I am doing an insert condition in oracle that when the record based on job and subjob doesnt exists, it shall insert otherwise, if it exists then it should update the rest of the value.
this is my procedure,
CREATE OR REPLACE PROCEDURE WELTESADMIN.SP_JOB_INS
(
JOB_V VARCHAR2,
SUBJOB_V VARCHAR2,
STARTDATE_V DATE,
ENDDATE_V DATE,
JOBWEIGHT_V NUMBER
)
AS BEGIN INSERT INTO PROJECT_SPAN (JOB, SUBJOB, STARTDATE, ENDDATE, WEIGHT) VALUES (JOB_V, SUBJOB_V, STARTDATE_V, ENDDATE_V, JOBWEIGHT_V);
EXCEPTION WHEN DUP_VAL_ON_INDEX THEN
UPDATE PROJECT_SPAN SET STARTDATE = STARTDATE_V, ENDDATE = ENDDATE_V, WEIGHT = JOBWEIGHT_V WHERE JOB = JOB_V AND SUBJOB = SUBJOB_V;
END;
/
and this is from PHP Call,
$insertJobSpanSql = "BEGIN SP_JOB_INS(:JOB, :SUBJOB, :SDATE, :EDATE, :WT); END;";
$insertJobSpanParse = oci_parse($conn, $insertJobSpanSql);
oci_bind_by_name($insertJobSpanParse, ":JOB", $jobValue);
oci_bind_by_name($insertJobSpanParse, ":SUBJOB", $subJobValue);
oci_bind_by_name($insertJobSpanParse, ":SDATE", $startDateValue);
oci_bind_by_name($insertJobSpanParse, ":EDATE", $endDateValue);
oci_bind_by_name($insertJobSpanParse, ":WT", $jobWeightValue);
$insertJobSpanRes = oci_execute($insertJobSpanParse);
if ($insertJobSpanRes){
oci_commit($conn);
} else {
oci_rollback($conn);
}
problem is it keeps inserting new row with the same job and subjob value. it should be an update to the new value.
First of all, I recommend to use MERGE in such cases:
CREATE OR REPLACE PROCEDURE WELTESADMIN.SP_JOB_INS
(
JOB_V VARCHAR2,
SUBJOB_V VARCHAR2,
STARTDATE_V DATE,
ENDDATE_V DATE,
JOBWEIGHT_V NUMBER
) AS
BEGIN
merge into PROJECT_SPAN ps
using (select JOB_V, SUBJOB_V, STARTDATE_V, ENDDATE_V, JOBWEIGHT_V
from dual) new_val
on (ps.SUBJOB = new_val.SUBJOB_V and ps.JOB = new_val.JOB_V)
when matched then update
set STARTDATE = new_val.STARTDATE_V,
ENDDATE = new_val.ENDDATE_V,
WEIGHT = new_val.JOBWEIGHT_V
when not matched then insert (JOB, SUBJOB, STARTDATE, ENDDATE, WEIGHT)
values (new_val.JOB_V, new_val.SUBJOB_V, new_val.STARTDATE_V,
new_val.ENDDATE_V, new_val.JOBWEIGHT_V );
END;
/
If it still not updating values, use package DBMS_OUTPUT or logging into a table to make sure, that new and old JOB and SUBJOB really the same.
Oracle raises DUP_VAL_ON_INDEX when a DDL statement violates a primary key or unique key constraint. So for your code to work you need to define a primary key on (job, subjob) or else build a unique index on that pair.
Oracle provides an easier way of implementing upserts, in the form of the MERGE statement:
CREATE OR REPLACE PROCEDURE WELTESADMIN.SP_JOB_INS
(
JOB_V VARCHAR2,
SUBJOB_V VARCHAR2,
STARTDATE_V DATE,
ENDDATE_V DATE,
JOBWEIGHT_V NUMBER
)
AS BEGIN
merge into project_span ps
using ( select job_v, subjob_v, startdate_v, enddate_v, jobweight_v from dual ) q
on (ps.job = q.job_v
and ps.subjob = q.subjob_v)
when not matched then
insert (job, subjob, startdate, enddate, weight)
values (q.job_v, q.subjob_v, q.startdate_v, q.enddate_v, q.jobweight_v);
when matched then
update
set startdate = q.startdate_v
, enddate = q.enddate_v
, weight = q.jobweight_v
;
END;
The MERGE statement is in the documentation. Find out more.
