xql

Extensible and dependency free SQL builder that provides an expression tree and supports multiple database dialects.

Usage no npm install needed!

<script type="module">
  import xql from 'https://cdn.skypack.dev/xql';
</script>

README

xql.js

Extensible and dependency free SQL builder and expression tree for node.js.

Disclaimer

This library is used in production, but it doesn't contain all possible features of all available DB engines. It started initially with only PostgreSQL support, but now also MySQL, MSSQL, and SQLite dialects are available.

Be prepared for some minor API changes before the library stabilizes.

Introduction

xql.js is a library designed to build SQL queries programmatically. It provides SQL expression tree that is created by high level API calls which mimic SQL syntax. It's a tool that helps to create the SQL expression tree that can be compiled into a single query string at the end of the building phase. The library has been designed primarily for DAO/DB layers, but use-cases are nearly unlimited.

There are several reasons why xql.js has been developed:

  1. Full support and focus on PostgreSQL (PG is the primary engine, but xql is getting support for MySQL and SQLite3 as well).
  2. High performance and low memory footprint, see jsstuff/xql-bench that compares with other engines.
  3. Schemaless by design, but allows to specify type-mapping so the input data can be properly escaped.
  4. Control of SQL parameters and the way they are formatted / escaped.
  5. Construction of SQL query shouldn't require RAW expressions to be written, but it should be easy to use RAW expressions in case they are needed.
  6. No more legacy JS (xql.js is based on ES6 classes), however, it doesn't dictate you how to write your own code.

There are several node.js libraries that focus on SQL query building, but none has satisfied all the needs. The closest library and huge inspiration for xql.js was Python's SqlAlchemy, which is much more advanced compared to any node.js SQL framework at the moment. However, xql.js is just a query builder that has a type-mapping feature, which is used describe column types for serialization, but they are not used to describe relations or anything else. There are no plans to add ORM support to xql.js in any future release.

To simplify the library design and use cases, xql.js itself doesn't implement any functionality to talk to a real database - is just a query builder. There is another project in preparation that will bridge xql.js with node.js SQL drivers, but since there are so many libraries that can be used (including libraries for SQL connection pooling) there was no real work done to create another library for this purpose yet.

At the beginning, xql.js has been designed to work primarily with PostgreSQL, but other dialects are already in-progress and some code that brings initial support for MySQL and SQLite3 has landed.

Basic Usage

To use xql.js in node.js add "xql" library to your package.json and then require("xql") it. You need to create a context before you compile your expressions:

const xql = require("xql");

// Create your context - context is used to hold database dialect and some
// options. It doesn't hold any intermediate data. It's perfectly fine to
// use one context for all your queries (and it's designed this way).
const ctx = xql.dialect.newContext({ dialect: "pgsql" /* [more options]*/ });

// Create some query.
var query = xql.SELECT("*")
  .FROM("cities")
  .WHERE("population", ">=", 1000000) // 3 form WHERE.
  .WHERE("capital", true);            // 2 form WHERE, implicit equality.

// Use context to compile the query.
console.log(query.compileStatement(ctx));
// SELECT * FROM "cities" WHERE "population" >= 1000000 AND "capital" = TRUE;

If you plan to pretty-print your queries for debugging purposes, use pretty and optionally indentation (default 2) option:

const xql = require("xql");
const ctx = xql.dialect.newContext({
  dialect: "pgsql"
  pretty: true
});

var query = xql.SELECT("*")
  .FROM("cities")
  .WHERE("population", ">=", 1000000)
  .WHERE("capital", true);

console.log(query.compileStatement(ctx));
// SELECT
//   *
// FROM
//   "cities"
// WHERE
//   "population" >= 1000000 AND "capital" = TRUE;

If you ask yourself why all SQL constructs are UPPERCASED the explanation is very simple: in the past xql.js supported both conventions (UPPERCASED and camelCased), but it led to confusion and ambiguity. The new API follows a very simple rule: if any function creates a new SQL expression or modifies an existing one based on SQL semantics it's name is always UPPERCASED, otherwise it's camelCased (utility functions, etc). This way it's very simple to visually distinguish between SQL building blocks and other logic in your own code. Please open an issue if you would like to discuss other possibilities.

