• Synopsis
• Support
• Installation
• Configuration
• Usage
• API Reference
• Contributors
• Stack

Synopsis

Antity-pgsql.js adds PostgreSQL features to Antity.js library.

• 🪶 Very lightweight
• 🧪 Thoroughly tested
• 🚚 Shipped as EcmaScrypt module
• 📝 Written in Typescript

Installation

$ npm i @dwtechs/antity-pgsql

Configuration

Antity-pgsql reads its PostgreSQL connection details from the process environment. Set these variables in your service's environment before opening any connection:

Variable	Required	Default	Description
DB_HOST	yes	—	PostgreSQL server hostname
DB_USER	yes	—	PostgreSQL user
DB_PWD	yes	—	PostgreSQL user password
DB_NAME	yes	—	PostgreSQL database name
DB_PORT	no	5432	PostgreSQL server port
DB_MAX	no	10	Maximum pool connections

Connection Pool

Since 0.22.0, the underlying pg-pool client is initialized lazily — it is opened on the first call to execute() or SQLEntity.query.sync(), not at module import.

• Consumers that only use the query-builder surface (SQLEntity.query.select, filter) without ever executing a query never open a socket.
• Combined with "sideEffects": false, the library is fully tree-shakable and safe to import at boot without side effects.
• Boot sequences that call execute() inside Promise.all([...init()]) will surface connection failures through that promise. Pair it with @dwtechs/servpico-express's failFast helper for a clean exit.

Usage

import{SQLEntity}from"@dwtechs/antity-pgsql";import{normalizeName,normalizeNickname}from"@dwtechs/checkard";// Create entity with default 'public' schemaconstentity=newSQLEntity("consumers",[// properties...]);// Or specify a custom schemaconstcustomEntity=newSQLEntity("consumers",[// properties...],"myschema");// Example with all propertiesconstentity=newSQLEntity("consumers",[{key: "id",type: "integer",min: 0,max: 120,isTypeChecked: true,isFilterable: true,requiredFor: ["PUT"],operations: ["SELECT","UPDATE"],isPrivate: false,sanitizer: null,normalizer: null,validator: null,},{key: "firstName",type: "string",min: 0,max: 255,isTypeChecked: true,isFilterable: false,requiredFor: ["POST","PUT"],operations: ["SELECT","UPDATE"],isPrivate: false,sanitizer: null,normalizer: normalizeName,validator: null,},{key: "lastName",type: "string",min: 0,max: 255,isTypeChecked: true,isFilterable: false,requiredFor: ["POST","PUT"],operations: ["SELECT","UPDATE"],isPrivate: false,sanitizer: null,normalizer: normalizeName,validator: null,},{key: "nickname",type: "string",min: 0,max: 255,isTypeChecked: true,isFilterable: true,requiredFor: ["POST","PUT"],operations: ["SELECT","UPDATE"],isPrivate: false,sanitizer: null,normalizer: normalizeNickname,validator: null,},]);router.get("/", ...,entity.get);// Using substacks (recommended) - combines normalize, validate, and database operationrouter.post("/", ...entity.addArraySubstack);router.put("/", ...entity.updateArraySubstack);router.put("/preferences", ...entity.syncArraySubstack);// Or manually chain middlewaresrouter.post("/manual",entity.normalizeArray,entity.validateArray, ...,entity.add);router.put("/manual",entity.normalizeArray,entity.validateArray, ...,entity.update);router.patch("/archive", ...,entity.archive);router.delete("/", ...,entity.delete);router.delete("/archived", ...,entity.deleteArchive);router.get("/:id/history", ...,entity.getHistory);

Expected table structure

CREATETABLEIF NOT EXISTS "service" (
id SERIALPRIMARY KEY,
name varchar(20) NOT NULL,
pattern TEXT,
archived BOOLEAN DEFAULT FALSE,
"archivedAt"TIMESTAMP,
"creatorId"INT,
"creatorName"TEXT,
"updaterId"INT,
"updaterName"TEXT,
"createdAt"TIMESTAMP DEFAULT NOW(),
"updatedAt"TIMESTAMPNULL
);