Related
I want to use one form to insert into two different Microsoft sql tables. I tryed to use 2 inserts, but didnt work.
if (isset($_GET['submit'])) {
$sth = $connection->prepare("INSERT INTO DB.dbo.Fehler (QualiID, TestaufstellungID, ModulinfoID, failAfter, Datum, Verbleib, DUTNr) VALUES ($QualiID, $TestaufstellungID,$ModulinfoID,'$failAfter','$Datum','$Verbleib','$DUTNr')");
echo "INSERT INTO DB.dbo.Fehler (QualiID, TestaufstellungID, ModulinfoID, failAfter, Datum, Verbleib, DUTNr) VALUES ($QualiID, $TestaufstellungID,$ModulinfoID,'$failAfter',$Datum,'$Verbleib','$DUTNr')";
$sth->execute();
if($sth)
{
echo "";
}
else
{
echo sqlsrv_errors();
}
$MID = $connection->prepare("MAX(MID) as MID FROM DB.dbo.Fehler WHERE DB.dbo.Fehler.TestaufstellungID = '". $TestaufstellungID . "'");
$MID->execute();
$sth2 = $connection->prepare("INSERT INTO DB.dbo.Fehlerinfo (MID, Tester, Test, Ausfallbedingungen, Fehlerbeschreibung, Ersteller) VALUES ($MID, '$Tester','$Test','$Ausfallbedingungen','$Fehlerbeschreibung','$Ersteller')");
$sth2->execute();
To understand MID is the Primary key of table Fehler and ist the foreign key in the second table Fehlerinfo
Thats why i have the select work around to get the last MID and want to save it in a variable $MID to insert it into the second table.
Is there a smarter solution possible?
As I mentioned in the comments, generally the better way is to do the insert in one batch. This is very over simplified, however, should put you in the right direction. Normally you would likely be passing the values for the Foreign Table in a Table Value Parameter (due to the Many to One relationship) and would encapsulate the entire thing in a TRY...CATCH and possibly a stored procedure.
I can't write this in PHP, as my knowledge of it is rudimentary, but this should get you on the right path to understanding:
USE Sandbox;
--Couple of sample tables
CREATE TABLE dbo.PrimaryTable (SomeID int IDENTITY(1,1),
SomeString varchar(10),
CONSTRAINT PK_PTID PRIMARY KEY NONCLUSTERED (SomeID));
CREATE TABLE dbo.ForeignTable (AnotherID int IDENTITY(1,1),
ForeignID int,
AnotherString varchar(10),
CONSTRAINT PK_FTID PRIMARY KEY NONCLUSTERED(AnotherID),
CONSTRAINT FK_FTPT FOREIGN KEY (ForeignID)
REFERENCES dbo.PrimaryTable(SomeID));
GO
--single batch example
--Declare input parameters and give some values
--These would be the values coming from your application
DECLARE #SomeString varchar(10) = 'abc',
#AnotherString varchar(10) = 'def';
--Create a temp table or variable for the output of the ID
DECLARE #ID table (ID int);
--Insert the data and get the ID at the same time:
INSERT INTO dbo.PrimaryTable (SomeString)
OUTPUT inserted.SomeID
INTO #ID
SELECT #SomeString;
--#ID now has the inserted ID(s)
--Use it to insert into the other table
INSERT INTO dbo.ForeignTable (ForeignID,AnotherString)
SELECT ID,
#AnotherString
FROM #ID;
GO
--Check the data:
SELECT *
FROM dbo.PrimaryTable PT
JOIN dbo.ForeignTable FT ON PT.SomeID = FT.ForeignID;
GO
--Clean up
DROP TABLE dbo.ForeignTable;
DROP TABLE dbo.PrimaryTable;
As i mentioned the answer how it works for me fine atm.
