UPDATE(7)

📦 version 18.6, in opensuse-leap 16.0

✨ UPDATE(7) ✨

okay bestie, this is the UPDATE page from the PostgreSQL 18.6 Documentation and it's genuinely one of the most-used commands in the whole game, so let's get into it 💅

🌸 NAME 🌸

UPDATE — update rows of a table

🌈 SYNOPSIS 🌈

[ WITH [ RECURSIVE ] with_query [, ...] ]
UPDATE [ ONLY ] table_name [ * ] [ [ AS ] alias ]
    SET { column_name = { expression | DEFAULT } |
          ( column_name [, ...] ) = [ ROW ] ( { expression | DEFAULT } [, ...] ) |
          ( column_name [, ...] ) = ( sub-SELECT )
        } [, ...]
    [ FROM from_item [, ...] ]
    [ WHERE condition | WHERE CURRENT OF cursor_name ]
    [ RETURNING [ WITH ( { OLD | NEW } AS output_alias [, ...] ) ]
                { * | output_expression [ [ AS ] output_name ] } [, ...] ]

💖 DESCRIPTION 💖

UPDATE changes the values of the specified columns in all rows that satisfy the condition. And listen — you only have to mention the columns you actually want to change in the SET clause. Every column you don't explicitly modify keeps its previous value. She's respectful like that 🫶

So there's two ways to modify a table using info that lives in other tables in the database: either you use sub-selects, or you specify extra tables in the FROM clause. Which technique is more appropriate depends on the specific circumstances. No cap, it's a vibe-by-vibe situation ✨

The optional RETURNING clause makes UPDATE compute and return value(s) based on each row that was actually updated. You can compute any expression using the table's columns, and/or columns of other tables mentioned in FROM. By default the new (post-update) values of the table's columns get used, but you can also request the old (pre-update) values if that's your thing. The syntax of the RETURNING list is identical to that of the output list of SELECT. iconic behavior honestly 👑

Permissions check, bestie — you must have the UPDATE privilege on the table, or at least on the column(s) that are listed to be updated. You must also have the SELECT privilege on any column whose values are read in the expressions or condition. Both. No skipping 🙈

🎀 PARAMETERS 🎀

with_query

The WITH clause lets you specify one or more subqueries that can be referenced by name in the UPDATE query. See Section 7.8 and SELECT(7) for details. iykyk 📚

table_name

The name (optionally schema-qualified) of the table to update. If ONLY is specified before the table name, matching rows are updated in the named table only. If ONLY is not specified, matching rows are also updated in any tables inheriting from the named table. Optionally, * can be specified after the table name to explicitly indicate that descendant tables are included. yes the asterisk is doing something, it's not just decor 💅

alias

A substitute name for the target table. And the way that when an alias is provided, it completely hides the actual name of the table. Like, for example, given UPDATE foo AS f, the remainder of the UPDATE statement must refer to this table as f not foo. She's committed to the bit, fr fr.

column_name

The name of a column in the table named by table_name. The column name can be qualified with a subfield name or array subscript, if needed. But do not include the table's name in the specification of a target column — for example, UPDATE table_name SET table_name.col = 1 is invalid. caught in 4k 😭

expression

An expression to assign to the column. The expression can use the old values of this and other columns in the table. so yes you can reference what was there before, go off 🌸

DEFAULT

Set the column to its default value (which will be NULL if no specific default expression has been assigned to it). An identity column will be set to a new value generated by the associated sequence. For a generated column, specifying this is permitted but merely specifies the normal behavior of computing the column from its generation expression. no surprises, just the standard lore ⭐

sub-SELECT

A SELECT sub-query that produces as many output columns as are listed in the parenthesized column list preceding it. The sub-query must yield no more than one row when executed. If it yields one row, its column values are assigned to the target columns; if it yields no rows, NULL values are assigned to the target columns. The sub-query can refer to old values of the current row of the table being updated. one row only bestie, she's monogamous 💀

from_item

A table expression allowing columns from other tables to appear in the WHERE condition and update expressions. This uses the same syntax as the FROM clause of a SELECT statement; for example, an alias for the table name can be specified. Do not repeat the target table as a from_item unless you intend a self-join (in which case it must appear with an alias in the from_item). lowkey a classic oopsie 🫠

condition

An expression that returns a value of type boolean. Only rows for which this expression returns true will be updated. simple as that ✨

cursor_name

The name of the cursor to use in a WHERE CURRENT OF condition. The row to be updated is the one most recently fetched from this cursor. The cursor must be a non-grouping query on the UPDATE's target table. Note that WHERE CURRENT OF cannot be specified together with a Boolean condition. See DECLARE(7) for more information about using cursors with WHERE CURRENT OF. 🌈

output_alias

An optional substitute name for OLD or NEW rows in the RETURNING list.

