Archive for the ‘sql’ Category
Multiple Column Lookups?
I’ve been working with Oracle so long, sometimes it’s frustrating when I find a syntax feature isn’t in another database. I ran into another example tonight. There isn’t a multiple column look up operator in MySQL. For example, you can do this in Oracle:
1 2 3 4 5 | DELETE FROM common_lookup WHERE (common_lookup_table,common_lookup_column) IN (('TRANSACTION','PAYMENT_METHOD_TYPE') ,('TRANSACTION','TRANSACTION_TYPE') ,('RENTAL_ITEM','RENTAL_ITEM_TYPE')); |
When I transformed it to comply with MySQL, it seems just as clean. In fact, with strings it’s simpler.
1 2 3 | DELETE FROM common_lookup WHERE common_lookup_table IN ('TRANSACTION','RENTAL_ITEM') AND common_lookup_column IN ('TRANSACTION_TYPE','PAYMENT_METHOD_TYPE','RENTAL_ITEM_TYPE'); |
Then, I thought about it. Oracle would let me write a single subquery returning the two columns, whereas MySQL requires two subqueries in their syntax. Likewise, MySQL doesn’t support the WITH
clause, which would let me reference a single query result in the scope of the master query (Oracle and SQL Server do support that).
I guess we can hope that Oracle will implement the feature in MySQL now that they own it. 🙂 Let me know if I’ve missed some fabulous syntax alternative.
Alice and Assignments
As I continue down the warren hole of Persistent Stored Modules (SQL/PSM) in MySQL, I keep wondering about that mad hare, Johnny Depp. Alice isn’t a programming language to teach me anything in this dream. Moreover, TIm Burton’s tale this seems oddly familiar, like a child’s story gone mad.
A quick update on comparative SQL expression assignments between PL/SQL and MySQL. When you want to filter a value through SQL functions before assigning it to another variable in MySQL, it’s not like PL/SQL. Just like the new Alice in Wonderland movie isn’t like the book.
The programmatic differences lies in their origins. PL/SQL evolved from Pascal through Ada to become a recursive language where you can call SQL from PL/SQL and PL/SQL from SQL. MySQL implemented PSMs from the ANSI SQL:2003 specification, which didn’t see it the same way, apparently (a disclaimer since I’ve not read the details of the specification).
Personally, I think PL/SQL is easier to write but I’ve been using it for almost 20 years. Naturally, there may be a consistency thread on this that I’m missing and an opportunity that I may exploit. After all, it is dark in this warren hole.
Oracle PL/SQL Assignments from SQL Expressions
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 | -- Enable output printing. SET SERVEROUTPUT ON SIZE 1000000 -- Define an anonymous block. DECLARE -- Declare a source variable. lv_right_operand VARCHAR2(10) := 'March'; -- Define a target variable for the assignment. lv_left_operand VARCHAR2(10); BEGIN -- Return the expression from a nested call parameter of the source variable. lv_left_operand := UPPER(SUBSTR(lv_right_operand,1,3)); -- Print it to console. dbms_output.put_line('Output ['||lv_left_operand||']'); END; / |
Oracle also supports this syntax, which isn’t frequently used because it’s much more verbose syntactically. It is also equivalent to the PSM syntax adopted by MySQL.
-- Define an anonymous block. DECLARE -- Declare a source variable. lv_right_operand VARCHAR2(10) := 'March'; -- Define a target variable for the assignment. lv_left_operand VARCHAR2(10); BEGIN -- Return the expression from a nested call parameter of the source variable. SELECT UPPER(SUBSTR(lv_right_operand,1,3)) INTO lv_left_operand FROM dual; -- Print it to console. dbms_output.put_line('Output ['||lv_left_operand||']'); END; / |
That means we can do it like the White Queen wants it or the Red Queen wants it in Oracle. Flexibility in PL/SQL is clearly broader because of the assignment options. Not so in MySQL, as you’ll see.
MySQL PSM Assignment from SQL Expressions
First, MySQL’s PSM approach doesn’t support anonymous blocks. The example must create a stored function or procedure, and then call it. A procedure seems like the best fit for the example.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | -- Conditionally drop procedure. SELECT 'DROP PROCEDURE IF EXISTS assignit' AS "Statement"; DROP PROCEDURE IF EXISTS assignit; -- Create the proceudre SELECT 'CREATE PROCEDURE assignit' AS "Statement"; DELIMITER $$ -- Define the procedure. CREATE PROCEDURE assignit() BEGIN /* Declare a source variable. */ DECLARE lv_right_operand VARCHAR(10) DEFAULT 'March'; /* Define a target variable for the assignment. */ DECLARE lv_left_operand VARCHAR(3); /* Assign the modified value through the SELECT-INTO model. */ SELECT UCASE(SUBSTRING(lv_right_operand,1,3)) INTO lv_left_operand; /* Display assigned value. */ SELECT lv_left_operand; END; $$ DELIMITER ; -- Call the procedure. CALL assignit(); |
The only question here in the warren is: Who’s the White Queen; and who’s the Red Queen. Which semantic should I choose? My I hope is that I wake up before it’s … oops, off with his head. Actually, 3D or not, I’ll probably not see it, that’s the new Alice in Wonderland film.
Likewise, when my students wake up and read this they’ll know I was just answering a question on how to perform assignments in MySQL stored procedures. By the way, I’ve updated this assignment process in my Debugging MySQL Procedures post.
As an aside, I’ve got a new MySQL debugger that I’m testing later in the week. When I complete the test cases, I’ll post a review.
MySQL Standard Group By
Teaching SQL is interesting because folks try syntax that experience would tell you shouldn’t work. It was interesting when I discovered what should be broken from my perspective but was expected behavior in MySQL. It became clearer to me why it’s there as I did some experimenting with it enabled and disabled. While I’d still argue it’s broken, it’s the only way to get support for advanced aggregation concepts.
The reason that I find that it broken is a matter of perspective not standards. The fact that you can select a set of non-aggregated columns with an aggregated column, and exclude one, more than one, or all of the non-aggregated columns from the GROUP BY
clause seemed like a broken behavior. The MySQL behavior is explained in Chapter 11.12.3 of the MySQL 5.1 Reference. What wasn’t clear from the documentation when I read it was why the behavior is allowed. Gary’s comment on the original post made me look more deeply into the behavior.
If you check the referenced documentation, the process is called GROUP BY
hidden columns. That label didn’t immediately resonate with me. That’s probably of my background with IBM’s DB2 and Oracle. I’m so accustom to how they work, that sometimes I think they set the standards but they don’t. It turns out that the GROUP BY
clause in the ANSI SQL standards doesn’t allow for expressions.
The default configuration of MySQL allows for the expressions (functions) in the GROUP BY
clause. The documentation refers to hidden columns, which are in point of fact expressions in the GROUP BY
clause. MySQL does support expressions in the GROUP BY
by default, and requires, like Oracle, that when you use an expression in the SELECT
clause that you mirror it in the GROUP BY
clause. This means you group on the result of the expression, not a column in the table or view.
In the default configuration, you have three options. They are determinate results, indeterminate results, and hidden column results. The following cover these components.
Determinate Results
Determinate results are straight forward. They require that all non-aggregated columns in the SELECT
clause are mirrored in the GROUP BY
clause. This means that the non-aggregated column values are the key upon which results are aggregated.
An example of determinate results is:
1 2 3 4 5 6 | SELECT key_one , key_two , SUM(counter) FROM GROUPING GROUP BY key_one , key_two; |
Indeterminate Results
Indeterminate results isn’t straight forward but isn’t hard to grasp. An indeterminate result set is returned when one or more non-aggregated columns in a SELECT
clause aren’t listed in the GROUP BY
clause. The columns listed in the SELECT
clause but excluded from the GROUP BY
clause return meaningless values because they’re column values chosen indeterminately from all pre-aggregated rows.
The following query runs in a generically configured MySQL instance without an error. It returns a meaningless key_two
column value from the pre-aggregated row set. In the example, the aggregation column counts the unique key_one
column values. This behavior makes key_one
a determinate value, and key_two
an indeterminate value.
1 2 3 4 5 | SELECT key_one , key_two , SUM(counter) FROM GROUPING GROUP BY key_one; |
You can fix this mixed return set by adding the key_two
column to the GROUP BY
clause, which would return a determinate set. Alternatively, you can prevent the default behavior for the GROUP BY
clause by adding the ONLY_FULL_GROUP_BY
mode variable to your SQL_MODE
system variable.
After setting the SQL_MODE>
variable, a GROUP BY
must contain all non-aggregated columns. When you make this change to the SQL_MODE
system variable, you also disable any queries that use an expression in their GROUP BY
clause. Those queries with expressions in the group by will now raise an ERROR 1055
exception, like this:
ERROR 1055 (42000): 'sampledb.grouping.key_two' isn't in GROUP BY |
If you want to prevent indeterminate results and don’t use expressions in the GROUP BY
clause, you can add the ONLY_FULL_GROUP_BY
mode to your SQL_MODE
system variable. You can do that during a session with the following syntax:
SET SQL_MODE=(SELECT CONCAT(@@sql_mode,',ONLY_FULL_GROUP_BY')); |
Or, you can start the mysqld
with the following option:
mysqld --sql_mode="sql_mode1,sql_mode2, ... ,sql_mode(n+1)" |
A better alternative, is to add it to the my.cnf
configuration file on Linux, or my.ini
configuration file on Windows. You can add it to this line, which is done at the end of the line.
# Set the SQL mode to strict sql-mode="STRICT_TRANS_TABLES,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION,ONLY_FULL_GROUP_BY" |
If you opt for changing it at the server configuration file, you must stop and restart the mysqld
process. You can do that on Windows from the command line, like this on Windows provide the service name is mysql
. If you’ve set the Windows service to mysql51
, then you need to substitute mysql51
for mysql
.
