Simulate CREATE DATABASE IF NOT EXISTS for PostgreSQL?

Question 1

Restrictions

You can ask the system catalog pg_database - accessible from any database in the same database cluster. The tricky part is that CREATE DATABASE can only be executed as a single statement. The manual:

CREATE DATABASE cannot be executed inside a transaction block.

So it cannot be run directly inside a function or DO statement, where it would be inside a transaction block implicitly. SQL procedures, introduced with Postgres 11, cannot help with this either.

Workaround from within psql

You can work around it from within psql by executing the DDL statement conditionally:

SELECT 'CREATE DATABASE mydb'
WHERE NOT EXISTS (SELECT FROM pg_database WHERE datname = 'mydb')\gexec

The manual:

\gexec

Sends the current query buffer to the server, then treats each column of each row of the query's output (if any) as a SQL statement to be executed.

Workaround from the shell

With \gexec you only need to call psql once:

echo "SELECT 'CREATE DATABASE mydb' WHERE NOT EXISTS (SELECT FROM pg_database WHERE datname = 'mydb')\gexec" | psql

You may need more psql options for your connection; role, port, password, ... See:

Run batch file with psql command without password

The same cannot be called with psql -c "SELECT ...\gexec" since \gexec is a psql meta‑command and the -c option expects a single command for which the manual states:

command must be either a command string that is completely parsable by the server (i.e., it contains no psql-specific features), or a single backslash command. Thus you cannot mix SQL and psql meta-commands within a -c option.

Workaround from within Postgres transaction

You could use a dblink connection back to the current database, which runs outside of the transaction block. Effects can therefore also not be rolled back.

Install the additional module dblink for this (once per database):

How to use (install) dblink in PostgreSQL?

Then:

DO
$do$
BEGIN
   IF EXISTS (SELECT FROM pg_database WHERE datname = 'mydb') THEN
      RAISE NOTICE 'Database already exists';  -- optional
   ELSE
      PERFORM dblink_exec('dbname=' || current_database()  -- current db
                        , 'CREATE DATABASE mydb');
   END IF;
END
$do$;

Again, you may need more psql options for the connection. See Ortwin's added answer:

Simulate CREATE DATABASE IF NOT EXISTS for PostgreSQL?

Detailed explanation for dblink:

How do I do large non-blocking updates in PostgreSQL?

You can make this a function for repeated use.

Question 2

another alternative, just in case you want to have a shell script which creates the database if it does not exist and otherwise just keeps it as it is:

psql -U postgres -tc "SELECT 1 FROM pg_database WHERE datname = 'my_db'" | grep -q 1 || psql -U postgres -c "CREATE DATABASE my_db"

I found this to be helpful in devops provisioning scripts, which you might want to run multiple times over the same instance.

For those of you who would like an explanation:

-c = run command in database session, command is given in string
-t = skip header and footer
-q = silent mode for grep 
|| = logical OR, if grep fails to find match run the subsequent command

Question 3

If you don't care about the data, you can drop database first and then recreate it:

DROP DATABASE IF EXISTS dbname;
CREATE DATABASE dbname;

Question 4

PostgreSQL does not support IF NOT EXISTS for CREATE DATABASE statement. It is supported only in CREATE SCHEMA. Moreover CREATE DATABASE cannot be issued in transaction therefore it cannot be in DO block with exception catching.

When CREATE SCHEMA IF NOT EXISTS is issued and schema already exists then notice (not error) with duplicate object information is raised.

To solve these problems you need to use dblink extension which opens a new connection to database server and execute query without entering into transaction. You can reuse connection parameters with supplying empty string.

Below is PL/pgSQL code which fully simulates CREATE DATABASE IF NOT EXISTS with same behavior like in CREATE SCHEMA IF NOT EXISTS. It calls CREATE DATABASE via dblink, catch duplicate_database exception (which is issued when database already exists) and converts it into notice with propagating errcode. String message has appended , skipping in the same way how it does CREATE SCHEMA IF NOT EXISTS.

CREATE EXTENSION IF NOT EXISTS dblink;

DO $$
BEGIN
PERFORM dblink_exec('', 'CREATE DATABASE testdb');
EXCEPTION WHEN duplicate_database THEN RAISE NOTICE '%, skipping', SQLERRM USING ERRCODE = SQLSTATE;
END
$$;

This solution is without any race condition like in other answers, where database can be created by external process (or other instance of same script) between checking if database exists and its own creation.

Moreover when CREATE DATABASE fails with other error than database already exists then this error is propagated as error and not silently discarded. There is only catch for duplicate_database error. So it really behaves as IF NOT EXISTS should.

You can put this code into own function, call it directly or from transaction. Just rollback (restore dropped database) would not work.