API Overview

xql.js library consists of several nested namespaces, however, they are rarely used outside of xql implementation:

Namespace Description
xql High-level SQL builder interface targeting end-users
xql.error Namespace that provides custom errors used by xql.js
xql.misc SQL utilities made public, contains also a VERSION member in a "major.minor.patch" form
xql.node SQL expression tree, contains xql.node.Node and all nodes that inherit from it

Error classes:

Error Description
xql.error.ValueError Error thrown if data is wrong
xql.error.CompileError Error thrown if query is wrong

Expression tree:

Node Description
xql.node.Node Base node, all SQL nodes inherit from it, it's safe to use instanceof operator to check whether an object is a xql.node.Node
xql.node.NodeArray Contains array of SQL nodes or values
xql.node.Raw Raw SQL expression intended to be used unescaped (the only way to pass something, which will not be escaped)
xql.node.Value SQL value base class
xql.node.Identifier SQL identifier, like table or column
xql.node.Unary SQL unary node (can contain a single child)
xql.node.Binary SQL binary node (can contain two children, left and right)
xql.node.Func SQL function or aggregate
xql.node.Case SQL CASE construct
xql.node.When SQL WHEN construct
xql.node.Logical Logical operator like AND and OR, which is based on NodeArray and can contain more than two expressions
xql.node.ConditionalMap Special node that contains key/value interface that can be used to construct WHERE like expressions without constructing xql.node.Logical nodes.
xql.node.Join SQL JOIN construct
xql.node.Sort SQL ORDER BY construct
xql.node.With Expression representing "identifier" AS (SELECT ...) part of WITH clause.
xql.node.Statement Base class representing a single SQL statement, which should end with semicolon
xql.node.QueryStatement Base class used by SELECT, INSERT, UPDATE, and DELETE statements
xql.node.SelectStatement SQL SELECT statement
xql.node.InsertStatement SQL INSERT statement
xql.node.UpdateStatement SQL UPDATE statement
xql.node.DeleteStatement SQL DELETE statement
xql.node.CompoundStatement SQL UNION, INTERSECT, and EXCEPT operators that can be used to combine multiple query statements

High-level SQL builder concepts:

SQL-Builder API Description
xql.TABLE(...) Create a xql.node.Identifier wrapping a table name
xql.COLUMN(...) Create a xql.node.Identifier wrapping a column name (in a format "column" or "table"."column" or "namespace"."table"."column")
xql.COL(...) Alias to xql.COLUMN
xql.VALUE(...) Create a xql.node.Value wrapping a value like null, boolean, number, or string
xql.VAL(...) Alias to xql.VALUE.
xql.VALUES(...) Create a xql.node.Value wrapping an array into SQL VALUES
xql.DATE(...) Create a xql.node.Value wrapping a DATE value
xql.TIME(...) Create a xql.node.Value wrapping a TIME value
xql.TIMESTAMP(...) Create a xql.node.Value wrapping a TIMESTAMP value
xql.TIMESTAMPTZ(...) Create a xql.node.Value wrapping a TIMESTAMPTZ value
xql.INTERVAL(...) Create a xql.node.Value wrapping a INTERVAL value
xql.ARRAY(...) Create a xql.node.Value wrapping an ARRAY value
xql.JSON_(...) Create a xql.node.Value wrapping a JSON value
xql.RAW(s, bindings) Create a RAW query xql.node.Raw node based on query string s and optional bindings
xql.OP(...) Create a xql.node.Unary or xql.node.Binary node depending on the count of parameters. The most used form is a 3 operand form, which is used to describe a binary expression.