Hidden Column Results
A hidden column result, is a result generated by an expression in the GROUP BY
clause. The following is a query that lets you sum transactions by the month name. This is supported by the default behavior of hidden columns. The following is an approach that you might find in Oracle because they don’t support aggregation by only part of the non-aggregated columns in a query.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 | SELECT il.month AS "MON-YEAR" , il.base AS "BASE" , il.plus10 AS "10_PLUS" , il.plus20 AS "20_PLUS" FROM (SELECT CONCAT(UPPER(SUBSTRING(MONTHNAME(t.transaction_date),1,3)),'-',EXTRACT(YEAR FROM t.transaction_date)) AS MONTH , MONTH(t.transaction_date) AS sortkey , LPAD(CONCAT('$',FORMAT(SUM(t.transaction_amount) * 1.0,2)),10,' ') AS base , LPAD(CONCAT('$',FORMAT(SUM(t.transaction_amount) * 1.1,2)),10,' ') AS plus10 , LPAD(CONCAT('$',FORMAT(SUM(t.transaction_amount) * 1.2,2)),10,' ') AS plus20 FROM TRANSACTION t WHERE EXTRACT(YEAR FROM t.transaction_date) = 2009 GROUP BY CONCAT(UPPER(SUBSTRING(MONTHNAME(t.transaction_date),1,3)),'-',EXTRACT(YEAR FROM t.transaction_date)) , MONTH(t.transaction_date)) il ORDER BY il.sortkey; |
The interesting thing about MySQL’s approach is that the Oracle equivalent of an inline view is unnecessary. You can simply switch the functions like the exmaple below.
1 2 3 4 5 6 7 8 9 10 | SELECT DATE_FORMAT(t.transaction_date,'%m-%Y') AS MONTH , MONTH(t.transaction_date) AS sortkey , LPAD(CONCAT('$',FORMAT(SUM(t.transaction_amount) * 1.0,2)),10,' ') AS base , LPAD(CONCAT('$',FORMAT(SUM(t.transaction_amount) * 1.1,2)),10,' ') AS plus10 , LPAD(CONCAT('$',FORMAT(SUM(t.transaction_amount) * 1.2,2)),10,' ') AS plus20 FROM TRANSACTION t WHERE EXTRACT(YEAR FROM t.transaction_date) = 2009 GROUP BY DATE_FORMAT(t.transaction_date,'%m-%Y') , MONTH(t.transaction_date) ORDER BY sortkey; |
When you add the ONLY_FULL_GROUP_BY
mode to your SQL_MODE
system variable, you disallow this type of behavior and the query would no longer work. It would return the following error:
ERROR 1055 (42000): 'sampledb.t.transaction_date' isn't in GROUP BY |
Complete Code Sample ↓
Expand this section to see the sample working code for indeterminate results.
This builds the GROUPING
table, inserts nine rows, and tests it before and after setting the variable in the session. The change reverts after you exit and re-enter the database.
-- Conditionally drop sample table. SELECT 'DROP TABLE IF EXISTS grouping' AS "Statement"; DROP TABLE IF EXISTS GROUPING; -- Create sample table. SELECT 'CREATE TABLE grouping' AS "Statement"; CREATE TABLE GROUPING ( key_one CHAR(5) , key_two CHAR(5) , counter INT UNSIGNED ); -- Insert nine rows into the sample table. SELECT 'INSERT INTO grouping' AS "Statement" INSERT INTO GROUPING VALUES ('One','Uno-1',1),('Two','Due-1',2),('Three','Tre-1',3) ('One','Uno-2',1),('Two','Due-2',2),('Three','Tre-2',3) ('One','Uno-3',1),('Two','Due-3',2),('Three','Tre-3',3); -- Query with hidden columns, no enforcement on non-aggregate columns. SELECT 'SELECT non-aggregates FROM grouping with hidden columns' AS "Statement"; SELECT key_one, key_two, SUM(counter) FROM GROUPING GROUP BY key_one; -- Check current system mode variable assignment. SELECT 'SELECT @@sql_mode' AS "Statement"; SELECT @@sql_mode; -- Append the mode to the existing system mode variable. SELECT 'SET SQL_MODE=(SELECT ...)' AS "Statement"; SET SQL_MODE=(SELECT CONCAT(@@sql_mode,',ONLY_FULL_GROUP_BY')); -- Check modified system mode variable assignment. SELECT 'SELECT @@sql_mode' AS "Statement"; SELECT @@sql_mode; |
Wrap a cursor function
A Gauss posted a question on my from last year’s Utah Oracle User’s Group Training Days presentation. If I understood his question correctly, this should help him work with his legacy code. Honestly, as I wrote the example something Bryn Llewellyn said kept banging around in my head, “Just because we can, doesn’t mean we should.” He was speaking of writing poorly engineered code.
Sometimes, we don’t get the opportunity to re-factor existing code. That leaves us with writing wrappers that aren’t pretty or effective. A realization and preface to showing everyone how to accomplish these tasks, and perhaps a watch out warning if you choose this path. I suspect that there may be a better way but I don’t know their code tree.
Here’s the question, as I understand it. They’ve got a library function in PL/SQL that returns a system reference cursor and is principally consumed by an external Java program. This type of architecture is more or less an Adapter OOAD pattern that I wrote about here, over a year and a half ago. The question comes to how to you wrap this approach and make it work in PL/SQL natively too.
The answer depends on some earlier posts because I don’t have a great deal of time to write new examples. It uses a COMMON_LOOKUP
table, which is more or less a bunch of small tables grouped into a big table for use in user interaction forms. That way the values don’t get lost in a large code base and are always consistently maintained. These types of tables exist in all major ERP and CRM applications.
The base code for the example is found here, where I discussed how you can effectively use object tables – collections of user-defined object types (Oracle 9iR2 forward if I remember correctly). You can grab the full code at the bottom of the page by clicking the Code Script widget to unfold the code. That code also depends on the Oracle Database 11g PL/SQL Programming downloadable code, which you can download by clicking the link to the zip file location.
Here are the steps to wrap a function that returns a PL/SQL reference cursor so that it can also return a PL/SQL associative array.
- Create a package specification to hold all the components that are required to manage the process. Assuming that they may have anchored the system reference cursor to something other than a table like a shared cursor, which is a cumbersome implementation design. (I actually chose to exclude this from the book because it’s a stretch as a good coding practice. At least, it is from my perspective. Also, I couldn’t find an example in the Oracle documentation, which led me to believe they didn’t think it’s a great idea either or I could have glossed over it.) You should note that the PL/SQL
RECORD
, Associative Array (collection), and theREF CURSOR
are defined in this package specification.
-- Create a package to hold the PL/SQL record structure. CREATE OR REPLACE PACKAGE example IS -- Force cursors to be read as if empty every time. PRAGMA SERIALLY_REUSABLE; -- Package-level record structure that mimics SQL object type. TYPE common_lookup_record IS RECORD ( common_lookup_id NUMBER , common_lookup_type VARCHAR2(30) , common_lookup_meaning VARCHAR2(255)); -- Package-level collection that mimics SQL object table. TYPE common_lookup_record_table IS TABLE OF common_lookup_record INDEX BY PLS_INTEGER; -- Cursor structure to support a strongly-typed reference cursor. CURSOR c IS SELECT common_lookup_id , common_lookup_type , common_lookup_meaning FROM common_lookup; -- Package-level strongly-typed system reference cursor. TYPE cursor_lookup IS REF CURSOR RETURN c%ROWTYPE; END; / |
- Write a function to return a strongly typed system reference cursor that’s anchored to a cursor defined in the package. This is fairly straightforward when the package specification is done right. You should notice right away that anchoring the original cursor in the package was a horrible practice because you must repeat it all again in the function. In my opinion, you shouldn’t anchor any system reference cursor explicitly to anything other than a table. The cursor could have used the generic weak cursor data type –
SYS_REFCURSOR
. Doing so, saves all the extra lines required by a potential shared cursor.
CREATE OR REPLACE FUNCTION get_common_lookup_cursor ( TABLE_NAME VARCHAR2, column_name VARCHAR2) RETURN example.cursor_lookup IS -- Define a local variable of a strongly-typed reference cursor. lv_cursor EXAMPLE.CURSOR_LOOKUP; BEGIN -- Open the cursor from a static cursor OPEN lv_cursor FOR SELECT common_lookup_id , common_lookup_type , common_lookup_meaning FROM common_lookup WHERE common_lookup_table = TABLE_NAME AND common_lookup_column = column_name; -- Return the cursor handle. RETURN lv_cursor; END; / |
- Write a wrapper function that takes the reference cursor as a formal parameter and returns an Associative Array. You should note that this can’t be called from a SQL context. You must only use it in a PL/SQL context because system reference cursors are PL/SQL only data types.
CREATE OR REPLACE FUNCTION convert_common_lookup_cursor ( pv_cursor EXAMPLE.CURSOR_LOOKUP) RETURN example.common_lookup_record_table IS -- Declare a local counter variable. counter INTEGER := 1; -- Local PL/SQL-only variable. out_record EXAMPLE.COMMON_LOOKUP_RECORD; out_table EXAMPLE.COMMON_LOOKUP_RECORD_TABLE; BEGIN -- Grab the cursor wrapper and return values to a PL/SQL-only record collection. LOOP FETCH pv_cursor INTO out_record; EXIT WHEN pv_cursor%NOTFOUND; -- Assign it one row at a time to an associative array. out_table(counter) := out_record; -- Increment the counter. counter := counter + 1; END LOOP; -- Return the record collection. RETURN out_table; END; / |
- You can test the program in an anonymous block, like the one below. It defines a local Associative Array variable and then assigns the system reference cursor through the wrapper.