Testing output (called two times via DO and then directly):

$ sudo -u postgres psql
psql (9.6.12)
Type "help" for help.

postgres=# \set ON_ERROR_STOP on
postgres=# \set VERBOSITY verbose
postgres=# 
postgres=# CREATE EXTENSION IF NOT EXISTS dblink;
CREATE EXTENSION
postgres=# DO $$
postgres$# BEGIN
postgres$# PERFORM dblink_exec('', 'CREATE DATABASE testdb');
postgres$# EXCEPTION WHEN duplicate_database THEN RAISE NOTICE '%, skipping', SQLERRM USING ERRCODE = SQLSTATE;
postgres$# END
postgres$# $$;
DO
postgres=# 
postgres=# CREATE EXTENSION IF NOT EXISTS dblink;
NOTICE:  42710: extension "dblink" already exists, skipping
LOCATION:  CreateExtension, extension.c:1539
CREATE EXTENSION
postgres=# DO $$
postgres$# BEGIN
postgres$# PERFORM dblink_exec('', 'CREATE DATABASE testdb');
postgres$# EXCEPTION WHEN duplicate_database THEN RAISE NOTICE '%, skipping', SQLERRM USING ERRCODE = SQLSTATE;
postgres$# END
postgres$# $$;
NOTICE:  42P04: database "testdb" already exists, skipping
LOCATION:  exec_stmt_raise, pl_exec.c:3165
DO
postgres=# 
postgres=# CREATE DATABASE testdb;
ERROR:  42P04: database "testdb" already exists
LOCATION:  createdb, dbcommands.c:467

Question 5

I had to use a slightly extended version @Erwin Brandstetter used:

DO
$do$
DECLARE
  _db TEXT := 'some_db';
  _user TEXT := 'postgres_user';
  _password TEXT := 'password';
BEGIN
  CREATE EXTENSION IF NOT EXISTS dblink; -- enable extension 
  IF EXISTS (SELECT 1 FROM pg_database WHERE datname = _db) THEN
    RAISE NOTICE 'Database already exists';
  ELSE
    PERFORM dblink_connect('host=localhost user=' || _user || ' password=' || _password || ' dbname=' || current_database());
    PERFORM dblink_exec('CREATE DATABASE ' || _db);
  END IF;
END
$do$

I had to enable the dblink extension, plus i had to provide the credentials for dblink. Works with Postgres 9.4.

Question 6

If you can use shell, try

psql -U postgres -c 'select 1' -d $DB &>dev/null || psql -U postgres -tc 'create database $DB'

I think psql -U postgres -c "select 1" -d $DB is easier than SELECT 1 FROM pg_database WHERE datname = 'my_db',and only need one type of quote, easier to combine with sh -c.

I use this in my ansible task

- name: create service database
  shell: docker exec postgres sh -c '{ psql -U postgres -tc "SELECT 1" -d {{service_name}} &> /dev/null && echo -n 1; } || { psql -U postgres -c "CREATE DATABASE {{service_name}}"}'
  register: shell_result
  changed_when: "shell_result.stdout != '1'"

Question 7

The best way is just running the SQL.

CREATE DATABASE MY_DATABASE;

if the database already exists, it throws "database already exists error" which you can do whatever you want to do, otherwise it creates the database. I do not think it will create a new database on top of yours. :D

Question 8

After reading through all these in my opinion complicated solutions that are terrible work arounds for the lack of the IF NOT EXIST option for postgres user creation, I forgot almost there is a simply way to just handle it at the shell level. Even though it might not be what some want, I think a lot of folks want simplicity and not creating procedures and complicated constructs.

I'm using docker, here are the important snippets from my bash script that loads data in a devsetup:

execute_psql_command_pipe () {
         $DOCKER_COMMAND exec -it $POSTGRES_CONTAINER bash -c "echo \"$1\"| psql -h localhost -U postgres || echo psql command failed - object likely exists"
}

read -r -d '' CREATE_USER_COMMANDS << EOM
create user User1 WITH PASSWORD 'password';
create user User2 WITH PASSWORD 'password';
EOM

execute_psql_command_pipe "$CREATE_USER_COMMANDS"

There are a few things wrong with it, but it's the simplest way I could find to make it do what I want: create on first pass of script, continue on second pass when existing. By the way, the echo output does not show, but the commands continue because the echo command exits with 0.

The same can be done for any command (like db create). This obviously fails (or succeeds, depending on perspective) for any other error that may occur too, but you get the psql output printer so more handling can be added.

Question 9

Just create the database using createdb CLI tool:

PGHOST="my.database.domain.com"
PGUSER="postgres"
PGDB="mydb"
createdb -h $PGHOST -p $PGPORT -U $PGUSER $PGDB

If the database exists, it will return an error:

createdb: database creation failed: ERROR:  database "mydb" already exists

Question 10

One simple clean way to do it that I ended up using:

createdb $DATABASE 2> /dev/null || echo "database already exists"

If you expect other error than database "x" already exists that obviously won't work (e.g. permission denied). In any case, if that is a concern, one can always perform such checks prior to this point.

Don't forget to set the value for DATABASE, and to pass in the required switches for the createdb command. Preferably you can also do like:

export PGHOST=localhost
export PGUSER=user
export PGPASSWORD=p455w0rd
...