if (isset($_GET['submit'])) {
$failInsert = ("INSERT INTO DB.dbo.Fehler (QualiID, TestaufstellungID, ModulinfoID, failAfter, Datum, Verbleib, DUTNr) VALUES ($QualiID, $TestaufstellungID,$ModulinfoID,'$failAfter','$Datum','$Verbleib','$DUTNr')");
$failInsert .= ("INSERT INTO DB.dbo.Fehlerinfo (MID, Tester, Test, Ausfallbedingungen, Fehlerbeschreibung, Ersteller) VALUES (NULL, '$Tester','$Test','$Ausfallbedingungen','$Fehlerbeschreibung','$Ersteller')");
$failInsert .= ("UPDATE DB.dbo.Fehlerinfo SET DB.dbo.Fehlerinfo.MID = i.MID FROM (SELECT MAX(MID)as MID FROM DB.dbo.Fehler) i WHERE DB.dbo.Fehlerinfo.TestID = ( SELECT MAX(TestID) as TestID FROM DB.dbo.Fehlerinfo)");
$sth = $connection->prepare($failInsert);
$sth->execute();
}
I have this PHP function below that Creates a new Project Task MySQL database record if one with the same ID does not exist already.
If one does already exist with the same ID, then it instead UPDATES the MySQL record.
This code has a lot of time invested into it as it is very important that the Task record's date_modified field is to ONLY be updated in the event that any of these 5 fields have changed to a new value!:
- name
- description
- status
- type
- priority
If all of these fields have the same value as the previous value when making an UPDATE, for example if the sort_order value is changed then it would NOT update the date_modified field!
It works great after lots of work crafting it to work just right. I mention all this as it is important to not break this functionality!
Now this question is because this function can CREATE and UPDATE Task records in MySQL.
I now have the need to know when each of these events happens.
If a new Task record is CREATED then I need to know that a new record was created.
If the Task record is simply UPDATED then I need to know that the record had already existed and was simply Updated.
Is there anyway to determine which of the 2 events happened with what I have?
private function _addOrUpdateTaskRecord($taskId, $projectId, $created_by_user_id, $modified_user_id, $name, $description, $status, $priority, $type, $date_entered, $date_modified, $date_started, $date_completed, $date_due, $milestone_id, $assigned_user_id, $sort_order, $heading){
$sql = "
INSERT INTO
$this->tasksDbTableName(task_id, project_id, created_by_user_id, modified_user_id, name, description, status, priority, type, date_entered, date_modified, date_started, date_completed, date_due, milestone_id, assigned_user_id, sort_order, heading)
VALUES
('$taskId', '$projectId', '$created_by_user_id', '$modified_user_id', '$name', '$description', '$status', '$priority', '$type', UTC_TIMESTAMP(), UTC_TIMESTAMP(), '$date_started', '$date_completed', '$date_due', '$milestone_id', '$assigned_user_id', '$sort_order', '$heading')
ON DUPLICATE KEY UPDATE
date_modified = (CASE
WHEN name <> values(name)
OR description <> values(description)
OR status <> values(status)
OR type <> values(type)
OR priority <> values(priority)
THEN UTC_TIMESTAMP()
ELSE date_modified
END),
modified_user_id='$modified_user_id',
name='$name',
description='$description',
status='$status',
priority='$priority',
type='$type',
date_started='$date_started',
date_completed='$date_completed',
date_due='$date_due',
milestone_id='$milestone_id',
assigned_user_id='$assigned_user_id',
sort_order='$sort_order',
heading='$heading'";
$insertOrUpdateTasks = $this->db->query($sql);
return $insertOrUpdateTasks;
}
Update
I am adding my MySQL Triggers code here for others to see as it might help someone someday with similar issue, or myself so I don't end up posting a question I have a solution to again!
