P3 is a modern, lean and mean PostgreSQL client for Pharo.

P3Client uses frontend/backend protocol 3.0 (PostgreSQL version 7.4 [2003] and later), implementing the simple and extended query cycles. It supports plaintext, md5 and scram-sha-256 password authentication. When SQL queries return row data, it efficiently converts incoming data to objects. P3Client supports most common PostgreSQL types.

P3Client can be configured manually or through a URL.

P3Client new url: 'psql://username:password@localhost:5432/databasename'.

Not all properties need to be specified, the minimum is the following URL.

P3Client new url: 'psql://user@localhost'.

P3Client has a minimal public protocol, basically #query: (#execute: is an alias).

Opening a connection to the server (#open) and running the authentication and startup protocols (#connect) are done automatically when needed from #query.

P3Client also supports SSL connections. Use #connectSSL to initiate such a connection. Alternatively you can add sslmode=require to the connection URL, as in 'psql://username:password@localhost:5432/databasename?sslmode=require'.

Through the #prepare: message, you can ask P3Client to prepare/parse an SQL statement or query with parameters. This will give you a P3PreparedStatement instance than you can then execute with specific parameters. Polymorphic to this there is also P3FormattedStatement which you create using the #format: message. These work at the textual, client side level.

Here is the simplest test that does an actual query, it should return true.

(P3Client new url: 'psql://sven@localhost') in: [ :client |
   [ client isWorking ] ensure: [ client close ] ].

This is how to create a simple table with some rows in it.

(P3Client new url: 'psql://sven@localhost') in: [ :client |
   client execute: 'DROP TABLE IF EXISTS table1'.
   client execute: 'CREATE TABLE table1 (id INTEGER, name TEXT, enabled BOOLEAN)'.
   client execute: 'INSERT INTO table1 (id, name, enabled) VALUES (1, ''foo'', true)'.
   client execute: 'INSERT INTO table1 (id, name, enabled) VALUES (2, ''bar'', false)'.
   client close ].

Now we can query the contents of the simple table we just created.

(P3Client new url: 'psql://sven@localhost') in: [ :client |
   [ client query: 'SELECT * FROM table1' ] ensure: [ client close ] ].

The result is an instance of P3Result

   a P3Result('SELECT 2' 2 records 3 columns)

P3Result contains 3 elements, results, descriptions & data:

• Results is a string (collection of strings for multiple embedded queries) indicating successful execution.
• Descriptions is a collection of row field description objects.
• Data is a collection of rows with fully converted field values as objects.

The data itself is an array with 2 sub arrays, one for each record.

#( #(1 'foo' true) #(2 'bar' false) )

Finally we can clean up.

(P3Client new url: 'psql://sven@localhost') in: [ :client |
   [ client execute: 'DROP TABLE table1' ] ensure: [ client close ] ].

• https://postgresql.org
• https://en.wikipedia.org/wiki/PostgreSQL
• https://www.postgresql.org/docs/9.6/static/protocol.html

Although you are free to create your SQL statements in any way you see fit, feeding them to #execute: and #query:, inserting arguments in SQL statements can be hard (because you have to know the correct syntax), error-prone (because you might violate syntax rules) and dangerous (due to SQL injection attacks).

P3 can help here with two mechanisms: prepared and formatted statements. They are mostly polymorphic and use the same template notation. They allow you to create a statement once, specifying placeholders with $n, and execute it once or multiple times with concrete arguments, with the necessary conversions happening automatically.

The difference between the two is that formatted statements are implemented using simple textual substitution on the client side, while prepared statements are evaluated on the server side with full syntax checking, and are executed with more type checks. Prepared statements are more efficient since the server can do part of its optimization in the prepare phase, saving time on each execution.

Here is a transcript of how to use them. First we set up a client and create a test table.

client := P3Client new url: 'psql://sven@localhost'.

client execute: 'DROP TABLE IF EXISTS table1'.
client execute: 'CREATE TABLE table1 (id INTEGER, name TEXT, weight REAL, enabled BOOLEAN)'.

Next we insert some data and then query it using prepared statements.

statement := client prepare: 'INSERT INTO table1 (id, name, weight, enabled) VALUES ($1, $2, $3, $4)'.

statement execute: { 1. 'foo'. 75.5. true }.
statement executeBatch: { { 2. 'bar'. 80.25. true }. { 3. 'foobar'. 10.75. false } }.

statement close.

statement := client prepare: 'SELECT id, name, weight FROM table1 WHERE id = $1 AND enabled = $2'.
statement query: { 1. true }.

statement close.

Note that prepared statements are server side resources that need to be closed when no longer needed. Prepared statements exist in the scope of a single session/connection.

Next we start over and do the same insert and query using formatted statements.

client execute: 'TRUNCATE TABLE table1'.

statement := client format: 'INSERT INTO table1 (id, name, weight, enabled) VALUES ($1, $2, $3, $4)'.

statement execute: { 1. 'foo'. 75.5. true }.
statement executeBatch: { { 2. 'bar'. 80.25. true }. { 3. 'foobar'. 10.75. false } }.

statement := client format: 'SELECT id, name, weight FROM table1 WHERE id = $1 AND enabled = $2'.
statement query: { 1. true }.

And finally we clean up.

client execute: 'DROP TABLE table1'.
client close.

P3 supports most common PostgreSQL types. Here are some tables with the details. As of PostgreSQL 9.6, there are 41 general purpose data types of which 32 are currently implemented.

These are the 32 general purpose data type currently implemented, with the Pharo class they map to.

Here are the 9 general purpose data types that are not yet implemented.

Additionally, the following 10 common types are also implemented, with the Pharo class they map to.

P3 also supports enums. Each enum definition creates a new type. You can send #loadEnums to P3Client to create mappings for all visible enums.