-- Open the session to see output from PL/SQL blocks. SET SERVEROUTPUT ON DECLARE -- Define a local associative array. process_table EXAMPLE.COMMON_LOOKUP_RECORD_TABLE; BEGIN -- Print title block. dbms_output.put_line('Converting a SYS_REFCURSOR to TABLE'); dbms_output.put_line('---------------------------------------------------'); -- Run the dynamic variables through the cursor generating function and then convert it. process_table := convert_common_lookup_cursor(get_common_lookup_cursor('ITEM','ITEM_TYPE')); -- Read the content of the Associative array. FOR i IN 1..process_table.COUNT LOOP dbms_output.put('['||process_table(i).common_lookup_id||']'); dbms_output.put('['||process_table(i).common_lookup_type||']'); dbms_output.put_line('['||process_table(i).common_lookup_meaning||']'); END LOOP; END; / |
I hope this answers Gauss’s question. While writing it, I could envision another question that might pop-up. How do you convert an object table type to a PL/SQL context. It was an omission not to include it in that original post on object table types. Here’s how you wrap an object table type into a PL/SQL scope collection.
You might have guessed. It’s done with another wrapper function. At least this is the easiest way to convert the SQL data type to a PL/SQL data type that I see. If you’ve another approach, a better way, let us know.
CREATE OR REPLACE FUNCTION get_common_lookup_record_table ( TABLE_NAME VARCHAR2 , column_name VARCHAR2 ) RETURN example.common_lookup_record_table IS -- Declare a local counter variable. counter INTEGER := 1; -- Define a dynamic cursor that takes two formal parameters. CURSOR c (table_name_in VARCHAR2, table_column_name_in VARCHAR2) IS SELECT * FROM TABLE(get_common_lookup_object_table(UPPER(table_name_in),UPPER(table_column_name_in))); -- A local PL/SQL-only collection variable. list EXAMPLE.COMMON_LOOKUP_RECORD_TABLE; BEGIN -- Grab the cursor wrapper and return values to a PL/SQL-only record collection. FOR i IN c(TABLE_NAME, column_name) LOOP list(counter) := i; counter := counter + 1; END LOOP; -- Return the record collection. RETURN list; END get_common_lookup_record_table; / |
You can then test this in an anonymous block, like so:
-- Open the session to see output from PL/SQL blocks. SET SERVEROUTPUT ON DECLARE -- Declare a local PL/SQL-only collection and assign the value from the function call. list EXAMPLE.COMMON_LOOKUP_RECORD_TABLE; BEGIN -- Print title block. dbms_output.put_line('Converting a SQL Collection to a PL/SQL Collection'); dbms_output.put_line('---------------------------------------------------'); -- Assign wrapped SQL collection to a PL/SQL-only collection. list := get_common_lookup_record_table('ITEM','ITEM_TYPE'); -- Call the record wrapper function. FOR i IN 1..list.COUNT LOOP dbms_output.put('['||list(i).common_lookup_id||']'); dbms_output.put('['||list(i).common_lookup_type||']'); dbms_output.put_line('['||list(i).common_lookup_meaning||']'); END LOOP; END; / |
As always, I hope this helps somebody without paying a fee for content. 😉
Stable set of rows?
My students are into the upload lab and they’ve encountered the wonderful ORA-30926
error from a MERGE
statement. An example of the MERGE
for an upload is in this earlier blog post.
This is the wonderful error message, which doesn’t seem to have meaning for many. The key is the non-deterministic where clauses phrase. That phrase means that the the query inside the USING
clause returns a non-unique set of rows. The returned rows can’t be matched uniquely against the target table in the ON
clause. The ON
clause is where the MERGE
statment matches the source query’s rows against the target table’s rows. The best join condition in a MERGE
statement is one between a surrogate primary and foreign key column.
Error: ORA-30926 Text: Unable TO GET a stable SET OF ROWS IN the SOURCE TABLES. -------------------------------------------------------------------- Cause: A stable SET OF ROWS could NOT be got because OF LARGE dml activity OR a non-deterministic WHERE clause. Action: Remove any non-deterministic WHERE clauses AND reissue the dml. |
The problem is very much like when you write what you think is a single-row subquery but find out it’s actually a multiple-row subquery when it return an ORA-01422
error in Oracle.
As a rule, I’ve found that using the DISTINCT
operator in the source SELECT
statement fixes it most of the time because the join isn’t returning a unique set of rows. Although, the better solution requires that you identify how to gain a unique result set. Alternatively, you need to re-examine the logic of your WHERE
clause. It also happens when the SELECT
clause returns date-time data types like SYSDATE
instead of date data types. A date-time can be converted by using the TRUNC
ate function like this:
MERGE INTO TABLE_NAME target USING (SELECT ... , TRUNC(SYSDATE) AS creation_date FROM ...) SOURCE ON (target.primary_key_column = SOURCE.primary_key_column) WHEN MATCHED THEN UPDATE SET last_updated_by = SOURCE.last_updated_by , last_update_date = SOURCE.last_update_date WHEN NOT MATCHED THEN INSERT VALUES ( column_list_of_values ); |
Hope this helps my students and others … 😉
Debugging MySQL Procedures
In my second database class we focus on PL/SQL but we’ve begun highlighting the alternatives in MySQL and SQL Server. A student asked how they could debug runtime variable values in a MySQL Stored Procedure (or subroutines according to some documentation). You can see this post for how to create an equivalent solutions for MySQL functions.
In Oracle, we debug with the DBMS_OUTPUT
package. Packages, like DBMS_OUTPUT
hold related functions and procedures, and are a corollary to System.out.println()
in Java.
Before you can see the output at the command-line in Oracle (that is if you’re not using SQL*Developer or Toad), you must set a SQL*Plus environment variable. These variables don’t exist in MySQL or SQL Server command-line tools because they never served the function of a report writer like SQL*Plus.
You enable output display in Oracle by setting this in SQL*Plus:
SQL> SET SERVEROUTPUT ON SIZE 1000000 |
You can test your anonymous or named block. Since MySQL doesn’t support anonymous named block, the examples using a trivial procedure that prints Hello World! (orginal, right ;-)).
1 2 3 4 5 6 7 8 9 10 11 12 | -- Create a procedure in Oracle. CREATE OR REPLACE PROCEDURE hello_world IS BEGIN -- Print a word without a line return. DBMS_OUTPUT.put('Hello '); -- Print the rest of the phrase and a line return. DBMS_OUTPUT.put_line('World!'); END; / -- Call the procedure. EXECUTE hello_world; |
It’s seems useless to print the output because it should be evident. MySQL procedures are a bit different because there’s no OR REPLACE
syntax. The equivalent to calling the DBMS_OUTPUT
package procedures in MySQL is to simply select a string. Now you can do this with or without the FROM dual
clause in MySQL, don’t we wish we could do the same thing in Oracle. 🙂
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 | -- Conditionally drop the procedure. SELECT 'DROP PROCEDURE hello_world' AS "Statement"; DROP PROCEDURE IF EXISTS hello_world; -- Reset the delimiter to write a procedure. DELIMITER $$ -- Create a procedure in Oracle. CREATE PROCEDURE hello_world() BEGIN -- Print the phrase and a line return. SELECT 'Hello World!'; END; $$ -- Reset the delimiter back to a semicolon to work again. DELIMITER ; -- Call the procedure. SELECT 'CALL hello_world' AS "Statement"; CALL hello_world(); |
Originally, I tried to keep this short but somebody wanted an example in a loop. Ouch, loops are so verbose in MySQL. Since I was modifying this post, it seemed like a good idea to put down some guidelines for successful development too.
Guidelines for Development of Procedures
Declaration Guidelines
The sequencing of components in MySQL procedures is important. Unlike, PL/SQL, there’s no declaration block, declarations must be at the top of the execution block. They also must appear in the following order:
- Variable declarations must go first, you can assign initial values with the
DEFAULT
keyword. While not required, you should:
- Consider using something like
lv_
to identify them as local variables for clarity and support of your code. - Consider grouping local variables that relate to handlers at the bottom of the list of variables.
- After local variables and before handlers, you put your cursor definitions. You should note that MySQL doesn’t support explicit dynamic cursors, which means you can’t define one with a formal signature. However, you do have prepared statements and they mimic dynamic cursor behaviors.
- Last in your declaration block, you declare your handler events.
Execution Guidelines
- Variable assignments are made one of two ways:
- You should start each execution block with a
START TRANSACTION
and then aSAVEPOINT
, which ensures the procedure acts like a cohesive programming unit. - You assign a
left_operand = right_operand;
as a statement. - You use the
SELECT column INTO variable;
syntax to filter a value through SQL functions and assign the resulting expression to a local variable. - You assign a single row cursor output to variables using a
SELECT column INTO variable FROM ...
.
- You must assign values from cursors called in a loop into local variables when you want to use the results in nested SQL statements or loops.
- You must reset looping variables, like the
fetched
control variable at the end of the loop to reuse the handler variable in subsequent loops. - You must assign values to local variables if you want to use them in the exception handler.
- If you’ve started a transaction, don’t forget to
COMMIT
your work.
Exception Guidelines
- Leave out the exception handler until you’ve tested all outcomes, and make sure you document them and add them as potential handlers.
- When you deploy exception blocks, they’re the last element at the bottom of the exception block.
- You should consider explicit exception handlers for each error unless the action taken is the same.
- You should consider grouping all exception handlers when the action taken is the same.
- You should include a
ROLLBACK
whenever you’ve performed two or more SQL statements that may modify data.
