It isn't possible to ALTER table ADD column foo timestamp default CURRENT_TIMESTAMP in sqlite, but are there any clever workarounds?
SQLite does not actually modify the table contents when adding a column, so the default must be a value that does not require reading the table.
To work around this, you could use a trigger instead of a default value:
Add the column without the default value:
ALTER TABLE MyTable ADD COLUMN foo timestamp DEFAULT NULL;
Use a trigger to set the default value:
ADD TRIGGER MyTable_foo_default
AFTER INSERT ON MyTable
FOR EACH ROW
WHEN NEW.foo IS NULL
BEGIN
UPDATE MyTable
SET foo = CURRENT_TIMESTAMP
WHERE rowid = NEW.rowid;
END;
Alternatively, modify the table contents first so that all rows have a value, then set the default value:
Add the column without the default value:
ALTER TABLE MyTable ADD COLUMN foo timestamp DEFAULT NULL /* replace me */;
Set the column value to something in all rows (the actual value does not matter, the important thing is that all rows now have the new column):
UPDATE MyTable SET foo = CURRENT_TIMESTAMP;
Change the default value (documentation: PRAGMA writable_schema, sqlite_master):
PRAGMA writable_schema = on;
UPDATE sqlite_master
SET sql = replace(sql, 'DEFAULT NULL /* replace me */',
'DEFAULT CURRENT_TIMESTAMP')
WHERE type = 'table'
AND name = 'MyTable';
PRAGMA writable_schema = off;
Reopen the database (otherwise, SQLite won't know about the new default value).
Related
My table has timestamp column. I want a trigger which sets timestamp to 0 on affected rows when a row is updated and the timestamp is not specified in the update statement.
If I use this trigger:
CREATE TRIGGER AFTER UPDATE ON mytable FOR EACH ROW
WHEN (NEW.timestamp IS NULL)
BEGIN
UPDATE mytable SET timestamp = 0 WHERE id = NEW.id;
END;
then the trigger doesn't fire for this update statement:
UPDATE mytable SET comecolumn='some'
I.e. timestamp of affected rows doesn't change to 0.
Can you please help me define the trigger?
The only way to make additional changes to a row in an UPDATE trigger is to execute another UPDATE on the same table afterwards.
The only way to detect whether a column value is changed is to compare the old and the new row values; the trigger does not know which columns actually were mentioned in the original UPDATE statement.
To prevent the trigger from triggering itself recursively, you should restrict it to be triggered by changes of all columns except the timestamp:
CREATE TRIGGER clear_timestamp
AFTER UPDATE OF all_the, other, columns ON MyTable
FOR EACH ROW
WHEN OLD.timestamp = NEW.timestamp
BEGIN
UPDATE MyTable
SET timestamp = 0
WHERE id = NEW.id;
END;
I think the problem is that in the SET statement is expanded to every column, with every column set to the current value in the database. So the original only trigger works, if the current timestamp column is NULL.
A solution could be to create another trigger that resets the timestamp column to NULL before an UPDATE.
CREATE TRIGGER "set_null"
BEFORE UPDATE ON "mytable" FOR EACH ROW
BEGIN
UPDATE mytable set timestamp = NULL where rowid = NEW.rowid;
END
This way the NEW.timestamp is NULL if it is not specified in the UPDATE SET.
Obviously now a NOT NULL constraint cannot be set on timestamp.
Another problem is that trigger recursion must be off when executing a update query:
PRAGMA recursive_triggers = OFF;
Here is another way:
import sqlite3
conn = sqlite3.connect(':memory:')
c = conn.cursor()
name = {'name':'jack'}
c.execute("""CREATE TABLE Programs (
id INTEGER PRIMARY KEY,
name VARCHAR(64) NOT NULL,
time_added INTEGER
);""")
c.execute("""CREATE TRIGGER program_time_added AFTER INSERT ON Programs
FOR EACH ROW
BEGIN
UPDATE Programs SET time_added =datetime('now', 'localtime') WHERE id = NEW.id;
END;""")
c.execute('INSERT INTO Programs (name) VALUES (?)', [name['name']])
I have an SQLite3 database that I would like to create. I want an INTEGER field (named "Length") to have a DEFAULT value that equals the length of the string in another field (named "Pattern").
CREATE TABLE knowledge (
Entry INTEGER PRIMARY KEY AUTOINCREMENT,
Priority TINYINT UNSIGNED CHECK (0 <= Priority < 15),
Pattern TEXT NOT NULL,
Length INTEGER UNSIGNED DEFAULT 'LENGTH(Pattern);'
);
However, the current table set-up does not "dynamically" set the value of "Length" as desired.
How can I properly set the DEFAULT value of "Length" to be the string length of the "Pattern" field?
The Default value you want to assign is dynamic which sqlite does not support. One solution is what CL. said. I would define the default value as 0 and use not one but two triggers (one for insert and another for update).
CREATE TRIGGER default_length_on_insert AFTER INSERT ON knowledge WHEN NEW.Length IS 0
BEGIN
UPDATE knowledge SET Length=length(NEW.Pattern) WHERE ROWID = NEW.ROWID;
END;
and
CREATE TRIGGER default_length_on_update AFTER UPDATE ON knowledge
BEGIN
UPDATE knowledge SET Length=length(NEW.Pattern) WHERE ROWID = NEW.ROWID;
END;
A default value must be a constant.