For example OP(COL("salary"), "+", 500).AS("newSalary") can be used to describe an expression like "salary" + 500 AS "newSalary". Please note that AND and OR operators should always use xql.node.Logical as xql.js can construct queries containing multiple AND and OR leaves
xql.EQ(a, b) Create a xql.node.Binary node describing a = b expression
xql.NE(a, b) Create a xql.node.Binary node describing a <> b expression
xql.LT(a, b) Create a xql.node.Binary node describing a < b expression
xql.LE(a, b) Create a xql.node.Binary node describing a <= b expression
xql.GT(a, b) Create a xql.node.Binary node describing a > b expression
xql.GE(a, b) Create a xql.node.Binary node describing a >= b expression
xql.IS(a, b) Create a xql.node.Binary node describing a IS b expression (you can use EQ as well which would detect IS case)
xql.IS_DISTINCT_FROM(a, b) Create a xql.node.Binary node describing a IS DISTINCT FROM b expression
xql.LIKE(a, b) Create a xql.node.Binary node describing a LIKE b expression
xql.ILIKE(a, b) Create a xql.node.Binary node describing a ILIKE b expression
xql.SIMILAR_TO(a, b) Create a xql.node.Binary node describing a SIMILAR TO b expression
xql.IN(a, b) Create a xql.node.Binary node describing a IN (b) expression
xql.NOT_IN(a, b) Create a xql.node.Binary node describing a NOT IN (b) expression
xql.BETWEEN(x, a, b) Create a xql.node.Func node describing x BETWEEN a AND b expression
xql.NOT_BETWEEN(x, a, b) Create a xql.node.Func node describing x NOT BETWEEN a AND b expression
xql.FUNCTION_NAME(...) Create a xql.node.Func node describing FUNCTION_NAME(...) expression. Note that FUNCTION_NAME has to be replaced by the name of the function to be used, for example xql.SIN(...) describes SIN() function and xql.COUNT(...) describes COUNT() aggregate
xql.AND(...) Create a xql.node.Logical expression describing AND expression
xql.OR(...) Create a xql.node.Logical expression describing OR expression
xql.SELECT(...) Create a xql.node.SelectStatement and pass optional arguments to the SelectStatement.FIELD(...) method
xql.INSERT(...) Create a xql.node.InsertStatement and use an optional first argument as a table name (FROM clause) if it's a string or an identifier, and pass all other arguments to SelectStatement.FIELD(...) method
xql.UPDATE(...) Create a xql.node.UpdateStatement and use an optional first argument as a table name (UPDATE ... clause) if it's a string or an identifier, and pass all other arguments to UpdateStatement.FIELD(...) method
xql.DELETE(...) Create a xql.node.DeleteStatement and use an optional first argument as a table name
xql.EXCEPT(...) Create a xql.node.CompoundStatement describing EXCEPT expression
xql.EXCEPT_ALL(...) Create a xql.node.CompoundStatement describing EXCEPT ALL query
xql.INTERSECT(...) Create a xql.node.CompoundStatement describing INTERSECT query
xql.INTERSECT_ALL(...) Create a xql.node.CompoundStatement describing INTERSECT ALL query
xql.UNION(...) Create a xql.node.CompoundStatement describing UNION query
xql.UNION_ALL(...) Create a xql.node.CompoundStatement describing UNION ALL query
xql.SORT(c, sort, nulls) Create a xql.node.Sort node wrapping an ORDER BY clause

Generic Interface

Since every node that is used to describe various constructs inherits directly or indirectly from xql.node.Node all nodes share a common interface:

xql.node.Node Description
.getType() Get the node type {String}. For example a xql.node.SelectStatement is a SELECT type, logical operator is AND or OR type, etc...
.setType(type) Set the node type (used internally)
.getLabel() Get the node label that is rendered as AS "label" in SQL
.setLabel(label) Set the node label
.canExecute() Can be used to check whether the node can be executed by SQL engine. Only SELECT, INSERT, UPDATE, and DELETE queries and UNION, INTERSECT, and EXCEPT operators can be executed.
.compileNode(ctx) Compile the node into a string. The ctx argument is currently not used, but it's designed in a way to pass an additional information to the compiler so multiple dialects can be used in the future.
.compileStatement(ctx?) Compile the query, it's basically a compileNode() call with semicolon ";" at the end. This method should be used to return the query to be executed by your DB engine. It's provided by all query nodes.
.AS(label) Alias to setLabel().
.EQ(b) Returns this = b expression.
.NE(b) Returns this <> b expression.
.LT(b) Returns this < b expression.
.LE(b) Returns this <= b expression.
.GT(b) Returns this > b expression.
.GE(b) Returns this >= b expression.
.IN(b) Returns this IN b expression.
.NOT_IN(b) Returns this NOT IN b expression.

For example COL("a").EQ(1) yields the same tree as OP(COL("a"), "=", 1)

The xql.node.Unary interface:

xql.node.Unary Description
.getValue() Get the child node or value
.setValue(value) Set the child node or value

The xql.node.Binary interface:

xql.node.Binary Description
.getLeft() Get the left node or value
.setLeft(left) Set the left node or value
.getRight() Get the right node or value
.setRight(right) Set the right node or value
.addLeft(left)
.addRight(right) Helpers, can only be used if the target value is an array, in such case the value left or right is pushed into it.

SELECT

Select query is described by xql.node.SelectStatement node and wrapped by xql.SELECT(...). It accepts arguments that are passed to the FIELD() method making the SELECT(...), SELECT([...]) and SELECT().FIELD(...) constructs equivalent.