Below is an example for putting debug code inside a loop.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 | -- Conditionally drop a sample table. SELECT 'DROP TABLE IF EXISTS sample' AS "Statement"; DROP TABLE IF EXISTS sample; -- Create a table. CREATE TABLE sample ( sample_id INT UNSIGNED PRIMARY KEY AUTO_INCREMENT , sample_msg VARCHAR(20)); -- Insert into sample. INSERT INTO sample (sample_msg) VALUES ('Message #1') ,('Message #2') ,('Message #3'); -- Conditionally drop the procedure. SELECT 'DROP PROCEDURE debug_loop' AS "Statement"; DROP PROCEDURE IF EXISTS debug_loop; -- Reset the delimiter to write a procedure. DELIMITER $$ -- Create a procedure in Oracle. CREATE PROCEDURE debug_loop() BEGIN /* Declare a counter variable. */ DECLARE lv_counter INT DEFAULT 1; /* Declare local control loop variables. */ DECLARE lv_sample_id INT; DECLARE lv_sample_msg VARCHAR(20); /* Declare a local variable for a subsequent handler. */ DECLARE duplicate_key INT DEFAULT 0; DECLARE fetched INT DEFAULT 0; /* Declare a SQL cursor fabricated from local variables. */ DECLARE sample_cursor CURSOR FOR SELECT * FROM sample; /* Declare a duplicate key handler */ DECLARE CONTINUE HANDLER FOR 1062 SET duplicate_key = 1; /* Declare a not found record handler to close a cursor loop. */ DECLARE CONTINUE HANDLER FOR NOT FOUND SET fetched = 1; /* Start transaction context. */ START TRANSACTION; /* Set savepoint. */ SAVEPOINT all_or_none; /* Open a sample cursor. */ OPEN sample_cursor; cursor_sample: LOOP /* Fetch a row at a time. */ FETCH sample_cursor INTO lv_sample_id , lv_sample_msg; /* Place the catch handler for no more rows found immediately after the fetch operation. */ IF fetched = 1 THEN LEAVE cursor_sample; END IF; -- Print the cursor values. SELECT CONCAT('Row #',lv_counter,' [',lv_sample_id,'][',lv_sample_msg,']') AS "Rows"; -- Increment counter variable. SET lv_counter = lv_counter + 1; END LOOP cursor_sample; CLOSE sample_cursor; /* This acts as an exception handling block. */ IF duplicate_key = 1 THEN /* This undoes all DML statements to this point in the procedure. */ ROLLBACK TO SAVEPOINT all_or_none; END IF; END; $$ -- Reset the delimiter back to a semicolon to work again. DELIMITER ; -- Call the procedure. SELECT 'CALL debug_loop' AS "Statement"; CALL debug_loop(); |
This post certainly answers the student question. Hopefully, it also helps other who must migrate Oracle skills to MySQL. Since IBM DB2 has introduced a PL/SQL equivalent, wouldn’t it be nice if Oracle did that for MySQL. That is, migrate PL/SQL to MySQL. Don’t tell me if you think that’s a pipe dream, I’d like to hope for that change.
MySQL Boolean Data Type
Somebody posted a quick question about the outcome of defining a table with a bool
data type in PHPMyAdmin. They were subsequently surprised when they checked the MySQL database and found that it was actually a tinyint(1)
. The natural question they had was: “What do you enter – true/false or 1/0?”
I promised to post an answer tonight, and morning counts too. You can enter a true
or false
because they’re synonyms for a 1
or 0
respectively. TINYINT
is the supported data type, and BIT
, BOOL
, and BOOLEAN
are synonyms for the base data type.
Here’s an example in MySQL:
mysql> CREATE TABLE data_type (TYPE bool); mysql> DESCRIBE data_type; +-------+------------+------+-----+---------+-------+ | FIELD | TYPE | NULL | KEY | DEFAULT | Extra | +-------+------------+------+-----+---------+-------+ | TYPE | tinyint(1) | YES | | NULL | | +-------+------------+------+-----+---------+-------+ mysql> INSERT INTO data_type VALUES (TRUE),(FALSE); mysql> SELECT * FROM data_type; +------+ | TYPE | +------+ | 1 | | 0 | +------+ |
The comment below raises the question of what happens with values in the valid range of TINYINT
that aren’t 0
or 1
, like 5. The simple answer is they’re not valid when compared against the true
and false
constants, as you can see by creating the following example.
-- Create a test table. CREATE TABLE verify ( verify_id INT UNSIGNED PRIMARY KEY AUTO_INCREMENT , range_1 tinyint UNSIGNED , range_2 tinyint ); -- Insert test values. INSERT INTO verify VALUES (NULL, 0, 0) ,(NULL, 1, 1) ,(NULL, 1,-1) ,(NULL, 2, 2); -- Query results. SELECT range_1 AS "Value" , CASE WHEN range_1 = TRUE THEN 'True' WHEN range_1 = FALSE THEN 'False' ELSE 'Invalid' END AS "Unsigned" , range_2 AS "Value" , CASE WHEN range_2 = TRUE THEN 'True' WHEN range_2 = FALSE THEN 'False' ELSE 'Invalid' END AS "Signed" FROM verify; |
The results of the test demonstrate that only a 0
or 1
value validates against the false
or true
constants, as shown:
+-------+----------+-------+---------+ | Value | Unsigned | Value | Signed | +-------+----------+-------+---------+ | 0 | False | 0 | False | | 1 | True | 1 | True | | 1 | True | -1 | Invalid | | 2 | Invalid | 2 | Invalid | +-------+----------+-------+---------+ |
Based on the comment, the 256 value range of TINYINT
is found here.
SQL Aggregation Tutorial
I’ve been working on a Basic Aggregation tutorial for my students. I think this might be close to what may benefit them. However, I thought it would be great to put it out there and solicit ideas. If you have some on improving this post, please let me know.
My first take at the post …
This is a lesson on basic aggregation in SQL. Aggregation in SQL means counting, adding, and grouping by results of counts or sums. Aggregation is a critical part of using the SQL language. At a basic level, aggregation includes the COUNT
, SUM
, AVERAGE
, MAX
, and MIN
aggregation functions; and the ORDER BY
, GROUP BY
, and HAVING
clauses.
You’ll find the setup scripts for these examples at the bottom of this blog page. The best way to use this page is to copy the setup code, run it in your database, and then test the examples as you work though them.
Data set ↓
This section discusses the data set. You unfold it by clicking on the bold text above.
The illustration of aggregation is a challenge because a trivial set of numbers doesn’t present much challenge or sometimes relevance but it shows the basic concept. A more applied sample set makes mental calculation and comprehension of functions more difficult. As a result of that trade off, this uses two copies of the ordinal numbers (ordinal numbers are the single digit integers from zero to nine).
After you run the setup script, you can display the data set with the following query.
-- Oracle only SQL*Plus formatting command to smooth output display. SET PAGESIZE 99 -- Query ordered by the sets and then the integers in ascending order. SELECT * FROM ordinal ORDER BY 2, 4; |
You could have sorted them more easily by using the ORDINAL_ID
surrogate key column but it wouldn’t demonstrate the ORDER BY
clause. In some database, like Oracle and MySQL, the ORDER BY
lets you use column names, column aliases, or the position order of the column in the SELECT
clause. Unfortunately, this isn’t true across all databases.
The ORDER BY
clause does let you see that you can set nested ordering, like the numbers inside the data sets that are qualified by the LIST_SET
column. This is demonstrated in the output below.
ORDINAL_ID LIST_SET LIST_NAME LIST_VALUE ---------- -------------------- ---------- ---------- 1 Value Set A Zero 0 2 Value Set A One 1 3 Value Set A Two 2 4 Value Set A Three 3 5 Value Set A Four 4 6 Value Set A Five 5 7 Value Set A Six 6 8 Value Set A Seven 7 9 Value Set A Eight 8 10 Value Set A Nine 9 11 Value Set A 12 Value Set B Zero 0 13 Value Set B One 1 14 Value Set B Two 2 15 Value Set B Three 3 16 Value Set B Four 4 17 Value Set B Five 5 18 Value Set B Six 6 19 Value Set B Seven 7 20 Value Set B Eight 8 21 Value Set B Nine 9 22 Value Set B |
Single row aggregation function use ↓
This section discusses and demonstrates the five aggregation functions in the post. They return a single row of data. Like the other sections, you unfold this by clicking on the bold text above.
COUNT
function
The COUNT
function has two behaviors that may apply. One is when you count rows and the other is when you count values. Both return only a single row.
Counting by Reference: The following example demonstrates counting rows. In this scenario, the COUNT
function takes a single call parameter, which is an asterisk (*
). The asterisk actually counts the references to the rows in the table, and in Oracle is equivalent to COUNT(ROWID)
. The pseudo column ROWID
points to the physical block address where the row is stored in the database. As such, the asterisk acts very much like a pointer in the C or C++ language (that is if you substitute a block address for a memory address).
SELECT COUNT(*) FROM ordinal; |
It returns
COUNT(*) ---------- 22 |
When you count rows, you count them whether they contain values or not. It is possible that you could insert a meaningless row with null values for all columns in a table that’s unconstrained. Moreover, the COUNT
function counts all rows no matter whether contain any values.
Counting by Value: The following examples demonstrate approaches to counting values. It’s important to note that when you count values, you ignore null values. You can also count distinct things or all things when you count by value.
The LIST_SET
column doesn’t contain any null values (as you can see by inspecting the data set early in this blog page). The following counts the number of values in the LIST_SET
column. You could also substitute COUNT(ALL list_set)
and it would return the same thing because ALL
is the default and always provided unless you override it with a DISTINCT
keyword.
SELECT COUNT(list_set) FROM ordinal; |
It returns:
COUNT(*) ---------- 22 |
Since the LIST_SET
column only contains one of two values, you can count that to make sure. You do it with the following syntax:
SELECT COUNT(DISTINCT list_set) FROM ordinal; |
It returns:
COUNT(*) ---------- 2 |
The LIST_NAME
and LIST_VALUE
both contain two null values. Let’s see what they return when we count all or distinct values. This example simplifies it a step more by performing both counts side-by-side. This is possible because both columns return a single row.
SELECT COUNT(ALL list_name) AS name_number , COUNT(DISTINCT list_value) AS value_number FROM ordinal; |
It returns the twenty total string values found in the first column and the ten unique numeric values found in the second column. The COUNT
function with the DISTINCT
filter acts like a COUNT
function with a GROUP BY
clause on the LIST_VALUE
column, as you’ll see later in the discussion.
