALTER COLUMN "c" NOT NULL;
> Query 1 of 6 elapsed time (seconds) - Total: 0.012, SQL query: 0.012,
> Building output: 0
> Query 2 of 6 elapsed time (seconds) - Total: 0.003, SQL query: 0.003,
> Building output: 0
> Query 3 of 6 elapsed time (seconds) - Total: 0.003, SQL query: 0.003,
> Building output: 0
> 1 Row(s) Inserted
> Query 4 of 6 elapsed time (seconds) - Total: 0.009, SQL query: 0.009,
> Building output: 0
> Query 5 of 6 elapsed time (seconds) - Total: 0.003, SQL query: 0.003,
> Building output: 0
> Error: java.sql.SQLException: Column 'C' is either not in any table in the
> FROM list or appears within a join specification and is outside the scope of
> the join specification or appears in a HAVING clause and is not in the GROUP
> BY list. If this is a CREATE or ALTER TABLE statement then 'C' is not a
> column in the target table., SQL State: 42X04, Error Code: -1
> -----Inline Message Follows-----
> Geoff hendrey wrote:
> > I think I have isolated a bug involving the use of double quotes to
> > define a column name.
> Hi Geoff, I agree, that is definitely a bug. Your script reproduces
> the problem for me, on the current Derby trunk.
> It appears that AlterTableConstantAction.validateNotNullConstraint
> is internally generating and executing a statement of the form:
> select count(*) from tab where not (col is not null)
> The code which generates this SQL staement is not properly enclosing
> the column name in double quotes, as you noticed, so the compiler
> converts the column name to upper case, and gets the no-such-column error.
> Can you open a problem report in Jira so that we can track this down
> and get it fixed?
> http://db.apache.org/derby/DerbyBugGuidelines.html
> thanks,
> bryan
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