By default, old values from the target table can be returned by writing OLD.column_name or OLD.*, and new values can be returned by writing NEW.column_name or NEW.*. When an alias is provided, these names are hidden and the old or new rows must be referred to using the alias. For example RETURNING WITH (OLD AS o, NEW AS n) o.*, n.*. that's the glow-up right there 💅

output_expression

An expression to be computed and returned by the UPDATE command after each row is updated. The expression can use any column names of the table named by table_name or table(s) listed in FROM. Write * to return all columns. give me everything, no crumbs 🍓

A column name or * may be qualified using OLD or NEW, or the corresponding output_alias for OLD or NEW, to cause old or new values to be returned. An unqualified column name, or *, or a column name or * qualified using the target table name or alias will return new values. so unqualified = new by default, periodt.

output_name

A name to use for a returned column.

🔥 OUTPUTS 🔥

On successful completion, an UPDATE command returns a command tag of the form

UPDATE count

The count is the number of rows updated, including matched rows whose values did not change. Note that the number may be less than the number of rows that matched the condition when updates were suppressed by a BEFORE UPDATE trigger. If count is 0, no rows were updated by the query (this is not considered an error, she's just built different). 💖

If the UPDATE command contains a RETURNING clause, the result will be similar to that of a SELECT statement containing the columns and values defined in the RETURNING list, computed over the row(s) updated by the command. 🦋

🦋 NOTES 🦋

When a FROM clause is present, what essentially happens is that the target table is joined to the tables mentioned in the from_item list, and each output row of the join represents an update operation for the target table. When using FROM you should ensure that the join produces at most one output row for each row to be modified. In other words, a target row shouldn't join to more than one row from the other table(s). If it does, then only one of the join rows will be used to update the target row, but which one will be used is not readily predictable. it's giving chaos, and not the cute kind 😭

Because of this indeterminacy, referencing other tables only within sub-selects is safer, though often harder to read and slower than using a join. tradeoffs, bestie 🎀

In the case of a partitioned table, updating a row might cause it to no longer satisfy the partition constraint of the containing partition. In that case, if there is some other partition in the partition tree for which this row satisfies its partition constraint, then the row is moved to that partition. If there is no such partition, an error will occur. Behind the scenes, the row movement is actually a DELETE and INSERT operation. sneaky little glow-up ✨

There is a possibility that a concurrent UPDATE or DELETE on the row being moved will get a serialization failure error. Suppose session 1 is performing an UPDATE on a partition key, and meanwhile a concurrent session 2 for which this row is visible performs an UPDATE or DELETE operation on this row. In such case, session 2's UPDATE or DELETE will detect the row movement and raise a serialization failure error (which always returns with an SQLSTATE code '40001'). Applications may wish to retry the transaction if this occurs. In the usual case where the table is not partitioned, or where there is no row movement, session 2 would have identified the newly updated row and carried out the UPDATE/DELETE on this new row version.

Note that while rows can be moved from local partitions to a foreign-table partition (provided the foreign data wrapper supports tuple routing), they cannot be moved from a foreign-table partition to another partition. one-way street, no cap 🚧

An attempt of moving a row from one partition to another will fail if a foreign key is found to directly reference an ancestor of the source partition that is not the same as the ancestor that's mentioned in the UPDATE query. 💀

✨ EXAMPLES ✨

Okay bestie, real talk: this is the part where it all clicks. Grab a snack 🍓, we're doing this.

Wanna swap a word out? Say you want the word Drama to become Dramatic in the kind column of the table films:

UPDATE films SET kind = 'Dramatic' WHERE kind = 'Drama';

Now let's mess with the temperature entries and reset prcp back to its default value, all in one single row of the table weather:

UPDATE weather SET temp_lo = temp_lo+1, temp_hi = temp_lo+15, prcp = DEFAULT
  WHERE city = 'San Francisco' AND date = '2003-07-03';

Same operation, but this time we want the updated entries back, plus the old precipitation value:

UPDATE weather SET temp_lo = temp_lo+1, temp_hi = temp_lo+15, prcp = DEFAULT
  WHERE city = 'San Francisco' AND date = '2003-07-03'
  RETURNING temp_lo, temp_hi, prcp, old.prcp AS old_prcp;

Feeling fancy? Use the alternative column-list syntax to do that exact same update:

UPDATE weather SET (temp_lo, temp_hi, prcp) = (temp_lo+1, temp_lo+15, DEFAULT)
  WHERE city = 'San Francisco' AND date = '2003-07-03';

