I have an SQL UPDATE statement using the compound primary key (Key1,Key2) of a table that looks something like this:
UPDATE TableName SET FieldName = CASE
WHEN (Key1=389 AND Key2=5594091315209354374) THEN 1320243147187
WHEN (Key1=397 AND Key2=8686441440518828409) THEN 1320243147562
WHEN (Key1=389 AND Key2=5717973625907258381) THEN 1320243147182
....
WHEN (Key1=394 AND Key2=5512452777552926025) THEN 1320243147389 END
WHERE Key2 IN (123782199165241826,5594091315209354374,...,3553840348728167644)
AND Key1 IN (400,394,391,389,397);
I might have, say, 20 or so WHENs in the CASE statement.
How might MySQL say this has updated slightly more rows than there are WHENs?
Your CASE statement has nothing to do with which rows are affected. It's entirely up to your WHERE clause:
WHERE Key2 IN (123782199165241826,5594091315209354374,...,3553840348728167644)
AND Key1 IN (400,394,391,389,397);
Every row that fits those conditions will be updated. The question, then, is what happens to the rows that don't match a CASE condition?
For this part I'm not 100% sure what MySQL will do, as I'm more of a sql server guy. I suspect, however, that your CASE statement results in NULL, which is then assigned to FieldName. It's also possible that MySQL will decide not to change anything, but I would still expect it to report all the rows that match your WHERE clause as updated.
You should probably have an ELSE FieldName at the end of your case statement to be certain you get the latter behavior (no change) rather than the former (set to NULL).
I think the real determining factors are the values in your WHERE clause -- do these exactly correspond to the WHEN statements?
Related
I need assistance trying to find a problem where 1 record/row was updated but the entire table got updated with the same value. Need to understand how this could have happened and implement a method to prevent it from doing so again.
update TABLE set pushID='1234567890' where userID='111222333' ;
What would that update statement do if the userID value equaled nothing or equaled NULL? Could that cause the update statement to update every single row with the same pushID?
IE: update TABLE set pushID='1234567890' where userID='' ;
Could a blank userID value cause this? If not, what could cause this? If so, how could I write the query statement to prevent this from happening again?
What would that update statement do if the userID value equaled nothing or equaled NULL?
If the userID is NULL, then the condition becomes userID = NULL. This will always evaluate as false. In other words, no record will be updated.
If the userID is the empty string, then the condition becomes userID = ''. This will only update records where userID is equal to the empty string. I would expect that userID is the primary key of your table, so it would be suprising the find an empty value. And even if there is one, it will be unique, so a unique record will be updated.
As you see, none of the two above use case would generate a massive update of the table. The most probable option is that the query was triggered without an actual WHERE clause, like:
update TABLE set pushID='1234567890'
In case these are altogether integer ID, you should not convert them to string with ''. Doing so may lead to unexpected results (never tried, but it seems to be a possible cause for the comparison in the WHERE condition to fail). For example:
UPDATE TABLE SET pushID=1234567890 WHERE userID=111222333;
When all records are being updated, the query lacks the WHERE condition. Make sure to have posted the correct query, because this query should update nothing when the WHERE condition fails to match.
I run a wrong delete query in mysql terminal which i terminated quickly. however, i would like to know the likely outcome.
delete from table where 123
instead of
delete from table where id=123
The WHERE clause expects a boolean expression (something that is either TRUE or FALSE). If the expression is not boolean, it will be casted to a numeric value. Then any numeric value which is not equal to 0 (zero) is considered to be TRUE. Zero or NULL are equivalent to FALSE.
Since 123 is a non-zero numeric value, WHERE 123 is equivalent to WHERE 1 or WHERE TRUE (or even just without the WHERE clause), which in you case means: Delete all rows in the table.
Beside that: Some clients have settings, which don't permit delete statements like this, and expect the primary key to be used in the WHERE clause. Those settings are introduced to avoid exactly this kind of (typo) errors. To delete all rows, you would then need to write WHERE id=id.
It will delete all the rows of the table. MySQL treats all the integer from 1 to 9 as true and 0 as false. So now the where clause will give the boolean true result.
To see what query will delete, you can run same query but with select clause, like:
select * from table where 123
in this particular case it must select all rows without any filtering.
Are there any performance gain with this query:
UPDATE tbl SET field = 1 WHERE field != 1
over this
UPDATE tbl SET field = 1
Does the SQL parser already know that he doenst' need to update row that already are field = 1 ?
The condition field != 1 will likely make it faster especially in the case where most values are already 1. Assuming an index is available for optimizing, the database engine will be able to avoid examining most of the records in that case.
More importantly, perhaps, is that the queries may not have the same result. If any of the field values are NULL, the first UPDATE statement will not update those values. The second query would set the NULLs to 1.
Another (fairly obvious) case where they would not be equivalent is that UPDATE triggers would fire for all records for the second query (without the condition) but they would not fire for the first query for the skipped rows.
If there are just a few fields which are != 1, then there are most definitely performance gains in adding the WHERE clause. Even if MySQL did not write the value to disk, it still would need to look at every row to see if it needs to write it or not, but it certainly does not add the WHERE clause by itself.
I don't think this is possible as I couldn't find anything but I thought I would check on here in case I am not searching for the correct thing.
I have a settings table in my database which has two columns. The first column is the setting name and the second column is the value.