You coud use a trigger instead:
CREATE TRIGGER knowledge_length_default
AFTER INSERT ON knowledge
FOR EACH ROW
WHEN NEW.Length IS NULL
BEGIN
UPDATE knowledge
SET Length = length(NEW.Pattern)
WHERE Entry = NEW.Entry;
END;
I am working on SQLite. I want to create a table namely user_role with two column role_id and role_name. And there is an another table namely default that contain role_name and default_val.
I want to set Default value of role_id in the time of creating the table user_role and the Default value have to be retrieved from the table default.
I am new to SQLite and have to idea about the way to doing such recursive query. Please help.
The documentation says:
An explicit DEFAULT clause may specify that the default value is NULL, a string constant, a blob constant, a signed-number, or any constant expression enclosed in parentheses.
You would need to use a subquery, which is not allowed.
However, you could use a trigger that sets the ID if none was specified:
CREATE TRIGGER user_role_id_default
AFTER INSERT ON user_role
FOR EACH ROW
WHEN NEW.role_id IS NULL
BEGIN
UPDATE user_role
SET role_id = (SELECT default_val
FROM "default"
WHERE role_name = NEW.role_name)
WHERE rowid = NEW.rowid;
END;
Problem: a simplest possible update trigger writes a new value to all table rows instead of just the row being updated. Here is the table:
[names]
id INTEGER PRIMARY KEY
name TEXT
len INTEGER
Now I want to create triggers to update 'len' with the length of 'name'. This INSERT trigger seems to be doing the job corectly:
CREATE TRIGGER 'namelen' AFTER INSERT ON 'names'
BEGIN
UPDATE 'names' SET len = length(NEW.name) WHERE (id=NEW.id);
END;
Problems begin when I add a similar UPDATE trigger:
CREATE TRIGGER 'namelenupd' AFTER UPDATE ON 'names'
BEGIN
UPDATE 'names' SET len = length(NEW.name) WHERE (OLD.id=NEW.id);
END;
The update trigger writes the new length to all rows of the table, despite the WHERE clause. For example, if I say
UPDATE 'names' SET name='foo' where id=1;
then the value of 'len' becomes 3 for all rows of the table. I've looked at sqlite trigger examples and I can't see my error. What else must I do to make sure the trigger updates the 'len' column only in the row(s) that are actually updated?
Both OLD.xxx and NEW.xxx refer to the table row that caused the trigger to run.
The UPDATE statement inside the trigger runs independently; if you want to restrict it to one table row, you have to explicitly do this in its WHERE clause by filtering on that statement's table values, i.e., names.id or just id.
When the original UPDATE statement does not change the id column, the old and new id values are the same, and the expression OLD.id=NEW.id is true for all records in the table, as seen by the inner UPDATE statement.
The correct trigger looks like this:
CREATE TRIGGER "namelenupd"
AFTER UPDATE OF name ON "names"
BEGIN
UPDATE "names" SET len = length(NEW.name) WHERE id = NEW.id;
END;
Had the same issue, here's the syntax from my trigger
You would change "ALTER" to "CREATE" depending on what you already have (or not)
You have "id" as your primary key
Your dbo is "names"
Obviously, this will set the name value to "foo" (not really what you wanted). The key seems to be the last line, where you set inner join inserted on names.Id = inserted.Id.
USE [yourDBname]
ALTER TRIGGER [dbo].[yourTrigger]
ON [dbo].[names]
After INSERT, UPDATE
AS
BEGIN
-- SET NOCOUNT ON added to prevent extra result sets from
-- interfering with SELECT statements.
SET NOCOUNT ON;
Select id from inserted
begin
update [dbo].names
set [dbo].names.name = 'foo'
from dbo.names
inner join inserted
on names.id = inserted.id
END
I can't add a not null constraint or remove a default constraint. I would like to add a datetime column to a table and have all the values set to anything (perhaps 1970 or year 2000) but it seems like i cant use not null without a default and I cant remove a default once added in. So how can i add this column? (once again just a plain datetime not null)
Instead of using ALTER TABLE ADD COLUMN, create a new table that has the extra column, and copy your old data. This will free you from the restrictions of ALTER TABLE and let you have a NOT NULL constraint without a default value.
ALTER TABLE YourTable RENAME TO OldTable;
CREATE TABLE YourTable (/* old cols */, NewColumn DATETIME NOT NULL);
INSERT INTO YourTable SELECT *, '2000-01-01 00:00:00' FROM OldTable;
DROP TABLE OldTable;
Edit: The official SQLite documentation for ALTER TABLE now warns against the above procedure because it “might corrupt references to that table in triggers, views, and foreign key constraints.” The safe alternative is to use a temporary name for the new table, like this:
CREATE TABLE NewTable (/* old cols */, NewColumn DATETIME NOT NULL);
INSERT INTO NewTable SELECT *, '2000-01-01 00:00:00' FROM YourTable;
DROP TABLE YourTable;
ALTER TABLE NewTable RENAME TO YourTable;