A simple pool implementation for postgresql with lightweight client code and helper functions. Pooling is done automatically but it allows you to take control for transactions. Since v1.2.0 the client is merely a single layer on top of the native bindings to libpq. Since v1.3.0 the client skips column name and type parsing for prepared queries.
The package libpq that provides the native bindings, requires the PostgreSQL client libraries & tools. To validate them, check if the command pg_config is known to your machine. If not, take the following steps:
brew install libpq
If necessary, search with:
sudo find / -name "pg_config" -print
Then add the result to your shell file, e.g.:
echo 'export PATH="/opt/homebrew/Cellar/libpq/15.2/bin:$PATH"' >> ~/.zshrc
apt-get install libpq-dev g++ make
yum install postgresql-devel
Install PostgreSQL (http://www.postgresql.org/download/windows/) and add your PostgreSQL installation's bin folder to the system path (e.g. C:\Program Files\PostgreSQL\X.X\bin). Make sure that both libpq.dll and pg_config.exe are in that folder.
Once the tools are installed, install the package:
npm install pg-pool-minimal
let pool = new Postgres({
user: 'postgres',
host: '127.0.0.1',
port: 5432,
database: 'template1',
schema: 'public',
socket: '/var/run/postgresql/',
password: '',
threads: 10,
queueSize: 257594,
escapeChar: '\\',
valuesOnly: false,
parseInt8AsString: false
});
await pool.initialize();
By default queries will return objects mapping column name to value. If you want the queries to only return an array of values instead of said object (column order will be guaranteed), you can set valuesOnly to true.
If you do not want to or can use a unix socket, leave the socket parameter undefined.
The queueSize defines the internal command queue of the pool, the sample size should suffice. If you think you have more elements waiting for a free connection, you can either scale up or consider checking for a bottle neck.
parseInt8AsString determines if a int8 should be parsed as a string so you can perform your own conversion since Number.MAX_SAFE_INTEGER is ~9 × 10^15 and the max value of pg's int8 is ~9 × 10^18 which would result in conversion problems.
let items: Item[] = await pool.query("get-player-items", "SELECT * FROM items WHERE player_id=$1", [player_id]);
Note: You do not have to supply a parameter
let players: Player[] = await pool.queryString("SELECT * FROM players");
Using the function queryCount and queryStringCount you can, instead of making the client parse a query result, simply get the number of affected rows.
Performing transactions requires you to use the same client. Use the connect function to aquire one and execute your queries. Make sure to finally release the client to the pool soo it does not clog and die horribly.
let client = await pool.connect();
try{
await client.queryString('START TRANSACTION ISOLATION LEVEL SERIALIZABLE;');
await client.query('update','UPDATE users SET value=$1',[100]);
await client.queryString('COMMIT');
} catch(error){
//Handle error
await client.queryString('ROLLBACK');
} finally{
client.release();
}
In case you want to allow searches based on user input in your backend, using LIKE '%userinput%' for example, you can escape the input using
let escaped = pool.escapeWildcards(input)
This will escape characters that would otherwise be considered as patterns. You can define your escape character in the ClientConfig of the pool constructor.
You can choose names for your prepared queries of course, but normally you would not care about the concrete name. You can generate an alias for your query that is unique and as small as possible
let identifier = pool.getPrepareIdentifier();
Example usage:
const getAllItemsName = Pool.getPrepareIdentifier();
const getAllItemsQuery = "SELECT * FROM items WHERE player_id=$1";
function getItemsForPlayer(player_id: number){
items : Item[] = await pool.query(getAllItemsName, getAllItemsQuery, [player_id]);
}