API Reference

typeOperation="SELECT"|"INSERT"|"UPDATE";typeRow=Record<string,string|number|boolean|Date|number[]>;typeComparator="="|"<"|">"|"<="|">="|"<>"|"IS"|"IS NOT"|"IN"|"NOT IN"|"LIKE"|"NOT LIKE"|"&&";typeMatchMode="startsWith"|"endsWith"|"contains"|"notContains"|"equals"|"notEquals"|"!="|"between"|"in"|"notIn"|"&&"|// array overlap — use with array-typed columns; generates: column && ARRAY[$1,$2]"lt"|"lte"|"gt"|"gte"|"is"|"isNot"|"before"|"after"|"st_contains"|"st_dwithin"|Comparator;// direct SQL comparators are also acceptedtypeFilters={[key: string]: Filter|Filter[];// Supports both simple (object) and complex (array) formats}// Any scalar value that can be bound as a query parameter or stored in a row/column.typeSqlValue=string|number|boolean|Date|number[]|null;typeFilter={value: SqlValue;matchMode?: MatchMode;// semantic mode or direct SQL comparatoroperator?: string;// 'and' | 'or' - Used when multiple filters apply to the same property}typePGClient={query(text: string,values?: unknown[]): Promise<PGResponse>;};typePGResponse={rows: Record<string, unknown>[];rowCount: number|null;total?: number;};typeSelectResponse={rows: Record<string, unknown>[];total?: number;};typeExpressMiddleware=(req: Request,res: Response,next: NextFunction)=>void;typeExpressMiddlewareAsync=(req: Request,res: Response,next: NextFunction)=>Promise<void>;typeSubstackTuple=[ExpressMiddleware,ExpressMiddleware,ExpressMiddlewareAsync];classSQLEntity{constructor(name: string,properties: Property[],schema?: string);getname(): string;gettable(): string;getschema(): string;getprivateProps(): string[];getproperties(): Property[];setname(name: string);settable(table: string);setschema(schema: string);// Middleware substacks (combine normalize, validate, and operation)getaddArraySubstack(): SubstackTuple;getaddOneSubstack(): SubstackTuple;getupdateArraySubstack(): SubstackTuple;getupdateOneSubstack(): SubstackTuple;getupsertArraySubstack(): SubstackTuple;getupsertOneSubstack(): SubstackTuple;getsyncArraySubstack(): SubstackTuple;query: {select: (first?: number,limit?: number|null,sortField?: string|null,sortOrder?: "ASC"|"DESC"|null,filters?: Filters|null)=>{query: string;args: SqlValue[];},operator?: LogicalOperator;update: (rows: Row[],consumer?: {userId?: number|string,nickname?: string})=>{query: string;args: unknown[];};archive: (rows: Row[],consumer?: {userId?: number|string,nickname?: string})=>{query: string;args: unknown[];};insert: (rows: Row[],consumer?: {userId?: number|string,nickname?: string},rtn?: string)=>{query: string;args: unknown[];};upsert: (rows: Row[],conflictTarget: string|string[],consumer?: {userId?: number|string,nickname?: string},rtn?: string)=>{query: string;args: unknown[];};delete: (ids: number[])=>{query: string;args: number[];};deleteArchive: ()=>string;return: (prop: string)=>string;};get: (req: Request,res: Response,next: NextFunction)=>void;add: (req: Request,res: Response,next: NextFunction)=>Promise<void>;update: (req: Request,res: Response,next: NextFunction)=>Promise<void>;upsert: (req: Request,res: Response,next: NextFunction)=>Promise<void>;sync: (req: Request,res: Response,next: NextFunction)=>Promise<void>;archive: (req: Request,res: Response,next: NextFunction)=>Promise<void>;delete: (req: Request,res: Response,next: NextFunction)=>Promise<void>;deleteArchive: (req: Request,res: Response,next: NextFunction)=>void;getHistory: (req: Request,res: Response,next: NextFunction)=>Promise<void>;}functionfilter(first: number,rows: number|null,sortField: string|null,sortOrder: Sort|null,filters: Filters|null,operator?: LogicalOperator,): {filterClause: string,args: SqlValue[]};functionexecute(query: string,args: SqlValue[],client: PGClient|null,): Promise<PGResponse>;

Middleware Methods for Express.js

• get(), add(), update(), upsert(), sync(), archive(), delete(), deleteArchive() and getHistory() methods are made to be used as Express.js middlewares.
• add(), update() and upsert() accept either req.body.rows (as an array of entities for bulk operations) or req.body itself (as a single entity).
• archive() and sync() look for data to work on exclusively in the req.body.rows parameter (as an array).
• get() reads req.body.first, req.body.limit, req.body.sortField, req.body.sortOrder, req.body.filters and req.body.operator instead. Page size is exclusively req.body.limit (a number); req.body.rows is intentionally ignored by get(), since that same key means "array of entities" for every other method above and reusing it for pagination was error-prone.
• delete() reads req.body.rows ([{id: 1}, {id: 2}]) if present, otherwise falls back to a single req.params.id.- The upsert() method additionally requires req.body.conflictTarget to specify which column(s) define uniqueness.
• The sync() method accepts an optional req.body.idField (defaults to 'id') and optional req.body.filters to scope which existing rows are considered part of the managed set.