Now let's bump up the sales count of the salesperson who manages the account for Acme Corporation, using that FROM clause syntax:

UPDATE employees SET sales_count = sales_count + 1 FROM accounts
  WHERE accounts.name = 'Acme Corporation'
  AND employees.id = accounts.sales_person;

Same thing again, but this time with a sub-select chilling in the WHERE clause:

UPDATE employees SET sales_count = sales_count + 1 WHERE id =
  (SELECT sales_person FROM accounts WHERE name = 'Acme Corporation');

Wanna make the contact names in an accounts table match whoever's actually assigned as the salesperson rn? Say less:

UPDATE accounts SET (contact_first_name, contact_last_name) =
    (SELECT first_name, last_name FROM employees
     WHERE employees.id = accounts.sales_person);

You could get a similar result with a join instead:

UPDATE accounts SET contact_first_name = first_name,
                    contact_last_name = last_name
  FROM employees WHERE employees.id = accounts.sales_person;
But heads up bestie, the second query can give you unexpected results if employees.id is not a unique key, whereas the first query is guaranteed to raise an error if there are multiple id matches. Also, if there's no match for a particular accounts.sales_person entry, the first query will set the corresponding name fields to NULL, whereas the second query won't update that row at all. 💀

Updating statistics in a summary table so they match the current data:

UPDATE summary s SET (sum_x, sum_y, avg_x, avg_y) =
    (SELECT sum(x), sum(y), avg(x), avg(y) FROM data d
     WHERE d.group_id = s.group_id);

Here's a cute trick 🎀 — try to insert a new stock item along with the quantity of stock, but if the item already exists, update the stock count of the existing item instead. To do that without failing the entire transaction, use savepoints:

BEGIN;
-- other operations
SAVEPOINT sp1;
INSERT INTO wines VALUES('Chateau Lafite 2003', '24');
-- Assume the above fails because of a unique key violation,
-- so now we issue these commands:
ROLLBACK TO sp1;
UPDATE wines SET stock = stock + 24 WHERE winename = 'Chateau Lafite 2003';
-- continue with other operations, and eventually
COMMIT;

Wanna change the kind column of the table films in the row where the cursor c_films is currently positioned? Yup, WHERE CURRENT OF is a whole vibe:

UPDATE films SET kind = 'Dramatic' WHERE CURRENT OF c_films;

Okay so this next bit is important, no cap. 🫶 Updates that affect many rows can have negative effects on system performance, such as table bloat, increased replica lag, and increased lock contention. In those situations it can make sense to do the operation in smaller batches, possibly with a VACUUM operation on the table between batches. While there's no LIMIT clause for UPDATE, you can get a similar effect through the use of a Common Table Expression and a self-join. With the standard PostgreSQL table access method, a self-join on the system column ctid is very efficient:

WITH exceeded_max_retries AS (
  SELECT w.ctid FROM work_item AS w
    WHERE w.status = 'active' AND w.num_retries > 10
    ORDER BY w.retry_timestamp
    FOR UPDATE
    LIMIT 5000
)
UPDATE work_item SET status = 'failed'
  FROM exceeded_max_retries AS emr
  WHERE work_item.ctid = emr.ctid;

This command will need to be repeated until no rows remain to be updated. (This use of ctid is only safe because the query is repeatedly run, avoiding the problem of changed ctids.) Use of an ORDER BY clause allows the command to prioritize which rows will be updated; it can also prevent deadlock with other update operations if they use the same ordering. If lock contention is a concern, then SKIP LOCKED can be added to the CTE to prevent multiple commands from updating the same row. However, then a final UPDATE without SKIP LOCKED or LIMIT will be needed to ensure that no matching rows were overlooked. 👑

🌈 COMPATIBILITY 🌈

Good news: this command conforms to the SQL standard! 🎉 With like, two lil caveats — the FROM and RETURNING clauses are PostgreSQL extensions, and so is the ability to use WITH with UPDATE.

Some other database systems offer a FROM option in which the target table is supposed to be listed again within FROM. That is not how PostgreSQL interprets FROM. So be careful when porting applications that use this extension. 🙈

According to the standard, the source value for a parenthesized sub-list of target column names can be any row-valued expression yielding the correct number of columns. PostgreSQL only allows the source value to be a row constructor or a sub-SELECT. An individual column's updated value can be specified as DEFAULT in the row-constructor case, but not inside a sub-SELECT.

And that's the whole page, bestie!! 💅 You just read an entire manpage and honestly? Iconic. Main character energy fr fr.

This page is from PostgreSQL 18.6, dated 2026, and it lives at UPDATE(7) in the manual. Go forth and update responsibly 🦋✨

📦 where this page comes from