NAME_NUMBER VALUE_NUMBER ----------- ------------ 20 10 |
SUM
, AVERAGE
, MAX
, and MIN
functions
Math Operations by Value: These math aggregations are done with the SUM
, AVERAGE
, MAX
, and MIN
aggregation functions. They’re restricted to columns that contain numeric values. Each of them support the default ALL
or DISTINCT
keywords and they behave much as the COUNT
function discussed earlier.
There is a significant difference between their functionality. You can count unique strings but you can’t sum a column of numbers based on that other column without a GROUP BY
clause. You see examples of the GROUP BY
in the next part of this post.
Since these behave more or less the same, their examples are grouped together in sections by summing and averaging together, and taking the maximum and minimum values together. For reference, the ordinal numbers sum to 45, two sets of them are naturally 90; and the average of evenly dispersed numbers is half, or 45, and their unique values 4.5.
SELECT SUM(ALL list_value) AS sum_all , SUM(DISTINCT list_value) AS sum_distinct , AVG(ALL list_value) AS avg_all , AVG(DISTINCT list_value) AS avg_distinct FROM ordinal; |
It returns:
SUM_ALL SUM_DISTINCT AVG_ALL AVG_DISTINCT ---------- ------------ ---------- ------------ 90 45 4.5 4.5 |
The next section makes the same comparison with the MAX
, and MIN
functions. The maximum or minimums are the same because they the uniqueness doesn’t change the minimum or maximum of the sets.
SELECT MIN(ALL list_value) AS min_all , MIN(DISTINCT list_value) AS min_distinct , MAX(ALL list_value) AS max_all , MAX(DISTINCT list_value) AS max_distinct FROM ordinal; |
It returns:
MIN_ALL MIN_DISTINCT MAX_ALL MAX_DISTINCT ---------- ------------ ---------- ------------ 0 0 9 9 |
Multiple row aggregation function use ↓
This section discusses and demonstrates the five aggregation functions where they return more than a single row of data, which involves grouping by non-aggregated columns. This section covers the GROUP BY
and HAVING
clauses. Like the other sections, you unfold this by clicking on the bold text above.
COUNT
function
Counting by Groups: The following example demonstrates how we can count rows or values by groups. The example counts rows by reference and value, but it does so by grouping the return values on the unique string values in the LIST_SET
column.
1 2 3 4 5 | SELECT list_set AS grouping_by_column , COUNT(*) , COUNT(list_value) FROM ordinal GROUP BY list_set; |
It returns the following values:
GROUPING_BY_COLUMN COUNT(*) COUNT(LIST_VALUE) -------------------- ---------- ----------------- VALUE SET A 11 10 VALUE SET B 11 10 |
The results show that there are eleven rows for each set, and that ten of the rows contain values while one holds a null value.
You can also group on columns that contain null values or sets of columns. The following example groups by two columns, where one contains a null value.
1 2 3 4 5 6 7 8 | SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , COUNT(*) , COUNT(list_value) FROM ordinal WHERE list_set = 'Value Set A' GROUP BY list_set , list_name; |
It returns the following values:
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO COUNT(*) COUNT(LIST_VALUE) -------------------- ----------------- ---------- ----------------- VALUE SET A Zero 1 1 VALUE SET A Five 1 1 VALUE SET A Three 1 1 VALUE SET A Four 1 1 VALUE SET A One 1 1 VALUE SET A Two 1 1 VALUE SET A Eight 1 1 VALUE SET A Nine 1 1 VALUE SET A Seven 1 1 VALUE SET A Six 1 1 VALUE SET A 1 0 |
The last row returned is the one where the grouping by column value and counted column value are null. You should note that the GROUP BY
applies to a string column and the return set includes a row grouped by its null value. The count of null column values is always zero.
Another thing that you may have not noticed is that the rows are no longer in ascending numeric order by LIST_NAME
or LIST_VALUE
columns. There’s a multiple edge sword when you examine why this occurs. Basically with regard to the LIST_NAME
column, Oracle and many other databases use advanced sorting algorithms which may leave non-ordered sets as outcomes. The LIST_VALUE
column is not in the GROUP BY
clause and therefore can’t be used as an ORDER BY
key.
When you attempt to use a column outside of the GROUP BY
clause to order the return set, like this:
1 2 3 4 5 6 7 8 9 | SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , COUNT(*) , COUNT(list_value) FROM ordinal WHERE list_set = 'Value Set A' GROUP BY list_set , list_name ORDER BY list_value; |
You would see the following error in an Oracle Database 10g instance:
ORDER BY list_value * ERROR at line 9: ORA-00979: NOT a GROUP BY expression |
You can filter this particular set because the conditions are limited, but this wouldn’t be a good idea in many other solution sets. However, it certainly highlights some of the potential for extraordinary sorting of result sets.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 | SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , COUNT(*) , COUNT(list_value) FROM ordinal WHERE list_set = 'Value Set A' GROUP BY list_set , list_name ORDER BY CASE WHEN list_name = 'Zero' THEN 0 WHEN list_name = 'One' THEN 1 WHEN list_name = 'Two' THEN 2 WHEN list_name = 'Three' THEN 3 WHEN list_name = 'Four' THEN 4 WHEN list_name = 'Five' THEN 5 WHEN list_name = 'Six' THEN 6 WHEN list_name = 'Seven' THEN 7 WHEN list_name = 'Eight' THEN 8 WHEN list_name = 'Nine' THEN 9 END; |
It returns the following values:
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO COUNT(*) COUNT(LIST_VALUE) -------------------- ----------------- ---------- ----------------- VALUE SET A Zero 1 1 VALUE SET A One 1 1 VALUE SET A Two 1 1 VALUE SET A Three 1 1 VALUE SET A Four 1 1 VALUE SET A Five 1 1 VALUE SET A Six 1 1 VALUE SET A Seven 1 1 VALUE SET A Eight 1 1 VALUE SET A Nine 1 1 VALUE SET A 1 0 |
A better way to sort this type of data would involve putting it into an inline view and then joining it against the base data set. This allows you to use a non-grouping column as the sort key in an ORDER BY
clause. Here’s an example of that approach but note it does take system overhead and should only be done when SQL must perform the sort operation.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 | SELECT ilv.grouping_by_not_null , ilv.group_by_null_too , ilv.row_count , ilv.column_count FROM (SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , COUNT(*) AS ROW_COUNT , COUNT(list_value) AS column_count FROM ordinal WHERE list_set = 'Value Set A' GROUP BY list_set , list_name) ilv LEFT JOIN ordinal o ON ilv.grouping_by_not_null = o.list_set AND ilv.group_by_null_too = o.list_name ORDER BY o.list_value; |
That returns a perfectly ordered set because the LEFT JOIN
lets us capture the null value in the inline view without matching the two columns in the join condition. It returns the following set:
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO ROW_COUNT COLUMN_COUNT -------------------- ----------------- ---------- ------------ Value Set A Zero 1 1 Value Set A One 1 1 Value Set A Two 1 1 Value Set A Three 1 1 Value Set A Four 1 1 Value Set A Five 1 1 Value Set A Six 1 1 Value Set A Seven 1 1 Value Set A Eight 1 1 Value Set A Nine 1 1 Value Set A 1 0 |
SUM
, AVERAGE
, MAX
, and MIN
functions
Math Operations by Groups: These math aggregations are done with the SUM
, AVERAGE
, MAX
, and MIN
aggregation functions. Like the previous single row return examples, they’re restricted to columns that contain numeric values. While they also support the default ALL
or DISTINCT
keywords, it didn’t seem necessary to demonstrate more of those behaviors here.
The following sums, averages, and gets both the minimum and maximum value by unique LIST_SET
column values:
SELECT list_set AS grouping_by_not_null , SUM(list_value) AS ordinal_sum , AVG(list_value) AS ordinal_avg , MIN(list_value) AS ordinal_min , MAX(list_value) AS ordinal_max FROM ordinal GROUP BY list_set; |
It returns the following two rows:
GROUPING_BY_NOT_NULL ORDINAL_SUM ORDINAL_AVG ORDINAL_MIN ORDINAL_MAX -------------------- ----------- ----------- ----------- ----------- VALUE SET A 45 4.5 0 9 VALUE SET B 45 4.5 0 9 |
Expanding the GROUP BY
criteria to include the LIST_NAME
column, you see what happens when the SUM
and AVG
functions work with only null values. They always return null when adding only column values with a null value. This is handy as you’ll see in the selective aggregation discussion in a subsequent tutorial.
As done in the prior example, this filters out the second value set, to focus on what’s happening with the rows grouped on two columns.
1 2 3 4 5 6 7 8 9 10 | SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , SUM(list_value) AS ordinal_sum , AVG(list_value) AS ordinal_avg , MIN(list_value) AS ordinal_min , MAX(list_value) AS ordinal_max FROM ordinal WHERE NOT list_set = 'Value Set B' GROUP BY list_set , list_name; |
It returns the following set of records, which are clearly out of order based on the algorithms that optimize aggregation. It’s not hard to read because there aren’t too many rows returned.