Schema Qualification

All SQL queries generated by Antity-pgsql use schema-qualified table names (e.g., schema.table). This provides:

• Security: Protection against search_path manipulation attacks, especially important when using SECURITY DEFINER functions
• Clarity: Explicit schema references make queries more readable and maintainable
• Flexibility: Easy multi-schema support within the same application

The default schema is 'public' but can be customized via the constructor's third parameter or the schema setter.

Middleware Substacks

Substacks are pre-composed middleware chains that combine normalization, validation, and database operations:

• addArraySubstack: Combines normalizeArray, validateArray, and add. Use this for POST routes with req.body.rows containing multiple objects.
• addOneSubstack: Combines normalizeOne, validateOne, and add. Use this for POST routes with req.body containing a single object.
• updateArraySubstack: Combines normalizeArray, validateArray, and update. Use this for PUT routes with req.body.rows containing multiple objects.
• updateOneSubstack: Combines normalizeOne, validateOne, and update. Use this for PUT routes with req.body containing a single object.
• upsertArraySubstack: Combines normalizeArray, validateArray, and upsert. Use this for upsert routes with req.body.rows containing multiple objects. Requires req.body.conflictTarget.
• upsertOneSubstack: Combines normalizeOne, validateOne, and upsert. Use this for upsert routes with req.body containing a single object. Requires req.body.conflictTarget.
• syncArraySubstack: Combines normalizeArray, validateArray, and sync. Use this for bulk-sync routes with req.body.rows containing the full desired state. Rows are inserted, updated, or deleted as needed. Accepts optional req.body.idField and req.body.filters.

Using substacks simplifies your route definitions and ensures consistent data processing.

Query Methods

• query.select(): Generates a SELECT query. When the limit parameter is provided (not null), pagination is automatically enabled and the query includes COUNT(*) OVER () AS total to return the total number of rows. The total count is extracted from results and returned separately from the row data. The sortField parameter is validated against the entity's known properties; an unrecognised value is silently dropped.
• getCache(): Loads active rows for an in-memory cache warm-up. Always ANDs archived IS FALSE (a caller-supplied archived filter is overwritten). Takes optional extra filters and a client. Returns a Promise of rows with no LIMIT, ordered by id ASC. An empty result resolves to [] — unlike get(), which forwards a 404 through next. Requires a filterable archived property on the entity.
• query.insert(): Generates an INSERT query. Accepts an array of Row objects with properties matching the entity definition. Consumer fields are appended directly to the query arguments — row objects are not mutated. Optionally appends consumer.userId as creatorId and consumer.nickname as creatorName for audit tracking. Supports RETURNING clause via the rtn parameter.
• query.update(): Generates an UPDATE query using CASE statements. Accepts an array of Row objects with id property. Optionally appends consumer.userId as updaterId and consumer.nickname as updaterName for audit tracking.
• query.upsert(): Generates an INSERT ... ON CONFLICT ... DO UPDATE query. (See Upsert section below.) Accepts an array of Row objects and a conflictTarget (single column name or array of column names) that defines uniqueness. If a conflict occurs on the specified column(s), the row is updated; otherwise, it is inserted. Properties are automatically included if they have both INSERT and UPDATE operations. Consumer fields are appended directly to the query arguments — row objects are not mutated. Optionally appends consumer.userId as creatorId and consumer.nickname as creatorName on INSERT, and as updaterId/updaterName on CONFLICT UPDATE, for audit tracking. Supports RETURNING clause via the rtn parameter.
• query.archive(): Generates a UPDATE ... SET archived = true WHERE id IN (...) query. Accepts an array of ids, or of Row objects with an id property. Optionally appends consumer.userId as updaterId and consumer.nickname as updaterName for audit tracking. Does not require an archived field in the rows — it is set directly in the SQL.
• sync(): Atomically synchronises the table with the provided rows inside a single PostgreSQL transaction. Missing rows are inserted, existing rows are updated, and rows absent from the list are deleted. Accepts optional idField (default 'id') and filters to restrict the scope of managed rows. Stores the result in res.locals.rows and a summary { inserted, updated, deleted } in res.locals.sync.
• delete(): Deletes rows by their IDs. Reads ids from req.body.rows (array of objects with id property: [{id: 1}, {id: 2}]) if present, otherwise falls back to a single req.params.id (e.g. a DELETE /resource/:id route). Calls next({ status: 400, message: "Missing rows in req.body or id in req.params for delete operation" }) if neither is provided.
• deleteArchive(): Deletes archived rows that were archived before a specific date using a PostgreSQL SECURITY DEFINER function. Expects req.body.date to be a Date object.
• getHistory(): Retrieves modification history for rows from the log.history table. Expects req.body.rows to be an array of objects with id property. Returns all historical records for the specified entity IDs.