The xql.node.SelectStatement implements the following interface:

xql.node.SelectStatement Description
.FIELD(...)
.FIELD([...]) Add a field or expression to be selected. It accepts a xql.node.Node, column name, or a dictionary defining columns and their expressions.

The FIELD() calls are usually chained. For example FIELD("a").FIELD("b") calls are the same as FIELD("a", "b"), FIELD(["a", "b"]), and FIELD({ a: true, b: true })
.DISTINCT(...) Add a DISTINCT or DISTINCT ON (if arguments are provided) clause to the query.
.FROM(...)
.FROM([...]) Add FROM clause to the query. The method accepts multiple arguments or a list of arguments. Most of the time FROM is used with a single argument describing the table to select from, however, multiple arguments forming an implicit CROSS JOIN construct, which matches the SQL specification, are allowed.

For example FROM(a) construct will generate SELECT ... FROM "a" query, while FROM(a, b) construct will generate SELECT ... FROM "a", "b" or SELECT ... FROM "a" CROSS JOIN "b" (these are equivalent, xql.js can generate any of these depending on the version and implementation changes)
.CROSS_JOIN(with, cond)
.INNER_JOIN(...)
.LEFT_JOIN(...)
.RIGHT_JOIN(...)
.FULL_JOIN(...) Add a JOIN clause to the query. Joins always join the current query with a new table. For example FROM("a").INNER_JOIN("b").LEFT_JOIN("c") construct will generate SELECT ... FROM "a" INNER JOIN "b" LEFT OUTER JOIN "c" query
.WHERE(node)
.WHERE(a, b)
.WHERE(a, op, b) Add a WHERE clause node, WHERE a = b, or WHERE a op b to the query (implicitly ANDed with other WHERE clauses if present)
.OR_WHERE(node)
.OR_WHERE(a, b)
.OR_WHERE(a, op, b) Add a WHERE clause node, WHERE a = b, or WHERE a op b to the query (implicitly ORed with other WHERE clauses if present)
.GROUP_BY(...)
.GROUP_BY([...]) Add a GROUP BY clause to the query. Group by can be specified as a column or a xql.node.Node
.HAVING(node)
.HAVING(a, b)
.HAVING(a, op, b) Add a HAVING clause node, HAVING a = b, or HAVING a op b to the query (implicitly ANDed with other HAVING clauses if present)
.OR_HAVING(node)
.OR_HAVING(a, b)
.OR_HAVING(a, op, b) Add a HAVING clause node, HAVING a = b, or HAVING a op b to the query (implicitly ORed with other HAVING clauses if present)
.ORDER_BY(col, dir, nulls) Add an ORDER BY expression of the form "col" [ASC/DESC] [NULLS FIRST/LAST]. If col is an array the builder will insert multiple sort clauses with the same dir and nulls order
.OFFSET(offset) Add an OFFSET clause to the query
.LIMIT(limit) Add a LIMIT clause to the query

Sample SQL selects:

JavaScript SQL
`SELECT(["id", "name"]).FROM("users")` `SELECT "id", "name" FROM "users"`
`SELECT().FIELD("id").FIELD("name").FROM("users")` `SELECT "id", "name" FROM "users"`
`SELECT([COL("id").AS("userId"), "name"]).FROM("users")` `SELECT "id" AS "userId", "name" FROM "users"`
`SELECT(["a", "b", "c"]).DISTINCT().FROM("x").WHERE("a", "<>", 42)` `SELECT DISTINCT "a", "b", "c" FROM "x" WHERE "a" <> 42`
`SELECT(["a", "b", "c"]).DISTINCT("a").FROM("x").WHERE("a", "<>", 42)` `SELECT DISTINCT ON ("a") "a", "b", "c" FROM "x" WHERE "a" <> 42`

Complex SQL selects are possible by combining various SQL expressions together:

var query = SELECT()
  .FIELD("name")
  .FIELD(
     SELECT(MAX(COL("pop")))
      .FROM("cities")
      .WHERE(COL("cities.state"), "=", COL("states.name"))
      .AS("population"))
  .FROM("states");

yields to:

SELECT
  "name",
  (SELECT MAX("pop") FROM "cities" WHERE "cities"."state" = "states"."name")
FROM
  "states";

INSERT

Insert query is described by xql.node.InsertStatement node and wrapped by xql.INSERT(...). Note that INSERT(...) accepts parameters that can describe a target table and data to be inserted.