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO ORDINAL_SUM ORDINAL_AVG ORDINAL_MIN ORDINAL_MAX -------------------- ----------------- ----------- ----------- ----------- ----------- VALUE SET A Zero 0 0 0 0 VALUE SET A Five 5 5 5 5 VALUE SET A Three 3 3 3 3 VALUE SET A Four 4 4 4 4 VALUE SET A One 1 1 1 1 VALUE SET A Two 2 2 2 2 VALUE SET A Eight 8 8 8 8 VALUE SET A Nine 9 9 9 9 VALUE SET A Seven 7 7 7 7 VALUE SET A Six 6 6 6 6 VALUE SET A |
If SQL must sort these into numeric order, it can do so like the prior example with the result from the COUNT
function. You must instrument a solution for it through an inline view. The following shows you how to do that.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 | SELECT ilv.grouping_by_not_null , ilv.group_by_null_too , ilv.ordinal_sum , ilv.ordinal_avg , ilv.ordinal_min , ilv.ordinal_max FROM (SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , SUM(list_value) AS ordinal_sum , AVG(list_value) AS ordinal_avg , MIN(list_value) AS ordinal_min , MAX(list_value) AS ordinal_max FROM ordinal WHERE list_set = 'Value Set B' GROUP BY list_set , list_name) ilv LEFT JOIN ordinal o ON ilv.grouping_by_not_null = o.list_set AND ilv.group_by_null_too = o.list_name ORDER BY o.list_value; |
Now you should see an ordered set like this:
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO ORDINAL_SUM ORDINAL_AVG ORDINAL_MIN ORDINAL_MAX -------------------- ----------------- ----------- ----------- ----------- ----------- VALUE SET B Zero 0 0 0 0 VALUE SET B One 1 1 1 1 VALUE SET B Two 2 2 2 2 VALUE SET B Three 3 3 3 3 VALUE SET B Four 4 4 4 4 VALUE SET B Five 5 5 5 5 VALUE SET B Six 6 6 6 6 VALUE SET B Seven 7 7 7 7 VALUE SET B Eight 8 8 8 8 VALUE SET B Nine 9 9 9 9 VALUE SET B |
Math Operations with the HAVING
clause: The HAVING
clause is very useful when you want to find rows that duplicate a column or set of column values. It’s also useful for sorting data sets.
The HAVING
clause effects all aggregate rows and acts somewhat like a filter in the WHERE
clause of a non-aggregated query. The difference is that the WHERE
clause filters the base result set, while the HAVING
clause filters the aggregated result set.
The following example captures the sum and average for only the odd LIST_VALUE
column values. It does that by using modular (or modulo) mathematics through the MOD
function in Oracle’s SQL implementation. Modulo math works by calculating the remainder of integer division. When you divide any number by two and the result is one, then you have an odd number.
Here’s the sample program that illustrates the HAVING
clause in a CASE
statement. The CASE
statement only adds the number when it is an odd number. Any even number is zeroed out. The comparison operator ensures that only rows are returned when the SUM
is greater than zero, which only happens for odd numbers.
1 2 3 4 5 6 7 8 9 10 11 | SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , SUM(list_value) AS ordinal_sum , AVG(list_value) AS ordinal_avg FROM ordinal WHERE list_set = 'Value Set A' HAVING SUM(CASE WHEN MOD(list_value,2) = 1 THEN list_value ELSE 0 END) > 0 GROUP BY list_set , list_name; |
The data set, as you’ve seen with others isn’t sorted in ascending order because of how the algorithms work. Since there are only five rows returned, it is easy to see that the HAVING
clause let us narrow the return set to odd numbers. What we might overlook is that the HAVING
is always a filtering statement. This means we need to have a comparative operator tied to the HAVING
return value.
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO ORDINAL_SUM ORDINAL_AVG -------------------- ----------------- ----------- ----------- Value Set A Five 5 5 Value Set A Three 3 3 Value Set A One 1 1 Value Set A Nine 9 9 Value Set A Seven 7 7 |
As you’ve seen in early examples, you can also put this inside an inline view and sort the aggregated results. Here’s the code to accomplish that.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 | SELECT ilv.grouping_by_not_null , ilv.group_by_null_too , ilv.ordinal_sum , ilv.ordinal_avg FROM (SELECT list_set AS grouping_by_not_null , list_name AS group_by_null_too , SUM(list_value) AS ordinal_sum , AVG(list_value) AS ordinal_avg FROM ordinal WHERE list_set = 'Value Set A' HAVING SUM(CASE WHEN MOD(list_value,2) = 1 THEN list_value ELSE 0 END) > 0 GROUP BY list_set , list_name) ilv LEFT JOIN ordinal o ON ilv.grouping_by_not_null = o.list_set AND ilv.group_by_null_too = o.list_name ORDER BY o.list_value; |
It naturally returns the ordered set of aggregated results from only odd numbers in the base set of values.
GROUPING_BY_NOT_NULL GROUP_BY_NULL_TOO ORDINAL_SUM ORDINAL_AVG -------------------- ----------------- ----------- ----------- Value Set A One 1 1 Value Set A Three 3 3 Value Set A Five 5 5 Value Set A Seven 7 7 Value Set A Nine 9 9 |
Setup Script ↓
The drop down items, unfold with source code to seed the examples.
Oracle SQL Join Supporting Script ↓
This has the setup script for the example tables.
-- Conditionally drop the table and sequence. BEGIN FOR i IN (SELECT TABLE_NAME FROM user_tables WHERE TABLE_NAME = 'ORDINAL') LOOP EXECUTE IMMEDIATE 'DROP TABLE '||i.table_name; END LOOP; FOR i IN (SELECT sequence_name FROM user_sequences WHERE sequence_name = 'ORDINAL_S1') LOOP EXECUTE IMMEDIATE 'DROP SEQUENCE '||i.sequence_name; END LOOP; END; / -- Create the aggregation sample table and sequence. CREATE TABLE ordinal ( ordinal_id NUMBER , list_set VARCHAR2(20) , list_name VARCHAR2(5) , list_value NUMBER); CREATE SEQUENCE ordinal_s1; -- Seeding values in the aggregation sample table. DECLARE -- Define local types. TYPE listg IS TABLE OF VARCHAR2(20); TYPE listn IS TABLE OF NUMBER; TYPE lists IS TABLE OF VARCHAR2(5); -- Declare local variables. groups LISTG := listg('Value Set A','Value Set B'); labels LISTS := lists('Zero','One','Two','Three','Four','Five','Six','Seven','Eight','Nine'); valuen LISTN := listn(0,1,2,3,4,5,6,7,8,9); BEGIN -- Outer loop sets the group level. FOR i IN 1..groups.COUNT LOOP -- Inner loop sets the row level. FOR j IN 1..labels.COUNT LOOP INSERT INTO ordinal VALUES (ordinal_s1.NEXTVAL,groups(i),labels(j),valuen(j)); END LOOP; -- Insert the null values for each group. INSERT INTO ordinal VALUES (ordinal_s1.NEXTVAL,groups(i),NULL,NULL); END LOOP; COMMIT; END; / -- Commit the inserts. COMMIT; |
MySQL SQL Join Supporting Script ↓
This has the setup script for the example tables.
-- Conditionally drop the table. SELECT 'DROP TABLE ordinal' AS Statement; DROP TABLE IF EXISTS ordinal; -- Create the table. SELECT 'CREATE TABLE ordinal' AS Statement; CREATE TABLE ordinal ( ordinal_id INT UNSIGNED PRIMARY KEY AUTO_INCREMENT , list_set VARCHAR(20) , list_name VARCHAR(5) , list_value INT UNSIGNED); -- Seed the row values. SELECT 'INSERT INTO ordinal' AS Statement; INSERT INTO ordinal VALUES (NULL,'Value Set A','Zero','0') ,(NULL,'Value Set A','One','1') ,(NULL,'Value Set A','Two','2') ,(NULL,'Value Set A','Three','3') ,(NULL,'Value Set A','Four','4') ,(NULL,'Value Set A','Five','5') ,(NULL,'Value Set A','Six','6') ,(NULL,'Value Set A','Seven','7') ,(NULL,'Value Set A','Eight','8') ,(NULL,'Value Set A','Nine','9') ,(NULL,'Value Set A',NULL,NULL) ,(NULL,'Value Set B','Zero','0') ,(NULL,'Value Set B','One','1') ,(NULL,'Value Set B','Two','2') ,(NULL,'Value Set B','Three','3') ,(NULL,'Value Set B','Four','4') ,(NULL,'Value Set B','Five','5') ,(NULL,'Value Set B','Six','6') ,(NULL,'Value Set B','Seven','7') ,(NULL,'Value Set B','Eight','8') ,(NULL,'Value Set B','Nine','9') ,(NULL,'Value Set B',NULL,NULL); -- Commit the inserts. COMMIT; |
Multi-row Merge in MySQL
After I wrote the post for students on the multiple row MERGE
statement for an upload through an external table in Oracle, I thought to check how it might be done with MySQL. More or less because I try to keep track of how things are done in several databases.
MySQL’s equivalent to a MERGE
statement is an INSERT
statement with an ON DUPLICATE KEY
clause, which I blogged about a while back. You may also use the REPLACE INTO
when you want to merge more than one row. At the time that I wrote this, I thought there wasn’t support for an INSERT ON DUPLICATE KEY
clause statement with a subquery but I found that I was wrong. Fortunately, somebody posted a comment to remind me about this and now both solutions are here for anybody that would like them.
The workaround with a VALUES
clause was to write a stored procedure with two cursor loops, explicitly pass the values from the cursor to local variables, and then put the local variables in the VALUES
clause. I’ll post the other with a subquery soon. On parity, clearly Oracle’s MERGE
statement (shown here) is far superior than MySQL’s approach.
Demonstration
Here are the steps to accomplish an import/upload with the INSERT
statement and ON DUPLICATE KEY
clause. In this example, you upload data from a flat file, or Comma Separated Value (CSV) file to a denormalized table (actually in unnormalized form). This type of file upload transfers information that doesn’t have surrogate key values. You have to create those in the scope of the transformation to the normalized tables.
Step #1 : Position your CSV file in the physical directory
After creating the virtual directory, copy the following contents into a file named kingdom_mysql_import.csv
in the C:\Data\Download
directory or folder. If you have Windows UAC enabled in Windows Vista or 7, you should disable it before performing this step.
Place the following in the kingdom_mysql_import.csv
file. The trailing commas are meaningful in MySQL and avoid problems when reading CSV files.