Bulk Sync

The sync functionality atomically replaces the managed set of rows in a table with the supplied list. It combines insert, update, and delete in a single PostgreSQL transaction — either all changes succeed or none do.

How It Works

1. Fetch existing IDs: A SELECT id FROM table is issued, optionally scoped by filters.
2. Diff: Incoming rows without an ID (or with an unknown ID) are inserted; rows with a known ID are updated; existing IDs absent from the incoming list are deleted — all within the same filter scope.
3. Transaction: All three operations run inside BEGIN / COMMIT. A failure at any step triggers ROLLBACK.
4. Result: res.locals.rows contains the full synced list (with generated IDs filled in for inserts). res.locals.sync contains { inserted, updated, deleted } counts.

Usage Examples

Using the middleware:

// Route definitionrouter.put('/users/sync', ...entity.syncArraySubstack);// Request body — send the entire desired state{rows: [{id: 1,name: 'John Updated',email: 'john@example.com',age: 31},// update{name: 'Jane New',email: 'jane@example.com',age: 25}// insert// id: 2 is absent → will be deleted],idField: 'id'// optional, defaults to 'id'}

Scoping with filters (only manage a subset of rows):

// Only sync rows where age >= 18 — rows outside this filter are left untouched{rows: [{id: 1,name: 'John',email: 'john@example.com',age: 30}],filters: {age: {value: 18,matchMode: 'gte'}}}

Response locals after sync:

res.locals.rows// full list of synced rows (inserts have their new id)res.locals.sync// { inserted: 1, updated: 1, deleted: 1 }

Important Notes

• Atomic: All insert / update / delete operations are wrapped in a single transaction.
• Filter scope: When filters are provided, only rows matching the filter are considered "managed". Rows outside the filter are never touched.
• Consumer tracking: consumer.userId and consumer.nickname from res.locals.consumer are forwarded to inserts as creatorId/creatorName and to updates as updaterId/updaterName for audit tracking.

Upsert (Insert or Update)

The upsert functionality uses PostgreSQL's INSERT ... ON CONFLICT ... DO UPDATE syntax to insert rows or update them if they already exist based on a unique constraint.

How It Works

1. Conflict Target: You specify which column(s) define uniqueness (e.g., 'id', 'email', or ['name', 'email'])
2. Property Selection: Properties are automatically included if they have bothINSERT and UPDATE in their operations array
3. On Conflict: When a conflict occurs, all columns except the conflict target are updated

Usage Examples

Using the middleware with a single conflict target:

// Route definitionrouter.post('/users/upsert', ...entity.upsertArraySubstack);// Request body{rows: [{id: 1,name: 'John Updated',email: 'john@example.com'},{name: 'Jane New',email: 'jane@example.com'}],conflictTarget: 'id'}

Using email as conflict target:

// If a user with this email exists, update their name; otherwise, insert{rows: [{name: 'John',email: 'john@example.com',age: 30}],conflictTarget: 'email'}

Using multiple columns as conflict target:

// Unique constraint on combination of name and email{rows: [{name: 'John',email: 'john@example.com',age: 30}],conflictTarget: ['name','email']}

Using the query generator directly:

const{ query, args }=entity.query.upsert([{id: 1,name: 'John',email: 'john@example.com'}],'id',{userId: 1,nickname: 'admin'},// consumer (optional)'RETURNING id'// return clause (optional));// Generates:// INSERT INTO public.users (name, email, "creatorId", "creatorName")// VALUES ($1, $2, $3, $4)// ON CONFLICT (id) DO UPDATE SET // name = EXCLUDED.name,// email = EXCLUDED.email,// "updaterId" = EXCLUDED."creatorId",// "updaterName" = EXCLUDED."creatorName"// RETURNING id

Property Configuration for Upsert

Properties are automatically included in upsert if they have both INSERT and UPDATE operations:

{key: 'name',operations: ['SELECT','INSERT','UPDATE']// Included in upsert}{key: 'id',operations: ['SELECT','UPDATE']// NOT included (no INSERT)}{key: 'createdAt',operations: ['SELECT','INSERT']// NOT included (no UPDATE)}

