[sql] Re-order columns of table in Oracle

I have a table with 50+ columns and I need to swap the order of the first two columns. What is the best way to accomplish this using Oracle? Assume the table name is ORDERDETAILS and as it is, the first two columns are ITEM_ID and ORDER_ID. Once the rename is complete, the table name should still be ORDERDETAILS but the first two columns will be ORDER_ID and ITEM_ID. FWIW, column types and the rest of the columns and their order is irelevent.

Correct me if I'm wrong, but I think the general steps are:

  1. Rename the existing table.
  2. Drop the primary key constraint.
  3. Re-create the table with the correct column order.
  4. List item
  5. Run INSERT INTO .. SELECT to move the data from temp to the table in step #3.
  6. Drop the temp table.

I have little experience with Oracle so perhaps I'm missing a step or two.

Does a primary key imply an index in Oracle? Does dropping the primary key also drop the index?

SQL examples are much appreciated.

EDIT: Un-sincere thanks to those who question why it needs done instead of providing help. To answer your question as to why it needs done: I'm following orders from someone else who says I need to do it this way and the order of the columns DOES matter. My thoughts/opinions on this are irrelevent.

This question is related to sql oracle rename

The answer is


Look at the package DBMS_Redefinition. It will rebuild the table with the new ordering. It can be done with the table online.

As Phil Brown noted, think carefully before doing this. However there is overhead in scanning the row for columns and moving data on update. Column ordering rules I use (in no particular order):

  • Group related columns together.
  • Not NULL columns before null-able columns.
  • Frequently searched un-indexed columns first.
  • Rarely filled null-able columns last.
  • Static columns first.
  • Updateable varchar columns later.
  • Indexed columns after other searchable columns.

These rules conflict and have not all been tested for performance on the latest release. Most have been tested in practice, but I didn't document the results. Placement options target one of three conflicting goals: easy to understand column placement; fast data retrieval; and minimal data movement on updates.


It's sad that Oracle doesn't allow this, I get asked to do this by developers all the time..

Here's a slightly dangerous, somewhat quick and dirty method:

  1. Ensure you have enough space to copy the Table
  2. Note any Constraints, Grants, Indexes, Synonyms, Triggers, um.. maybe some other stuff - that belongs to a Table - that I haven't thought about?
  3. CREATE TABLE table_right_columns AS SELECT column1 column3, column2 FROM table_wrong_columns; -- Notice how we correct the position of the columns :)
  4. DROP TABLE table_wrong_columns;
  5. 'ALTER TABLE table_right_columns RENAME TO table_wrong_columns;`
  6. Now the yucky part: recreate all those items you noted in step 2 above
  7. Check what code is now invalid, and recompile to check for errors

And next time you create a table, please consider the future requirements! ;)


Use the View for your efforts in altering the position of the column: CREATE VIEW CORRECTED_POSITION AS SELECT co1_1, col_3, col_2 FROM UNORDERDED_POSITION should help.

This requests are made so some reports get produced where it is using SELECT * FROM [table_name]. Or, some business has a hierarchy approach of placing the information in order for better readability from the back end.

Thanks Dilip


I followed the solution above from Jonas and it worked well until I needed to add a second column. What I found is that when making the columns visible again Oracle does not necessarily set them visible in the order listed in the statement.

To demonstrate this follow Jonas' example above. As he showed, once the steps are complete the table is in the order that you'd expect. Things then break down when you add another column as shown below:

Example (continued from Jonas'):

Add another column which is to be inserted before column C.

ALTER TABLE t ADD (b2 INT);

Use the technique demonstrated above to move the newly added B2 column before column C.

ALTER TABLE t MODIFY (c INVISIBLE, d INVISIBLE, e INVISIBLE);
ALTER TABLE t MODIFY (c VISIBLE, d VISIBLE, e VISIBLE);

DESCRIBE t;

Name
----
A
B
B2
D
E
C

As shown above column C has moved to the end. It seems that the ALTER TABLE statement above processed the columns in the order D, E, C rather than in the order specified in the statement (perhaps in physical table order). To ensure that the column is placed where desired it is necessary to make the columns visible one by one in the desired order.

ALTER TABLE t MODIFY (c INVISIBLE, d INVISIBLE, e INVISIBLE);
ALTER TABLE t MODIFY c VISIBLE;
ALTER TABLE t MODIFY d VISIBLE;
ALTER TABLE t MODIFY e VISIBLE;

DESCRIBE t;

Name
----
A
B
B2
C
D
E

Since the release of Oracle 12c it is now easier to rearrange columns logically.

Oracle 12c added support for making columns invisible and that feature can be used to rearrange columns logically.

Quote from the documentation on invisible columns:

When you make an invisible column visible, the column is included in the table's column order as the last column.

Example

Create a table:

CREATE TABLE t (
    a INT,
    b INT,
    d INT,
    e INT
);

Add a column:

ALTER TABLE t ADD (c INT);

Move the column to the middle:

ALTER TABLE t MODIFY (d INVISIBLE, e INVISIBLE);
ALTER TABLE t MODIFY (d VISIBLE, e VISIBLE);

DESCRIBE t;

Name
----
A
B
C
D
E

Credits

I learned about this from an article by Tom Kyte on new features in Oracle 12c.


Examples related to sql

Passing multiple values for same variable in stored procedure SQL permissions for roles Generic XSLT Search and Replace template Access And/Or exclusions Pyspark: Filter dataframe based on multiple conditions Subtracting 1 day from a timestamp date PYODBC--Data source name not found and no default driver specified select rows in sql with latest date for each ID repeated multiple times ALTER TABLE DROP COLUMN failed because one or more objects access this column Create Local SQL Server database

Examples related to oracle

concat yesterdays date with a specific time ORA-28001: The password has expired how to modify the size of a column How to create a blank/empty column with SELECT query in oracle? Find the number of employees in each department - SQL Oracle Query to display all tablespaces in a database and datafiles When or Why to use a "SET DEFINE OFF" in Oracle Database How to insert date values into table error: ORA-65096: invalid common user or role name in oracle In Oracle SQL: How do you insert the current date + time into a table?

Examples related to rename

Gradle - Move a folder from ABC to XYZ How to rename a component in Angular CLI? How do I completely rename an Xcode project (i.e. inclusive of folders)? How to rename a directory/folder on GitHub website? How do I rename both a Git local and remote branch name? Convert row to column header for Pandas DataFrame, Renaming files using node.js Replacement for "rename" in dplyr Rename multiple columns by names Rename specific column(s) in pandas