Trigger to create an Event record when a Task is CREATED
DROP TRIGGER IF EXISTS project_task_new_event;
CREATE TRIGGER `project_task_new_event` AFTER INSERT ON `apoll_web_projects_tasks`
FOR EACH ROW
INSERT INTO apoll_web_projects_events (event_type, project_id, task_id, created_by_user_id, description, date_created) VALUES ('6', NEW.project_id, NEW.task_id, NEW.modified_user_id, NEW.name, UTC_TIMESTAMP());
Trigger to create an Event record when a Task is UPDATED
DROP TRIGGER IF EXISTS project_task_update_event;
DELIMITER $$
CREATE TRIGGER `project_task_update_event` AFTER UPDATE ON `apoll_web_projects_tasks` FOR EACH ROW
BEGIN
IF NOT (NEW.date_modified <=> OLD.date_modified)
THEN
INSERT INTO apoll_web_projects_events (event_type, project_id, task_id, created_by_user_id, description, date_created) VALUES ('7', NEW.project_id, NEW.task_id, NEW.modified_user_id, NEW.name, UTC_TIMESTAMP());
END IF;
END $$
DELIMITER ;
You can use affected_rows from your db class. (I am uncertain what class that is exactly). From the manual:
With ON DUPLICATE KEY UPDATE, the
affected-rows value per row is 1 if
the row is inserted as a new row and 2
if an existing row is updated.
Your code is insecure. You have SQL Injection problem.
To know if the record was created or updated, simple compare if create_date != update_date
When you insert data, set create_date and update_date to the same value. In "ON DUPLICATE" section update "update_date" field.
Another solution is to use triggers, something like this:
DELIMITER $$
CREATE TRIGGER `update_trigger` BEFORE UPDATE ON `tasks_table` FOR EACH ROW
BEGIN
/* Add any fields you want to compare here */
IF !(OLD.name <=> NEW.name OR OLD.description <=> NEW.description OR OLD.description <=> NEW.description OR OLD.status <=> NEW.status OR OLD.type <=> NEW.type OR OLD.priority <=> NEW.priority) THEN
NEW.date_modified = NOW()
END IF;
END;$$
DELIMITER ;
I call the procedure with php and the relevant variables. I need the latest IDs to use it for the next insert, so I set variables with SCOPE_IDENTITY. The return ist always the value of appointment_id ?!
ALTER proc [dbo].[insertPersonWithCmoFmo]
#appointment_id int,
#kostenstelle varchar(50),
#vorname varchar(50),
#nachname varchar(50),
#ci_nummer int,
#anzahl_monitore_old int,
#raum varchar(50),
#gebäude varchar(50),
#bemerkung text,
#hardware_typ varchar(50),
#anzahl_monitore_new varchar(50),
#zubehör text
as
DECLARE
#latestPersonID int,
#latestCmoID int,
#latestFmoID int
BEGIN
INSERT INTO [RC.Persons] (kostenstelle, vorname, nachname) VALUES (#kostenstelle, #vorname, #nachname);
SET #latestPersonID = (SELECT SCOPE_IDENTITY())
INSERT INTO [RC.CMO] (ci_nummer, anzahl_monitore, raum, gebäude, bemerkung) values (#ci_nummer, #anzahl_monitore_old, #raum, #gebäude, #bemerkung);
SET #latestCmoID = (SELECT SCOPE_IDENTITY())
INSERT INTO [RC.FMO] (hardware_typ, anzahl_monitore, zubehör) values (#hardware_typ, #anzahl_monitore_new, #zubehör);
SET #latestFmoID = (SELECT SCOPE_IDENTITY())
INSERT INTO [RC.Appointments_RC.CMO] (cmo_id, appointment_id) values (#latestCmoID, #appointment_id);
INSERT INTO [RC.Persons_RC.CMO] (cmo_id, person_id) VALUES (#latestCmoID, #latestPersonID);
INSERT INTO [RC.Persons_RC.FMO] (fmo_id, person_id) VALUES (#latestFmoID, #latestPersonID);
return #latestFmoID
END
This is the exec code. Why the is a "N" before all varchar type?
USE [Testtable]
GO
DECLARE #return_value int
EXEC #return_value = [dbo].[insertPersonWithCmoFmo]
#appointment_id = 52,
#kostenstelle = N'54',
#vorname = N'testname',
#nachname = N'testlastname',
#ci_nummer = 111222333,
#anzahl_monitore_old = 2,
#raum = N'255',
#gebäude = N'KWA12',
#bemerkung = N'blablabla',
#hardware_typ = N'Desktop',
#anzahl_monitore_new = N'4',
#zubehör = N'Test'
SELECT 'Return Value' = #return_value
GO
SQL Output:
Meldung 2601, Ebene 14, Status 1, Prozedur insertPersonWithCmoFmo, Zeile 36
Cannot insert duplicate key row in object 'dbo.RC.FMO' with unique index 'NonClusteredIndex-20140116-143317'. The duplicate key value is ().