When you do a query that results in data of an unknown type you will get an error, P3 cannot convert typeOid XXX, where XXX is the oid in the pg_type table.

P3 connects over the network (TCP) to PostgreSQL and supports plain (#connect) and TLS/SSL encrypted (#connectSSL) connections.

It is out of the scope of this README to explain how to install and configure an advanced database like PostgreSQL. There is extensive high quality documentation available covering all aspect of PostgreSQL, see https://postgresql.org

Out of the box, most PostgreSQL installations do not allow for network connections from other machines, only for local connections. Check the listen_addresses directive in postgresql.conf

As for authentication, CleartextPassword, MD5Password and SCRAM-SHA-256 are supported. This means that SCMCredential, GSS, SSPI are currently not (yet) supported. An error will be signalled when the server requests an unsupported authentication.

You have to create database users, called roles and give them a password. In SQL you can do this with CREATE|ALTER ROLE user1 LOGIN PASSWORD 'secret'

Next you have to tell PostgreSQL how network users should authenticate themselves. This is done by editing pg_hba.conf choosing specific methods, trust (no password, no authentication), password, md5 and scram-sha-256 work with P3.

Note that for SCRAM-SHA-256 to work, you need to change the password_encryption directive in postgresql.conf to scam-sha-256, restart and reenter all user passwords.

Included is P3DatabaseDriver, an interface between Glorp, an advanced object-relational mapper, and P3Client.

To install this driver (after loading Glorp itself), do

PharoDatabaseAccessor DefaultDriver: P3DatabaseDriver.

Configure your session using a Glorp Login object

Login new
   database: PostgreSQLPlatform new;
   username: 'username';
   password: 'password';
   connectString: 'host:5432_databasename';
   encodingStrategy: #utf8;
   yourself.

The default group loads P3Client and its basic dependencies NeoJSON and ZTimestamp

Metacello new
   baseline: 'P3';
   repository: 'github://svenvc/P3';
   load.

The glorp group loads P3DatabaseDriver and the whole of Glorp (warning: large download)

Metacello new
   baseline: 'P3';
   repository: 'github://svenvc/P3';
   load: 'glorp'.

P3ClientTest holds unit tests for the P3 PSQL client.

Configure it by setting its class side's connection URL.

P3ClientTest url: 'psql://sven:secret@localhost:5432/database'.

The minimal being the following:

P3ClientTest url: 'psql://sven@localhost'.

P3 uses object logging, an advanced form of code instrumentation. This means that during execution instances of subclasses of P3LogEvent are created (some including timing information) and sent to an Announcer (accessible via P3LogEvent announcer). Interested parties can subscribe to these log events and use the information contained in them to learn about P3 code execution.

The standard print method of a P3LogEvent can be used to generate textual output. The following expression enables logging to the Transcript.

P3LogEvent logToTranscript.

Executing the four expressions of the Basic Usage section yields the following output.

2020-09-21 16:27:57 001 [P3] 63731 #Connect sven@localhost:5432 Trust
2020-09-21 16:27:57 002 [P3] 63731 #Query SELECT 565 AS N
2020-09-21 16:27:57 003 [P3] 63731 #Result SELECT 1, 1 record, 1 colum, 4 ms
2020-09-21 16:27:57 004 [P3] 63731 #Close

2020-09-21 16:28:07 005 [P3] 63733 #Connect sven@localhost:5432 Trust
2020-09-21 16:28:07 006 [P3] 63733 #Query DROP TABLE IF EXISTS table1
2020-09-21 16:28:07 007 [P3] 63733 #Error P3Notification PostgreSQL table "table1" does not exist, skipping
2020-09-21 16:28:07 008 [P3] 63733 #Result DROP TABLE, 6 ms
2020-09-21 16:28:07 009 [P3] 63733 #Query CREATE TABLE table1 (id INTEGER, name TEXT, enabled BOOLEAN)
2020-09-21 16:28:07 010 [P3] 63733 #Result CREATE TABLE, 50 ms
2020-09-21 16:28:07 011 [P3] 63733 #Query INSERT INTO table1 (id, name, enabled) VALUES (1, 'foo', true)
2020-09-21 16:28:07 012 [P3] 63733 #Result INSERT 0 1, 4 ms
2020-09-21 16:28:07 013 [P3] 63733 #Query INSERT INTO table1 (id, name, enabled) VALUES (2, 'bar', false)
2020-09-21 16:28:07 014 [P3] 63733 #Result INSERT 0 1, 0 ms
2020-09-21 16:28:07 015 [P3] 63733 #Close

2020-09-21 16:28:20 016 [P3] 63737 #Connect sven@localhost:5432 Trust
2020-09-21 16:28:20 017 [P3] 63737 #Query SELECT * FROM table1
2020-09-21 16:28:20 018 [P3] 63737 #Result SELECT 2, 2 records, 3 columns, 2 ms
2020-09-21 16:28:20 019 [P3] 63737 #Close

2020-09-21 16:39:52 020 [P3] 63801 #Connect sven@localhost:5432 Trust
2020-09-21 16:39:52 021 [P3] 63801 #Query DROP TABLE table1
2020-09-21 16:39:52 022 [P3] 63801 #Result DROP TABLE, 13 ms
2020-09-21 16:39:52 023 [P3] 63801 #Close

Remember that the information inside the log events can be used to build other applications.

The main goal of P3 is to be a modern, lean and mean PostgreSQL client for Pharo. Right now, P3 is functional and usable.

The quality of open source software is determined by it being alive, supported and maintained.

The first way to help is to simply use P3 in your projects and tells us about your successes and the issues that you encounter. You can ask questions on the Pharo mailing lists.

Development happens on GitHub, where you can create issues.

Contributions should be done with pull requests solving specific issues.