The xql.node.InsertStatement implements the following interface:

xql.node.InsertStatement Description
.TABLE(table)
.INTO(table) Specify a target table
.VALUES(data) Specify a data to be inserted. The data argument can be both array or object. If an array is passed each element describes one row (it has to be array of objects), of an object is passed, it describes only one row. If VALUES() is called multiple times it pushes more rows to be inserted by the query
.RETURNING(...) Specify a RETURNING clause, uses the same syntax as SELECT()

Sample SQL insert:

// INSERT("tasks", {...}).RETURNING(...) would also work.
var query = INSERT()
  .INTO("tasks")
  .VALUES({
    title: "Try xql.js",
    duration: 5
  })
  .RETURNING("id");

yields to:

INSERT INTO
  "tasks" ("title", "duration")
VALUES
  ('Try xql.js', 5)
RETURNING
  "id";

UPDATE

Update query is described by xql.node.UpdateStatement node and wrapped by xql.UPDATE(...). Please note that UPDATE(...) accepts parameters that can describe a target table and data to be updated.

The xql.node.UpdateStatement implements the following interface:

xql.node.UpdateStatement Description
.TABLE(table) Specify a target table
.FROM(...) Specify a FROM clause, uses the same syntax as FROM() defined by SELECT query
.WHERE(node)
.WHERE(a, b)
.WHERE(a, op, b) Add a WHERE clause node, WHERE a = b, or WHERE a op b to the query (implicitly ANDed with other WHERE clauses if present)
.OR_WHERE(node)
.OR_WHERE(a, b)
.OR_WHERE(a, op, b) Add a WHERE clause node, WHERE a = b, or WHERE a op b to the query (implicitly ORed with other WHERE clauses if present)
.RETURNING(...) Specify a RETURNING clause, uses the same syntax as FIELD() defined by SELECT query

Sample SQL update:

var query = UPDATE("users")
  .VALUES({
    address: "Friedrichstrasse 50, Berlin",
    addressChanged: OP(COL("addressChanged"), "+", 1)
  })
  .WHERE("userId", "=", 1);

yields to:

UPDATE
  "users"
SET
  "address" = 'Friedrichstrasse 50, Berlin',
  "addressChanged" = "addressChanged" + 1
WHERE
  "userId" = 1;

DELETE

Delete query is described by xql.node.DeleteStatement node and wrapped by xql.DELETE(...).

The xql.node.DeleteStatement implements the following interface:

xql.node.DeleteStatement Description
.TABLE(table)
.FROM(table) Specify a target table
.USING(...) Specify a USING clause, uses the same syntax as FROM() defined by SELECT query
.WHERE(node)
.WHERE(a, b)
.WHERE(a, op, b) Add a WHERE clause node, WHERE a = b, or WHERE a op b to the query (implicitly ANDed with other WHERE clauses if present)
.OR_WHERE(node)
.OR_WHERE(a, b)
.OR_WHERE(a, op, b) Add a WHERE clause node, WHERE a = b, or WHERE a op b to the query (implicitly ORed with other WHERE clauses if present)
.RETURNING(...) Specify a RETURNING clause, uses the same syntax as FIELD() defined by SELECT query.

Sample SQL delete:

var query = DELETE().FROM("tasks").WHERE("completed", "=", true)

yields to:

DELETE FROM "tasks" WHERE "completed" = TRUE;

Type Mapping

xql.js has a feature called TypeMapping, which allows to override a default serialization of data used by INSERT and UPDATE. The type mapping is an object where a key/value defines a column/data-type pair. It can be set by setTypeMapping() and get by getTypeMapping() methods of the query object.

The following example illustrates how type mapping may affect data serialization:

var typeMapping = {
  tagsArray: "ARRAY",
  tagsJson : "JSON"
};

var query = UPDATE("users")
  .VALUES({
    tagsArray : ["accounting", "customer support"],
    tagsJson  : ["accounting", "customer support"]
  })
  .WHERE("userId", "=", 1)
  .setTypeMapping(typeMapping);
UPDATE
  "users"
SET
  "tagsArray" = ARRAY['accounting', 'customer support'],  -- Using PG ARRAY syntax.
  "tagsJson" = '["accounting", "customer support"]'::json -- Using PG JSON syntax.
WHERE
  "userId" = 1;

More Examples

There is a project called xql-fiddle, which can be used to explore xql.js possibilities by playing with it online. It contains more snippets and tries to teach by examples.