The statement has been terminated.
Is there something about the error message you don't understand? One or more of the tables has a unique constraint (or index) and you are trying to insert the same values in the table. For example, the persons table might already have the person in it.
The N before the string explicitly makes the string use wide characters.
Your stored procedure probably needs to be rewritten. You need to check errors that might occur along the way. Traditionally, a value would be returned using an OUTPUT parameter.
The safest way to get the new id value is to use the output clause of the insert statement (see the documentation here).
The N indicates nvarchar instead of varchar.
See:
https://stackoverflow.com/questions/10025032/what-is-the-meaning-of-the-prefix-n-in-t-sql-statements
Working on some postgreSQL queries. As I said in a previous question..my knowledge of SQL logic is quite limited..
I have this query that inserts a row.
$timestamp = date('Y-m-d G:i:s.u');
$check_time = "start"; //can also be stop
$check_type = "start_user"; //can also be stop_user
$insert_query = "INSERT INTO production_order_process_log (
production_order_id,
production_order_process_id,
$check_time,
$check_type)
VALUES (
'$production_order_id',
'$production_order_process_id',
'$timestamp',
'$user')
";
Unfortunately, that is adding a new row every time. I would like to add conditional SQL so that
if the production_order_process_id doesn't exist, do the INSERT as it's written in the query above. That is, add the new row with all the new information
but if the production_order_process_id does exist and the check_type is stop_user then UPDATE the row to fill the column stop with the $timestamp and fill the column stop_user with $user.
I understand this is complicated.. Or, at least for me it is ^^ Thanks much for the help!
This is usually called MERGE or upsert. PostgreSQL doesn't have explicit support for this operation.
The best article I've seen on the topic of MERGE in PostgreSQL is this one by depesz .
After insert type ON DUPLICATE KEY UPDATE...
Use MERGE statement
Here is the usage
MERGE INTO table [[AS] alias]
USING [table-ref | query]
ON join-condition
[WHEN MATCHED [AND condition] THEN MergeUpdate | DELETE]
[WHEN NOT MATCHED [AND condition] THEN MergeInsert]
MergeUpdate is
UPDATE SET { column = { expression | DEFAULT } |
( column [, ...] ) = ( { expression | DEFAULT } [, ...] ) }
[, ...]
(yes, there is no WHERE clause here)
MergeInsert is
INSERT [ ( column [, ...] ) ]
{ DEFAULT VALUES | VALUES ( { expression | DEFAULT } [, ...] )
[, ...]}
(no subquery allowed)
I'm sure you'll find more articles/examples if you search for it.
It would be good if you can create a stored procedure and call while insert new record.
DELIMITER $$
DROP PROCEDURE IF EXISTS `DB`.`InsertNewRow` $$
CREATE PROCEDURE `db`.`InsertNewRow` ()
BEGIN
DECLARE V_EXIST INT DEFAULT 0;
DECLARE V_check_type VARCHAR(20);
SELECT production_order_process_id,check_type INTO V_EXIST,V_check_type FROM production_order_process_log;
IF V_EXIST=0 THEN
INSERT INTO production_order_process_log (
production_order_id,
production_order_process_id,
$check_time,
$check_type)
VALUES (
'$production_order_id',
'$production_order_process_id',
'$timestamp',
'$user');
ELSEIF V_check_type='stop_user' THEN
/* UPDATE QUERY HERE */
END IF;
END $$
DELIMITER ;
Just ad a WHERE CLAUSE to the insert:
INSERT INTO production_order_process_log
( production_order_id, production_order_process_id, check_time, check_type)
VALUES ( '$production_order_id', '$production_order_process_id', '$timestamp', '$user')
WHERE NOT EXISTS ( SELECT *
FROM production_order_process_log nx
--
-- assuming production_order_id is the Primary Key, here
--
WHERE nx.production_order_id = '$production_order_id'
);
UPDATE: I was confused by the parameters and the VALUE() . The fragment below works without parameters, but
with immediate values:
INSERT INTO tmp.production_order_process_log
( production_order_id, production_order_process_id, check_time, check_type)
SELECT 1, 2, '2012-07-19 12:12:12', 'Lutser'
WHERE NOT EXISTS ( SELECT *
FROM tmp.production_order_process_log nx
--
-- assuming production_order_id is the Primary Key, here
--
WHERE nx.production_order_id = 1
);
(you'll have to change it a bit to re-add the parameters)
I am looking to create a function that gets me a random item from a mySQL table, but let's me keep the returned as the "item of the day". In other words, the item that was "the item of the day" yesterday should not be shown again until all other items have been shown as item of the day.
