SQLite / MySQL compatibility issue - mysql

I know SQL in SQLite is not completely implemented the same way as in MySql. My problem with the following queries is, that they are not compatible and I like to avoid a conditional if <DBMS> ... else
SQLite query
UPDATE sorties SET state = '#'
WHERE `key` IN (
SELECT `key` FROM sorties
INNER JOIN reports AS r
ON r.sortieId=sorties.`key`);
Error on MySQL:
SQL Error (1093): Table 'sorties' is specified twice, both as a target for 'UPDATE' and as a separate source for data
MySQL query (adapted from here)
UPDATE sorties AS s SET s.state='#'
WHERE s.`key` IN (
SELECT t.sortieId FROM (
SELECT r.sortieId AS sortieId
FROM reports AS r
INNER JOIN sorties AS sort
ON sort.`key`=r.sortieId)
AS t);
Error on SQLite:
SQLiteManager: Likely SQL syntax error: UPDATE sorties AS s SET s.state='#'
WHERE s.key IN ( SELECT t.sortieId FROM (
SELECT DISTINCT r.sortieId AS sortieId
FROM reports AS r
INNER JOIN sorties AS sort
ON sort.key=r.sortieId) AS t); [ near "AS": syntax error ]
Exception Name: NS_ERROR_FAILURE
Exception Message: Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [mozIStorageConnection.createStatement]
I can't figure out how to make this queries work on both systems equally!
All I want to have is, that each state of sorties must be '#' when it's key can be found in reports.sortieId.
Maybe there is a different approach for this?
Thank you

The first command reads the key value from the sorties table in the subquery, and then checks whether those key values exist in the sorties table in the outer statement. That check is superfluous; you can just compare the values to the ones in reports directly:
UPDATE sorties
SET state = '#'
WHERE key IN (SELECT sortieId
FROM reports);
As for the second command, SQLite does not support aliasing a table used in INSERT/UPDATE/DELETE because those commands work only on a single table. You can just remove the AS s and replace s with sorties everywhere.

Related

How to resolve ambiguous error in the query

I am using laravel framework for developing API's ,i have one query that is executed without where condition without any error i need to execute with where condition but it's throwing an error
query
select count(*) as aggregate
from `users`
left join `books` on `books`.`book_id` = `books`.`id`
where `access_id` = 5054
SQL Error [1052] [23000]: Column 'access_id' in where clause is
ambiguous
after searching google i got something we have to specify reference of a table name , i added reference name like this where users.access_id =5054 but it's throwing an error like unknown column but in my db i have that column in both users and books table
The problem is its consider as a column so that's why syntax error is coming,try following way it will resolve your problem
select count(*) as aggregate
from `users`
left join `books` on `books`.`book_id` = `books`.`id`
where `users`.`access_id` = 5054

DELETE FROM throws a SQL Error [1064] [42000]: (conn=5159) on a query, on which a SELECT works

