Miss any of our Open RFC calls?Watch the recordings here! »


0.7.2 • Public • Published


Build Status

simple elegant sql for nodejs

mesa is not an orm. it aims to help as much as possible with the construction, composition and execution of sql queries while not restricting full access to the underlying database driver and database in any way.

mesa builds on top of mohair, a simple fluent sql query builder.

it adds the ability to run queries on connections, process query results, to declare and include associations (hasOne, belongsTo, hasMany, hasManyThrough) and more.

mesa has been battle tested in a medium sized (8 heroku dynos) production environment for half a year.

mesa uses criterion for sql-where-conditions. consult the criterion readme and mohair readme to get the full picture of what is possible with mesa.


npm install mesa


mesa has a fluent interface where every method returns a new object. no method ever changes the state of the object it is called on. this enables a functional programming style.


var mesa = require('mesa');


mesa only works with node-postgres at the moment

tell mesa how to get a connection from the pool:

var pg = require('pg');
var mesaWithConnection = mesa.connection(function(cb) {
    pg.connect('tcp://username@localhost/database', cb);

mesaWithConnection will now use the provided function to get connections for the commands you execute. these connections are under mesa's control. mesa will properly call done() on every connection it has obtained from the pool.


specify the table to use:

var userTable = mesaWithConnection.table('user');


        name: 'alice'
    }, function(err, id) {

attributes() sets the properties to pick from data in the create() and update() methods. attributes() prevents mass assignment and must be called before using the create() or update() methods.

insert multiple records
        {name: 'alice'},
        {name: 'bob'}
    ], function(err, ids) {
insert with some raw sql
    .attributes(['name', 'created'])
        name: 'alice',
        created: userTable.raw('NOW()')
    }, function(err, id) {

raw() can be used to inject arbitrary sql instead of binding a parameter.

userTable.where({id: 3}).delete(function(err) {

see the criterion readme for all the ways to specify where conditions in mesa.

    .where({id: 3})
    .where({name: 'alice'})
    .update({name: 'bob'}, function(err) {

multiple calls to where are anded together.


find the first
userTable.where({id: 3}).first(function(err, user) {
test for existence
userTable.where({id: 3}).exists(function(err, exists) {
find all
userTable.where({id: 3}).find(function(err, user) {
select, join, group, order, limit, offset
    .select('user.*, count(project.id) AS project_count')
    .where({id: 3})
    .where('name = ?', 'foo')
    .join('JOIN project ON user.id = project.user_id')
    .order('created DESC, name ASC')
    .find(function(err, users) {


has one

use hasOne if the foreign key is in the other table (addressTable in this example)

var userTable = userTable.hasOne('address', addressTable, {
    primaryKey: 'id',               // optional with default: 'id'
    foreignKey: 'user_id'           // optional with default: userTable.getTable() + '_id'

the second argument can be a function which must return a mesa object. this can be used to resolve tables which are not yet created when the association is defined. it's also a way to do self associations.

belongs to

use belongsTo if the foreign key is in the table that belongsTo is called on (projectTable in this example)

var projectTable = projectTable.belongsTo('user', userTable, {
    primaryKey: 'id',               // optional with default: 'id'
    foreignKey: 'user_id'           // optional with default: userTable.getTable() + '_id'
has many

use hasMany if the foreign key is in the other table (userTable in this example) and there are multiple associated records

var userTable = userTable.hasMany('projects', projectTable, {
    primaryKey: 'id',               // optional with default: 'id'
    foreignKey: 'user_id'           // optional with default: userTable.getTable() + '_id'
has many through

use hasManyThrough if the association uses a join table

var userProjectTable = mesaWithConnection.table('user_project');
var userTable = userTable.hasManyThrough('projects', projectTable, userProjectTable,
    primaryKey: 'id',               // optional with default: 'id'
    foreignKey: 'user_id',          // optional with default: userTable.getTable() + '_id'
    otherPrimaryKey: 'id',          // optional with default: 'id'
    otherForeignKey: 'project_id'   // optional with default: projectTable.getTable() + '_id'
including associated

associations are only fetched if you include them:

userTable.includes({address: true}).find(function(err, users) {

includes can be nested arbitrarily deep:

        shipping_address: {
            street: true,
            town: true
        billing_address: true,
        friends: {
            billing_address: true
    .find(function(err, users) {

advanced use

extending mesa's fluent interface

every mesa object prototypically inherits from the object before it in the fluent call chain.

this means that every mesa object is very lightweight since it shares structure with objects before it in the fluent call chain.

it also makes it very easy to extend mesa's fluent interface:

var userTable = mesa.table('user');
userTable.activeAdmins = function() {
    return this.where({visible: true, role: 'admin'});
userTable.whereCreatedBetween = function(from, to) {
    return this.where('created BETWEEN ? AND ?', from, to);
    .order('created DESC')
    .whereCreatedBetween(new Date(2013, 4, 10), new Date(2013, 4, 12))
    .find(function(err, users) {
user controlled connections

sometimes, when using a transaction, you need to run multiple commands over multiple tables on the same connection.

use getConnection() to get a raw connection from mesa. you can then run arbitrary sql on that connection. use connection() with a connection object to tell mesa to explicitely use that connection instead of getting a new one from the pool:

userTable.getConnection(function(err, connection, done) {
    connection.query('BEGIN', function(err) {
            // use the transactional connection explicitely
            .insert({name: 'alice'}, function(err, id) {
                // run more commands in the transaction
                // possibly on other tables
                connection.query('COMMIT', function(err) {

when you are done using the connection you need to call done() to tell node-postgres to return the connection to the pool. otherwise you will leak that connection, which is very bad since your application will run out of connections and hang.

license: MIT


npm i mesa

DownloadsWeekly Downloads






Last publish


  • avatar