Any suggestions on how to do this in an elegant way?
Thanks
Add a bool column "UsedAsItemOfTheDay" set to false (0). Update to true when item is picked. Exclude already used items from the picking process.
SELECT * FROM `table`
WHERE UsedAsItemOfTheDay = 0
ORDER BY RAND() LIMIT 1;
(Note: this is not the fastest way to return a random row in MySql; it will be slow on huge tables)
See also: quick selection of a random row from a large table in mysql
SELECT <fields> FROM <table> WHERE <some logic to exclude already used> ORDER BY RAND() LIMIT 1 will get you a random row from the table.
Add a column to store whether the item has been used:
ALTER TABLE your_table ADD COLUMN isused BOOL DEFAULT 0;
Get a random item of the day:
SELECT t.*
FROM your_table t
WHERE t.isused = 0
ORDER BY RAND()
LIMIT 1
Now update that record so it can't be used in the future:
UPDATE your_table
SET isused = 1
WHERE id = id_from_select_random_statement
People who "know" SQL will look for declarative solutions and will shun procedural code. Flagging rows is a "smell" for procedural code.
Is the set of Items static (never changes) or stable (rarely changes)? If yes, it would be easier to do a one-off exercise of generating a lookup table of values from now until the end of time, rather than scheduling a proc to running daily to look for unused flags and update the flag for today and clear all flags if all have been used etc.
Create a table of sequential dates between today and a far future date representing the lifetime of your application (you could consider omitting non-business days, of course). Add a column(s) referencing the key in you Items table (ensure you opt for ON DELETE NO ACTION referential action just in case those Items prove not to be static!) Then randomly assign the whole set of Items one per day until each has been used once. Repeat again for the whole set of Items until the table is full. You could easily generate this data using a spreadsheet and import it (or pure SQL if you are hardcore ;)
Quick example using Standard SQL:
Say there are only five Items in the set:
CREATE TABLE Items
(
item_ID INTEGER NOT NULL UNIQUE
);
INSERT INTO Items (item_ID)
VALUES (1),
(2),
(3),
(4),
(5);
You lookup table would be as simple as this:
CREATE TABLE ItemsOfTheDay
(
cal_date DATE NOT NULL UNIQUE,
item_ID INTEGER NOT NULL
REFERENCES Items (item_ID)
ON DELETE NO ACTION
ON UPDATE CASCADE
);
Starting with today, add the whole set of Items in random order:
INSERT INTO Items (item_ID)
VALUES ('2010-07-13', 2),
('2010-07-14', 4),
('2010-07-15', 5),
('2010-07-16', 1),
('2010-07-17', 3);
Then, starting with the most recent unfilled date, add the whole set of Items in (hopefully a different) random order:
INSERT INTO Items (item_ID)
VALUES ('2010-07-18', 1),
('2010-07-19', 3),
('2010-07-20', 4),
('2010-07-21', 5),
('2010-07-22', 2);
...and again...
INSERT INTO Items (item_ID)
VALUES ('2010-07-23', 2),
('2010-07-24', 3),
('2010-07-25', 5),
('2010-07-26', 1),
('2010-07-27', 4);
..and so on until the table is full.
Then it would then simply be a case of looking up today's date in the lookup table as and when required.
If the set of Items changes then the lookup table would obviously need to be regenerated, so you need to balance out the simplicity of design against the need for manual maintenance.