I need to update all of these at the same time. I wanted to see if there was a way to update these values at the same time one query like the following
UPDATE table SET col1='setting name' WHERE col2='1 value' AND SET col1='another name' WHERE col2='another value';
I know the above isn't a correct SQL format but this is the sort of thing that I would like to do so was wondering if there was another way that this can be done instead of having to perform separate SQL queries for each setting I want to update.
Thanks for your help.
You can use INSERT INTO .. ON DUPLICATE KEY UPDATE to update multiple rows with different values.
You do need a unique index (like a primary key) to make the "duplicate key"-part work
Example:
INSERT INTO table (a,b,c) VALUES (1,2,3),(4,5,6)
ON DUPLICATE KEY UPDATE b = VALUES(b), c = VALUES(c);
-- VALUES(x) points back to the value you gave for field x
-- so for b it is 2 and 5, for c it is 3 and 6 for rows 1 and 4 respectively (if you assume that a is your unique key field)
If you have a specific case I can give you the exact query.
UPDATE table
SET col2 =
CASE col1
WHEN 'setting1'
THEN 'value'
ELSE col2
END
, SET col1 = ...
...
I decided to use multiple queries all in one go. so the code would go like
UPDATE table SET col2='value1' WHERE col1='setting1';
UPDATE table SET col2='value2' WHERE col1='setting1';
etc
etc
I've just done a test where I insert 1500 records into the database. Do it without starting a DB transaction and it took 35 seconds, blanked the database and did it again but starting a transaction first, then once the 1500th record inserted finish the transaction and the time it took was 1 second, so definetely seems like doing it in a db transaction is the way to go.
You need to run separate SQL queries and make use of Transactions if you want to run as atomic.
UPDATE table SET col1=if(col2='1 value','setting name','another name') WHERE col2='1 value' OR col2='another value'
#Frits Van Campen,
The insert into .. on duplicate works for me.
I am doing this for years when I want to update more than thousand records from an excel import.
Only problem with this trick is, when there is no record to update, instead of ignoring, this method inserts a record and on some instances it is a problem. Then I need to insert another field, then after import I have to delete all the records that has been inserted instead of update.
I have this query:
UPDATE phonecalls
SET Called = "Yes"
WHERE PhoneNumber = "999 29-4655"
My table is phonecalls, I have a column named PhoneNumber. All I want to update is a column named Called to "yes".
Any idea what I am doing wrong? when I return my query it says 0 rows affected.
If the such value already exists, mysql won't change it and will therefore return "0 rows affected". So be sure to also check the current value of called
Another reason for 0 affected rows that I have observed: wrong data type. If the column you want to update is an integer or boolean, and you set it to a string, it won't be updated - but you will also get no error.
To sum up the other strategies/ideas from this post:
Check with a SELECT statement, whether your WHERE works and returns results.
Check whether your columns do already have the value you want to set.
Check if your desired value suits the data type of the column.
If the values are the same, MySQL will not update the row (without triggering any warning or error), so the affected row count will be 0.
The problem might be that there are no records with PhoneNumber == "999 29-4655".
Try this query:
SELECT * FROM phonecalls where PhoneNumber = '999 29-4655'
If it doesn't return anything, then there are no rows that match.
For the benefit of anyone here from Google, this problem was caused by me because I was trying to append to an empty field using CONCAT().
UPDATE example SET data=CONCAT(data, 'more');
If data is NULL, then CONCAT() returns NULL (ignoring the second parameter), so the value does not change (updating a NULL value to be a NULL value), hence the 0 rows updated.
In this case changing to the CONCAT_WS() function instead fixed the problem.
Try select count(*) from phonecalls where PhoneNumber = "999 29-4655"; That will give you the number of matching rows. If the result is 0, then there isn't a row in the database that matches.-
Check to make sure this returns some result.
SELECT * FROM phonecalls WHERE PhoneNumber = '999 29-4655'
If it doesn't return any result than the filter WHERE PhoneNumber = '999 29-4655' is not correct.
Does it say Rows matched: 1 Changed: 0 Warnings: 0? Then maybe it's already set to that value.
Did you try single quotes vs. double quotes?
"999 29-4655" is the space a space or a tab and is it consistent in your query and the database?
That's my sugestion:
UPDATE `phonecalls` SET `Called` = 'yeah!' WHERE `PhoneNumber` = '999 29-4655' AND `Called` != 'yeah!'
And make sure with the case-sensitive name of table and field`s.
Just ran into an obscure case of this. Our code reads a list of records from the database, changes a column, and writes them back one by one. The UPDATE's WHERE clause contains only two conditions: WHERE key=? AND last_update_dt=?. (The timestamp check is for optimistic locking: if the record is changed by another process before we write ours, 0 rows are updated and we throw an error.)
But for one particular row the UPDATE was failing- zero rows effected.
After much hair-pulling I noticed that the timestamp for the row was 2019-03-10 02:59. In much of the U.S. that timestamp wouldn't exist- Daylight Savings Time causes the time to skip directly from 2:00 to 3:00. So I guessed that during the round trip from MySQL to Java back to MySQL, some part of the code was interpreting that timestamp differently from the rest, making the timestamps in the WHERE clause not match.
Changing the row's timestamp by one hour avoided the problem.
(Of course, the correct fix is to abolish Daylight Savings Time. I created a Jira but the U.S. Government has not responded to it yet.)
In my case, I was trying to update a column of text to correct a truncation problem with it. Trying to update to the correct text was yielding 0 rows updated because the text in the row wasn't changing.
Once I extended the column in the table structure to accommodate for the correct number of characters, I was able to see the desired results.