Trying to delete a specific amount of rows in a MySQL query, I am able to SELECT whatever I want to delete with the following command, getting the results I need:
select * from ns_cos ns where ns.created_at <>
(select max(nsa.created_at) from ns_cos nsa
where nsa.month_year = ns.month_year)
However, when I try to delete the selected data with:
delete from ns_cos ns where ns.created_at not exists
(select max(nsa.created_at) from ns_cos nsa
where nsa.month_year = ns.month_year)
I get:
SQL Error [1064] [42000]: (conn=5159) You have an error in your SQL syntax; check the manual that corresponds to your MariaDB server version for the right syntax to use near 'ns where ns.created_at not exists (select max(nsa.created_at) from ns_cos nsa wh' at line 1
What am I doing wrong?
Your immediate issue is that not all MySQL versions support aliasing the table directly in delete from. Furthermore, though, you cannot re-open the table you delete from in the from clause.
Consider using the delete ... join syntax.
delete ns
from ns_cos ns
inner join (
select month_year, max(nsa.created_at) created_at
from ns_cos nsa
group by month_year
) ns1 on ns1.month_year = ns.month_year and ns1.created_at <> ns.created_at
EXISTS in there not posible use IN clause, but you need to enclose the table in a seprate select, so that mysql thinks it is another table
delete from ns_cos ns
where ns.created_at not IN (select max(nsa.created_at) from (SELECT * FROM ns_cos) nsa where nsa.month_year = ns.month_year)
This happens when you join tables from two schemas, no matter whether you take aliases for the table names or not.
Error
SQL Error [1064] [42000]: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near '...'
delete !ALIAS! from table ALIAS ...
Strangely, only with aliases, you can get around this error in a one-step SQL. Tested in MySQL 5.7.
Try this pattern, using the alias of the table you want to delete from between delete and from:
delete t1 from table t1
join table2 t2
on t1.id = t2.id
Your code would be:
delete ns from ns_cos ns where ns.created_at not exists
(select max(nsa.created_at) from ns_cos nsa
where nsa.month_year = ns.month_year)
Other steps I checked before finding out the alias trick (do not read)
I tried it with a view of the other schema's table in the same schema, that is not enough.
One way to get around this is to make a copy of the table in the same schema (and delete that copy afterwards).
You might also somehow make a full "linked server link" like in T-SQL [linkedservername].DB1.Schema.Table1 at How to join two tables if they are in different schemas Your code is on the same server, but this might still help jumping over to the other schema, untested.

Django QuerySet.update() fails on MySQL if a subquery references the base table

I am trying to make this transaction mysql compatible. As mysql don't allow the current operation. MySQL doesn't allow updating the table you are already using in an inner select as the update criteria.
Getting this error
django.db.utils.OperationalError: (1093, "You can't specify target table 'catalogue_category' for update in FROM clause")
Code:
included_in_non_public_subtree = self.__class__.objects.filter(
is_public=False, path__rstartswith=OuterRef("path"), depth__lt=OuterRef("depth")
)
self.get_descendants_and_self().update(
ancestors_are_public=Exists(
included_in_non_public_subtree.values("id"), negated=True)
)
https://github.com/django-oscar/django-oscar/pull/3050#pullrequestreview-461576714
You can do subquery by yourself. Split one query into multiple queries.
e.g.
included_in_non_public_subtree = not self.__class__.objects.filter(
is_public=False, path__rstartswith=OuterRef("path"), depth__lt=OuterRef("depth")
).exists()
self.get_descendants_and_self().update(
ancestors_are_public=included_in_non_public_subtree,
)