If you have fixed items you can add column
ALTER TABLE your_table ADD COLUMN item_day INT DEFAULT 0;
then selecting item use
WHERE item_day = DATE_FORMAT('%j')
If you get empty result then you can format new list of day items:
<?php
$qry = " UPDATE your_table SET item_day = 0";
$db->execute($qry);
// You only need 355 item to set as item of the day
for($i = 0; $i < 355; $i++) {
$qry = "UPDATE your_table SET item_day = ".($i+1)." WHERE item_day = 0 ORDER BY RAND() LIMIT 1";
$rs = $db->execute($qry);
// If no items left stop update
if (!$rs) { break; }
}
?>
Here's a stored procedure which selects a random row without using ORDER BY RAND(), and which resets the used flag once all items have been used:
DELIMITER //
DROP PROCEDURE IF EXISTS random_iotd//
CREATE PROCEDURE random_iotd()
BEGIN
# Reset used flag if all the rows have been used.
SELECT COUNT(*) INTO #used FROM iotd WHERE used = 1;
SELECT COUNT(*) INTO #rows FROM iotd;
IF (#used = #rows) THEN
UPDATE iotd SET used = 0;
END IF;
# Select a random number between 1 and the number of unused rows.
SELECT FLOOR(RAND() * (#rows - #used)) INTO #rand;
# Select the id of the row at position #rand.
PREPARE stmt FROM 'SELECT id INTO #id FROM iotd WHERE used = 0 LIMIT ?,1';
EXECUTE stmt USING #rand;
# Select the row where id = #id.
PREPARE stmt FROM 'SELECT id, item FROM iotd WHERE id = ?';
EXECUTE stmt USING #id;
# Update the row where id = #id.
PREPARE stmt FROM 'UPDATE iotd SET used = 1 WHERE id = ?';
EXECUTE stmt USING #id;
DEALLOCATE PREPARE stmt;
END;
//
DELIMITER ;
To use:
CALL random_iotd();
The procedure assumes a table structure like this:
CREATE TABLE `iotd` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`item` varchar(255) NOT NULL,
`used` BOOLEAN NOT NULL DEFAULT 0,
INDEX `used` (`used`),
PRIMARY KEY (`id`)
) ENGINE=InnoDB;
Here's one way to get the result from PHP (to keep things simple, error checking has been removed):
$mysqli = new mysqli('localhost', 'root', 'password', 'database');
$stmt = $mysqli->prepare('CALL random_iotd()');
$stmt->execute();
$stmt->bind_result($id, $item);
$stmt->fetch();
echo "$id, $item\n";
// 4, Item 4
UPADATE
This version should return the same result repeatedly on a given date. I've not really had time to test this, so be sure to do some testing of your own...
DELIMITER //
DROP PROCEDURE IF EXISTS random_iotd//
CREATE PROCEDURE random_iotd()
BEGIN
# Get today's item.
SET #id := NULL;
SELECT id INTO #id FROM iotd WHERE ts = CURRENT_DATE();
IF ISNULL(#id) THEN
# Reset used flag if all the rows have been used.
SELECT COUNT(*) INTO #used FROM iotd WHERE used = 1;
SELECT COUNT(*) INTO #rows FROM iotd;
IF (#used = #rows) THEN
UPDATE iotd SET used = 0;
END IF;
# Select a random number between 1 and the number of unused rows.
SELECT FLOOR(RAND() * (#rows - #used)) INTO #rand;
# Select the id of the row at position #rand.
PREPARE stmt FROM 'SELECT id INTO #id FROM iotd WHERE used = 0 LIMIT ?,1';
EXECUTE stmt USING #rand;
# Update the row where id = #id.
PREPARE stmt FROM 'UPDATE iotd SET used = 1, ts = CURRENT_DATE() WHERE id = ?';
EXECUTE stmt USING #id;
END IF;
# Select the row where id = #id.
PREPARE stmt FROM 'SELECT id, item FROM iotd WHERE id = ?';
EXECUTE stmt USING #id;
DEALLOCATE PREPARE stmt;
END;
//
DELIMITER ;
And the table structure:
CREATE TABLE `iotd` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`item` varchar(255) NOT NULL,
`used` BOOLEAN NOT NULL DEFAULT 0,
`ts` DATE DEFAULT 0,
INDEX `used` (`used`),
INDEX `ts` (`ts`),
PRIMARY KEY (`id`)
) ENGINE=InnoDB;
Why don't you use sequence?
Sequence serves your purpose easily...