Narnia, 77600,'Peter the Magnificent',12720320,12920609, Narnia, 77600,'Edmund the Just',12720320,12920609, Narnia, 77600,'Susan the Gentle',12720320,12920609, Narnia, 77600,'Lucy the Valiant',12720320,12920609, Narnia, 42100,'Peter the Magnificent',15310412,15310531, Narnia, 42100,'Edmund the Just',15310412,15310531, Narnia, 42100,'Susan the Gentle',15310412,15310531, Narnia, 42100,'Lucy the Valiant',15310412,15310531, Camelot, 15200,'King Arthur',06310310,06861212, Camelot, 15200,'Sir Lionel',06310310,06861212, Camelot, 15200,'Sir Bors',06310310,06351212, Camelot, 15200,'Sir Bors',06400310,06861212, Camelot, 15200,'Sir Galahad',06310310,06861212, Camelot, 15200,'Sir Gawain',06310310,06861212, Camelot, 15200,'Sir Tristram',06310310,06861212, Camelot, 15200,'Sir Percival',06310310,06861212, Camelot, 15200,'Sir Lancelot',06700930,06821212, |
Step #2 : Connect as the student
user
Disconnect and connect as the student user, or reconnect as the student user. The reconnect syntax that protects your password is:
mysql -ustudent -p |
Connect to the sampledb
database, like so:
mysql> USE sampledb; |
Step #3 : Run the script that creates tables and sequences
Copy the following into a create_mysql_kingdom_upload.sql
file within a directory of your choice. Then, run it as the student
account.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 | -- This enables dropping tables with foreign key dependencies. -- It is specific to the InnoDB Engine. SET FOREIGN_KEY_CHECKS = 0; -- Conditionally drop objects. SELECT 'KINGDOM' AS "Drop Table"; DROP TABLE IF EXISTS KINGDOM; SELECT 'KNIGHT' AS "Drop Table"; DROP TABLE IF EXISTS KNIGHT; SELECT 'KINGDOM_KNIGHT_IMPORT' AS "Drop Table"; DROP TABLE IF EXISTS KINGDOM_KNIGHT_IMPORT; -- Create normalized kingdom table. SELECT 'KINGDOM' AS "Create Table"; CREATE TABLE kingdom ( kingdom_id INT UNSIGNED PRIMARY KEY AUTO_INCREMENT , kingdom_name VARCHAR(20) , population INT UNSIGNED) ENGINE=INNODB; -- Create normalized knight table. SELECT 'KNIGHT' AS "Create Table"; CREATE TABLE knight ( knight_id INT UNSIGNED PRIMARY KEY AUTO_INCREMENT , knight_name VARCHAR(24) , kingdom_allegiance_id INT UNSIGNED , allegiance_start_date DATE , allegiance_end_date DATE , CONSTRAINT fk_kingdom FOREIGN KEY (kingdom_allegiance_id) REFERENCES kingdom (kingdom_id)) ENGINE=INNODB; -- Create external import table in memory only - disappears after rebooting the mysqld service. SELECT 'KINGDOM_KNIGHT_IMPORT' AS "Create Table"; CREATE TABLE kingdom_knight_import ( kingdom_name VARCHAR(20) , population INT UNSIGNED , knight_name VARCHAR(24) , allegiance_start_date DATE , allegiance_end_date DATE) ENGINE=MEMORY; |
Step #4 : Load the data into your target upload table
There a number of things that could go wrong but when you choose LOCAL
there generally aren’t any problems. Run the following query from the student
account while using the sampledb
database, and check whether or not you can access the kingdom_import.csv
file.
1 2 3 4 5 6 | LOAD DATA LOCAL INFILE 'c:/Data/kingdom_mysql_import.csv' INTO TABLE kingdom_knight_import FIELDS TERMINATED BY ',' ENCLOSED BY '"' ESCAPED BY '\\' LINES TERMINATED BY '\r\n'; |
Step #5 : Create the upload procedure
Copy the following into a create_mysql_upload_procedure.sql
file within a directory of your choice. You should note that unlike Oracle’s MERGE
statement, this is done with the ON DUPLICATE KEY
clause and requires actual values not a source query. This presents few options other than a stored routine, known as a stored procedure. As you can see from the code, there’s a great deal of complexity to the syntax and a much more verbose implementation than Oracle’s equivalent PL/SQL.
Then, run it as the student
account. As you look at the structure to achieve this simple thing, the long standing complaint about PL/SQL being a verbose language comes to mind. Clearly, stored procedures are new to MySQL but they’re quite a bit more verbose than PL/SQL.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 | -- Conditionally drop the procedure. SELECT 'UPLOAD_KINGDOM' AS "Drop Procedure"; DROP PROCEDURE IF EXISTS upload_kingdom; -- Reset the execution delimiter to create a stored program. DELIMITER $$ -- The parentheses after the procedure name must be there or the MODIFIES SQL DATA raises an compile time exception. CREATE PROCEDURE upload_kingdom() MODIFIES SQL DATA BEGIN /* Declare local variables. */ DECLARE lv_kingdom_id INT UNSIGNED; DECLARE lv_kingdom_name VARCHAR(20); DECLARE lv_population INT UNSIGNED; DECLARE lv_knight_id INT UNSIGNED; DECLARE lv_knight_name VARCHAR(24); DECLARE lv_kingdom_allegiance_id INT UNSIGNED; DECLARE lv_allegiance_start_date DATE; DECLARE lv_allegiance_end_date DATE; /* Declare a handler variables. */ DECLARE duplicate_key INT DEFAULT 0; DECLARE foreign_key INT DEFAULT 0; DECLARE fetched INT DEFAULT 0; /* Cursors must come after variables and before event handlers. */ /* Declare a SQL cursor with a left join on the natural key. */ DECLARE kingdom_cursor CURSOR FOR SELECT DISTINCT k.kingdom_id , kki.kingdom_name , kki.population FROM kingdom_knight_import kki LEFT JOIN kingdom k ON kki.kingdom_name = k.kingdom_name AND kki.population = k.population; /* Declare a SQL cursor with a join on the natural key. */ DECLARE knight_cursor CURSOR FOR SELECT kn.knight_id , kki.knight_name , k.kingdom_id , kki.allegiance_start_date AS start_date , kki.allegiance_end_date AS end_date FROM kingdom_knight_import kki INNER JOIN kingdom k ON kki.kingdom_name = k.kingdom_name AND kki.population = k.population LEFT JOIN knight kn ON k.kingdom_id = kn.kingdom_allegiance_id AND kki.knight_name = kn.knight_name AND kki.allegiance_start_date = kn.allegiance_start_date AND kki.allegiance_end_date = kn.allegiance_end_date; /* Event handlers must always be last in the declaration section. */ /* Declare a duplicate key handler */ DECLARE CONTINUE HANDLER FOR 1062 SET duplicate_key = 1; DECLARE CONTINUE HANDLER FOR 1216 SET foreign_key = 1; /* Declare a not found record handler to close a cursor loop. */ DECLARE CONTINUE HANDLER FOR NOT FOUND SET fetched = 1; /* ---------------------------------------------------------------------- */ /* Start transaction context. */ START TRANSACTION; /* Set savepoint. */ SAVEPOINT both_or_none; /* Open a local cursor. */ OPEN kingdom_cursor; cursor_kingdom: LOOP FETCH kingdom_cursor INTO lv_kingdom_id , lv_kingdom_name , lv_population; /* Place the catch handler for no more rows found immediately after the fetch operation. */ IF fetched = 1 THEN LEAVE cursor_kingdom; END IF; INSERT INTO kingdom VALUES ( lv_kingdom_id , lv_kingdom_name , lv_population ) ON DUPLICATE KEY UPDATE kingdom_name = lv_kingdom_name; END LOOP cursor_kingdom; CLOSE kingdom_cursor; /* Reset the continue handler to zero. */ SET fetched = 0; /* Open a local cursor. */ OPEN knight_cursor; cursor_knight: LOOP /* Fetch records until they're all read, and a NOT FOUND SET is returned. */ FETCH knight_cursor INTO lv_knight_id , lv_knight_name , lv_kingdom_allegiance_id , lv_allegiance_start_date , lv_allegiance_end_date; /* Place the catch handler for no more rows found immediately after the fetch operation. */ IF fetched = 1 THEN LEAVE cursor_knight; END IF; INSERT INTO knight VALUES ( lv_knight_id , lv_knight_name , lv_kingdom_allegiance_id , lv_allegiance_start_date , lv_allegiance_end_date ) ON DUPLICATE KEY UPDATE knight_name = lv_knight_name; END LOOP cursor_knight; CLOSE knight_cursor; /* Reset the continue handler to zero. */ SET fetched = 0; /* ---------------------------------------------------------------------- */ /* This acts as an exception handling block. */ IF duplicate_key = 1 OR foreign_key = 1 THEN /* This undoes all DML statements to this point in the procedure. */ ROLLBACK TO SAVEPOINT both_or_none; ELSE /* This commits the writes. */ COMMIT; END IF; END; $$ -- Reset the delimiter to the default. DELIMITER ; |
Here’s the better option with an embedded query:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 | -- Conditionally drop the procedure. SELECT 'UPLOAD_KINGDOM' AS "Drop Procedure"; DROP PROCEDURE IF EXISTS upload_kingdom; -- Reset the execution delimiter to create a stored program. DELIMITER $$ -- The parentheses after the procedure name must be there or the MODIFIES SQL DATA raises an compile time exception. CREATE PROCEDURE upload_kingdom() MODIFIES SQL DATA BEGIN /* Declare a handler variables. */ DECLARE duplicate_key INT DEFAULT 0; DECLARE foreign_key INT DEFAULT 0; /* Declare a duplicate key handler */ DECLARE CONTINUE HANDLER FOR 1062 SET duplicate_key = 1; DECLARE CONTINUE HANDLER FOR 1216 SET foreign_key = 1; /* ---------------------------------------------------------------------- */ /* Start transaction context. */ START TRANSACTION; /* Set savepoint. */ SAVEPOINT both_or_none; /* Using subqueries update the targets. */ INSERT INTO knight ( SELECT kn.knight_id , kki.knight_name , k.kingdom_id , kki.allegiance_start_date AS start_date , kki.allegiance_end_date AS end_date FROM kingdom_knight_import kki INNER JOIN kingdom k ON kki.kingdom_name = k.kingdom_name AND kki.population = k.population LEFT JOIN knight kn ON k.kingdom_id = kn.kingdom_allegiance_id AND kki.knight_name = kn.knight_name AND kki.allegiance_start_date = kn.allegiance_start_date AND kki.allegiance_end_date = kn.allegiance_end_date ) ON DUPLICATE KEY UPDATE knight_id = kn.knight_id; INSERT INTO knight ( SELECT kn.knight_id , kki.knight_name , k.kingdom_id , kki.allegiance_start_date AS start_date , kki.allegiance_end_date AS end_date FROM kingdom_knight_import kki INNER JOIN kingdom k ON kki.kingdom_name = k.kingdom_name AND kki.population = k.population LEFT JOIN knight kn ON k.kingdom_id = kn.kingdom_allegiance_id AND kki.knight_name = kn.knight_name AND kki.allegiance_start_date = kn.allegiance_start_date AND kki.allegiance_end_date = kn.allegiance_end_date ) ON DUPLICATE KEY UPDATE knight_id = kn.knight_id; /* ---------------------------------------------------------------------- */ /* This acts as an exception handling block. */ IF duplicate_key = 1 OR foreign_key = 1 THEN /* This undoes all DML statements to this point in the procedure. */ ROLLBACK TO SAVEPOINT both_or_none; ELSE /* This commits the writes. */ COMMIT; END IF; END; $$ -- Reset the delimiter to the default. DELIMITER ; |
Step #6 : Run the upload procedure
You can run the file by calling the stored procedure built by the script. The procedure ensures that records are inserted or updated into their respective tables.