How do I write an AREL UpdateManager query for MySQL that uses a subquery

I am tying to run an update query with a subquery against a MySQL database using ruby. I am using ruby 1.9.3 and rails 4.1.
The query I am trying to create is as below:
UPDATE `items`
SET
`items`.`status_id` = 12
WHERE
`items`.`id` IN (SELECT DISTINCT
`items`.`id`
FROM
`items`
LEFT OUTER JOIN
`statuses` ON `items`.`status_id` = `statuses`.`id`
LEFT OUTER JOIN
`resources` ON `items`.`resource_id` = `resources`.`id`
WHERE
`statuses`.`title` LIKE 'On Loan'
AND `items`.`duedate` < '2015-04-24'
AND `items`.`return_date` IS NULL
ORDER BY `items`.`duedate`)
I can produce this query in ruby using AREL with the code shown below:
# Declare Arel objects
i = Item.arel_table
s = Status.arel_table
r = Resource.arel_table
# This is the AREL query that returns the data
overdues = i.project(i[:id]).
join(s, Arel::Nodes::OuterJoin).on(i[:status_id].eq(s[:id])).
join(r, Arel::Nodes::OuterJoin).on(i[:resource_id].eq(r[:id])).
where(s[:title].matches("On Loan").
and(i[:duedate].lt(DateTime.now.to_date)).
and(i[:return_date].eq(nil))
).
order(i[:duedate])
# Note: You can't chain distinct, otherwise "overdues" becomes a string with the value "DISTINCT".
overdues.distinct
# This creates the update...
u = Arel::UpdateManager.new i.engine
u.table(i)
u.set([[i[:status_id], 10]]).where(i[:id].in(overdues))
This does not work and returns an error message:
ActiveRecord::StatementInvalid: Mysql2::Error: You can't specify target table 'items' for update in FROM clause:
I tried using AR "update_all" but it produces the same SQL and hence the same error.
Item.where(i[:id].in(overdues)).update_all(:status_id => (Status.find_by(:title => "Overdue").id))
Having done some research I have found that you cannot run a update with a subquery that references the table you want to update in MySQL. I have seen a number of posts on this site and the wider internet that detail work arounds.
One suggestion says that the update should use a join instead of a sub query. Having looked at the code behind the update manager it has no "join" so I can't do that.
Another says run this in two parts but I can't see how to because AREL and AciveRecord both chain actions.
The only way I can see of doing this is by aliasing the table and adding an additional select (see below). This isn't great but it would be useful to see if it is possible to do.
UPDATE `items`
SET `status_id` = 10
WHERE `items`.`id` IN (
SELECT x.id
FROM
(SELECT DISTINCT `items`.`id`
FROM `items`
LEFT OUTER JOIN `statuses` ON `items`.`status_id` = `statuses`.`id`
LEFT OUTER JOIN `resources` ON `items`.`resource_id` = `resources`.`id`
WHERE `statuses`.`title` LIKE 'On Loan'
AND `items`.`duedate` < '2015-04-24'
AND `items`.`return_date` IS NULL
ORDER BY `items`.`duedate`) x
);
If I can't get this to work I could adopt two other approaches:
1) I could just hard-code the SQL but I want to use ActiveRecord and reference the models to keep it database agnostic.
2) The other way is to return an instance of all the records and loop through them doing individual updates. This will have a performance issue but I can accept this because its a background job that won't be updating more than a handful of records each day.
Update
I have the AREL query below that produces the subquery in the format I need.
x = Arel::Table.new('x')
overdues = Item.select(x[:id]).from(
Item.select(Item.arel_table[:id]).where(
Status.arel_table[:title].matches("On Loan").and(
Item.arel_table[:duedate].lt(DateTime.now.to_date).and(
Item.arel_table[:return_date].eq(nil))
)
).joins(
Item.arel_table.join(Status.arel_table, Arel::Nodes::OuterJoin).on(
Item.arel_table[:status_id].eq(Status.arel_table[:id])
).join_sources
).joins(
Item.arel_table.join(Resource.arel_table, Arel::Nodes::OuterJoin).on(
Item.arel_table[:resource_id].eq(Resource.arel_table[:id])
).join_sources
).order(Item.arel_table[:duedate]).uniq.as('x')
)
Sadly it returns an error when I use it in my update statement.
TypeError: Cannot visit Item::ActiveRecord_Relation
Having revisited this question I am at the conclusion that it's not possible to do this because of a limitation with MySQL:
ActiveRecord::StatementInvalid: Mysql2::Error: You can't specify target table 'items' for update in FROM clause:
It should be possible to do with other databases (although I haven't tested that).
I could create a temporary table, which is the copy of the original table, reference that and then drop the temporary table like this post suggests:
http://richtextblog.blogspot.co.uk/2007/09/mysql-temporary-tables-and-rails.html. That seems a lot of overhead to do a simple subquery.
What I am going to do is find all the ID's and loop through them and update the records that way (using a simple find and update). This has an overhead but it should only be updating a handful of records each run (no more than 100). The update will be running as a scheduled job outside user working hours so it won't impact performance.
I still find it bizarre that in all other flavours of SQL I have never encountered this problem before. Still you live and learn.
UPDATE:
Since updating my version of MySQL the select statement now works. I had to take out the order by for it to work.
ORDER BY `items`.`duedate`
I am now using version: 5.7.19.