CALL upload_kingdom; |
Step #7 : Test the results of the upload procedure
You can test whether or not it worked by running the following queries.
-- Check the kingdom table. SELECT * FROM kingdom; SELECT * FROM knight; |
It should display the following information:
+------------+--------------+------------+ | kingdom_id | kingdom_name | population | +------------+--------------+------------+ | 1 | Narnia | 77600 | | 2 | Narnia | 42100 | | 3 | Camelot | 15200 | +------------+--------------+------------+ +-----------+-------------------------+-----------------------+-----------------------+---------------------+ | knight_id | knight_name | kingdom_allegiance_id | allegiance_start_date | allegiance_end_date | +-----------+-------------------------+-----------------------+-----------------------+---------------------+ | 1 | 'Peter the Magnificent' | 1 | 1272-03-20 | 1292-06-09 | | 2 | 'Edmund the Just' | 1 | 1272-03-20 | 1292-06-09 | | 3 | 'Susan the Gentle' | 1 | 1272-03-20 | 1292-06-09 | | 4 | 'Lucy the Valiant' | 1 | 1272-03-20 | 1292-06-09 | | 5 | 'Peter the Magnificent' | 2 | 1531-04-12 | 1531-05-31 | | 6 | 'Edmund the Just' | 2 | 1531-04-12 | 1531-05-31 | | 7 | 'Susan the Gentle' | 2 | 1531-04-12 | 1531-05-31 | | 8 | 'Lucy the Valiant' | 2 | 1531-04-12 | 1531-05-31 | | 9 | 'King Arthur' | 3 | 0631-03-10 | 0686-12-12 | | 10 | 'Sir Lionel' | 3 | 0631-03-10 | 0686-12-12 | | 11 | 'Sir Bors' | 3 | 0631-03-10 | 0635-12-12 | | 12 | 'Sir Bors' | 3 | 0640-03-10 | 0686-12-12 | | 13 | 'Sir Galahad' | 3 | 0631-03-10 | 0686-12-12 | | 14 | 'Sir Gawain' | 3 | 0631-03-10 | 0686-12-12 | | 15 | 'Sir Tristram' | 3 | 0631-03-10 | 0686-12-12 | | 16 | 'Sir Percival' | 3 | 0631-03-10 | 0686-12-12 | | 17 | 'Sir Lancelot' | 3 | 0670-09-30 | 0682-12-12 | +-----------+-------------------------+-----------------------+-----------------------+---------------------+ |
You can rerun the procedure to check that it doesn’t alter any information, then you could add a new knight to test the insertion portion.
Merge Statement for ETL
While working through examples for my students on uploading data, I thought it would be interesting to demonstrate how to create a re-runnable upload. Especially when chatting with a friend who was unaware that you could use joins inside the source element of a MERGE
statement. Naturally, the MERGE
statement seemed like the best approach in an Oracle database because with my criteria:
- The source file would not include any surrogate key values.
- The source file would have denormalized record sets with data that should belong to parent and child tables, technically unnormalized form (UNF).
- Primary and foreign key values would be determined on load to the tables.
- There could be a one-to-many relationship between the parent and child tables in the original source.
- Subsequent data sets may replicate data already seeded or not in the tables.
- Avoid any complex PL/SQL structures.
Step #1 : Create a Virtual Directory
You can create a virtual directory without a physical directory but it won’t work when you try to access it. Therefore, you should create the physical directory first. Assuming you’ve created a C:\Data\Download
file directory on the Windows platform, you can then create a virtual directory and grant permissions to the student
user as the SYS
privileged user. The syntax for these steps is:
CREATE DIRECTORY download AS 'C:\Data\Download'; GRANT READ, WRITE ON DIRECTORY download TO student; |
If you want more detail on these steps, check this older post on the blog.
Step #2 : Create the Target and External Tables
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 | -- Conditionally drop tables and sequences. BEGIN FOR i IN (SELECT TABLE_NAME FROM user_tables WHERE TABLE_NAME IN ('KINGDOM','KNIGHT','KINGDOM_KNIGHT_IMPORT')) LOOP EXECUTE IMMEDIATE 'DROP TABLE '||i.table_name||' CASCADE CONSTRAINTS'; END LOOP; FOR i IN (SELECT sequence_name FROM user_sequences WHERE sequence_name IN ('KINGDOM_S1','KNIGHT_S1')) LOOP EXECUTE IMMEDIATE 'DROP SEQUENCE '||i.sequence_name; END LOOP; END; / -- Create normalized kingdom table. CREATE TABLE kingdom ( kingdom_id NUMBER , kingdom_name VARCHAR2(20) , population NUMBER); -- Create a sequence for the kingdom table. CREATE SEQUENCE kingdom_s1; -- Create normalized knight table. CREATE TABLE knight ( knight_id NUMBER , knight_name VARCHAR2(24) , kingdom_allegiance_id NUMBER , allegiance_start_date DATE , allegiance_end_date DATE); -- Create a sequence for the knight table. CREATE SEQUENCE knight_s1; -- Create external import table. CREATE TABLE kingdom_knight_import ( kingdom_name VARCHAR2(20) , population NUMBER , knight_name VARCHAR2(24) , allegiance_start_date DATE , allegiance_end_date DATE) ORGANIZATION EXTERNAL ( TYPE oracle_loader DEFAULT DIRECTORY download ACCESS PARAMETERS ( RECORDS DELIMITED BY NEWLINE CHARACTERSET US7ASCII BADFILE 'DOWNLOAD':'kingdom_import.bad' DISCARDFILE 'DOWNLOAD':'kingdom_import.dis' LOGFILE 'DOWNLOAD':'kingdom_import.log' FIELDS TERMINATED BY ',' OPTIONALLY ENCLOSED BY "'" MISSING FIELD VALUES ARE NULL ) LOCATION ('kingdom_import.csv')) REJECT LIMIT UNLIMITED; |
Step #3 : Create a Procedure to ensure an all or nothing transaction
The procedure ensures that an all or nothing transaction occurs to both tables. Inside the procedure you have two MERGE
statements.
The first MERGE
statement uses a LEFT JOIN
to ensure that any new kingdom_name
will be added to the kingdom
table. The kingdom_name
and population
columns are the natural key in this model. The second MERGE
statement uses an INNER JOIN
to ensure that knight
rows are only inserted when they belong to an existing kingdom_name
. Naturally, the primary key capture occurs in this statement and it maps the primary key to the foreign key column in the knight
table.
The complete procedure code follows:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 | -- Create a procedure to wrap the transaction. CREATE OR REPLACE PROCEDURE upload_kingdom IS BEGIN -- Set save point for an all or nothing transaction. SAVEPOINT starting_point; -- Insert or update the table, which makes this rerunnable when the file hasn't been updated. MERGE INTO kingdom target USING (SELECT DISTINCT k.kingdom_id , kki.kingdom_name , kki.population FROM kingdom_knight_import kki LEFT JOIN kingdom k ON kki.kingdom_name = k.kingdom_name AND kki.population = k.population) SOURCE ON (target.kingdom_id = SOURCE.kingdom_id) WHEN MATCHED THEN UPDATE SET kingdom_name = SOURCE.kingdom_name WHEN NOT MATCHED THEN INSERT VALUES ( kingdom_s1.nextval , SOURCE.kingdom_name , SOURCE.population); -- Insert or update the table, which makes this rerunnable when the file hasn't been updated. MERGE INTO knight target USING (SELECT k.kingdom_id , kki.knight_name , kki.allegiance_start_date AS start_date , kki.allegiance_end_date AS end_date FROM kingdom_knight_import kki INNER JOIN kingdom k ON kki.kingdom_name = k.kingdom_name AND kki.population = k.population) SOURCE ON (target.kingdom_allegiance_id = SOURCE.kingdom_id) WHEN MATCHED THEN UPDATE SET allegiance_start_date = SOURCE.start_date , allegiance_end_date = SOURCE.end_date WHEN NOT MATCHED THEN INSERT VALUES ( knight_s1.nextval , SOURCE.knight_name , SOURCE.kingdom_id , SOURCE.start_date , SOURCE.end_date); -- Save the changes. COMMIT; EXCEPTION WHEN OTHERS THEN ROLLBACK TO starting_point; RETURN; END; / |
Step #4 : Test the Process
You can test it by calling the procedure. Rerunning it will demonstrate that it doesn’t violate any of the rules.
EXECUTE upload_kingdom; |
As always, I hope this is useful to somebody besides me.