MySQL Syntax error message "Operand should contain 1 column(s)"

I tried running the following statement:
INSERT INTO VOUCHER (VOUCHER_NUMBER, BOOK_ID, DENOMINATION)
SELECT (a.number, b.ID, b.DENOMINATION)
FROM temp_cheques a, BOOK b
WHERE a.number BETWEEN b.START_NUMBER AND b.START_NUMBER+b.UNITS-1;
which, as I understand it, should insert into VOUCHER each record from temp_cheques with the ID and DENOMINATION fields corresponding to entries in the BOOK table (temp_cheques comes from a database backup, which I'm trying to recreate in a different format). However, when I run it, I get an error:
Error: Operand should contain 1 column(s)
SQLState: 21000
ErrorCode: 1241
I'm running this in SQuirrel and have not had issues with any other queries. Is there something wrong with the syntax of my query?
EDIT:
The structure of BOOK is:
ID int(11)
START_NUMBER int(11)
UNITS int(11)
DENOMINATION double(5,2)
The structure of temp_cheques is:
ID int(11)
number varchar(20)
Try removing the parenthesis from the SELECT clause. From Microsoft TechNet, the correct syntax for an INSERT statement using a SELECT clause is the following.
INSERT INTO MyTable (PriKey, Description)
SELECT ForeignKey, Description
FROM SomeView
The error you're getting, "The SELECT would examine more than MAX_JOIN_SIZE rows; check your WHERE and use SET SQL_BIG_SELECTS=1 or SET SQL_MAX_JOIN_SIZE=# if the SELECT is okay.", is actually correct, assuming you have many rows in both BOOK and temp_cheques. You are trying to query all rows from both tables and make a cross-reference, resulting in an m*n size query. SQL Server is trying to warn you of this, before performing a potentially long operation.
Set SQL_BIG_SELECTS = 1 before running this statement, and try again. It should work, but note that this operation may take a long time.
Does B contain the UNITS column?
What is the table structure for temp_cheques and Book?
EDIT: As I said in comments, all the columns should be numeric when doing +/- and when comparing. Does the following simple SELECT work?
SELECT b.START_NUMBER+b.UNITS-1 FROM Books B
I don't have a MySQL instance handy, but my first guess is the WHERE clause:
WHERE a.number BETWEEN b.START_NUMBER AND b.START_NUMBER+b.UNITS-1;
I imagine that the MySQL parser may be interpreting that as:
WHERE number
(BETWEEN start_number AND start_number) + units - 1
Try wrapping everything in parentheses, ie:
WHERE a.number BETWEEN b.START_NUMBER AND (b.START_NUMBER + b.UNITS - 1);
The final version of the query is as follows:
Set SQL_BIG_SELECTS = 1;
INSERT INTO VOUCHER (VOUCHER_NUMBER, BOOK_ID, DENOMINATION)
SELECT a.number, b.ID, b.DENOMINATION
FROM temp_cheques a, BOOK b
WHERE a.number BETWEEN b.START_NUMBER AND (b.START_NUMBER+b.UNITS-1);
The parsing of the BETWEEN statement required parentheses, the SELECT did not, and because of the size of the two tables (215000 records in temp_cheques, 8000 in BOOK) I was breaking a limit on the select size, requiring me to set SQL_BIG_SELECTS = 1.
I ran into the same error when using Spring Repositories.
My repository contained a method like:
List<SomeEntity> findAllBySomeId(List<String> ids);
This is working fine when running integration tests against an in-memory database (h2). However against a stand alone database like MySql is was failing with the same error.
I've solved it by changing the method interface to:
List<someEntity findBySomeIdIn(List<String> ids);
Note: there is no difference between find and findAll. As described here: Spring Data JPA difference between findBy / findAllBy