Important Notes

• Conflict Target Required: The conflictTarget parameter must specify an existing unique constraint or primary key
• Mixed Rows: You can upsert rows with and without IDs in the same request if your conflict target handles it (e.g., using SERIAL primary key)
• Atomic Operation: Unlike separate insert/update calls, upsert is a single atomic database operation
• Concurrent Safety: Prevents race conditions when multiple requests try to create the same record

Filters

Filters support two formats for maximum flexibility:

Simple Format (Single Filter per Property)

Backward-compatible format using a single filter object:

constfilters={name: {value: 'John',matchMode: 'contains'},age: {value: 30,matchMode: 'equals'},archived: {value: false,matchMode: 'equals'}};// Direct SQL comparators are also acceptedconstfilters={age: {value: 30,matchMode: '>='},status: {value: null,matchMode: 'IS NOT'}};

Complex Format (Multiple Filters per Property)

Array-based format supporting multiple filters with logical operators:

constfilters={// Multiple filters on the same property with OR operatorname: [{value: 'John',matchMode: 'contains',operator: 'or'},{value: 'Jane',matchMode: 'contains',operator: 'or'}],// Age range with AND operatorage: [{value: 18,matchMode: 'gte',operator: 'and'},{value: 65,matchMode: 'lte',operator: 'and'}],// Single filter in array formatarchived: [{value: false,matchMode: 'equals'}]};

This generates SQL like:

WHERE (name LIKE'%John%'OR name LIKE'%Jane%') AND (age >=18AND age <=65) AND archived = false

Top-level Logical Operator

By default, top-level filter properties are combined with AND. You can pass an optional operator argument to filter() or query.select() (or read from req.body.operator when using the get() middleware) to combine them with OR instead:

import{filter}from"@dwtechs/antity-pgsql";constfilters={name: [{value: 'John',matchMode: 'contains'}],age: [{value: 30,matchMode: 'equals'}]};constresult=filter(0,null,null,null,filters,"OR"// top-level logical operator);

This generates SQL like:

WHERE name LIKE'%John%'OR age =30

Mixing Property-Level & Top-Level Logical Operators

You can specify logical operators between conditions on the same property to build compound rules, and combine these top-level property filters with a different top-level logical operator (such as OR):

import{filter}from"@dwtechs/antity-pgsql";constfilters={name: [{value: 'John',matchMode: 'equals'}],nickname: [{value: null,matchMode: 'is',operator: 'or'},{value: 'John',matchMode: 'equals',operator: 'or'}]};constresult=filter(0,null,null,null,filters,"OR"// top-level logical operator combining 'name' and 'nickname' conditions);

This generates SQL like:

WHERE name = $1OR (nickname IS NULLOR nickname = $2)

Notes:is/isNotwithnull,trueorfalse: these are rendered as a SQL literal (col IS NULL, col IS NOT NULL, col IS TRUE, col IS NOT FALSE, etc.) rather than a bound parameter, since PostgreSQL's IS operator only accepts the NULL/TRUE/FALSE/UNKNOWN keywords. These filters don't consume a placeholder index or push a value into the returned args. Using is/isNot with any other value type (e.g. a string or number) still generates a bound parameter, unchanged.

Match modes

matchMode accepts either a semantic match mode (listed below) or a direct SQL comparator (=, <, >, <=, >=, <>, IS, IS NOT, IN, NOT IN, LIKE, NOT LIKE).

Using a direct comparator bypasses the semantic layer. Note that when using LIKE or NOT LIKE directly, wildcard characters (%) must be included manually in the value.

List of possible semantic match modes :

Name	alias	types	Description
startsWith		string	Whether the value starts with the filter value
contains		string	Whether the value contains the filter value
endsWith		string	Whether the value ends with the filter value
notContains		string	Whether the value does not contain filter value
equals		string	number
notEquals		string	number
!=	notEquals	string	number
in		string[]	number[]
notIn		string[]	number[]
lt		string	number
lte		string	number
gt		string	number
gte		string	number
is		date	boolean
isNot		date	boolean
before		date	Whether the date value is before the filter date
after		date	Whether the date value is after the filter date
dateIs	is	date	Alias of is for date fields
dateIsNot	isNot	date	Alias of isNot for date fields
dateBefore	before	date	Alias of before for date fields
dateAfter	after	date	Alias of after for date fields
between		date[2]	number[2]
st_contains		geometry	Whether the geometry completely contains other geometries
st_dwithin		geometry	Whether geometries are within a specified distance from another geometry

Types

List of compatible match modes for each property types.

Name	Match modes
string	startsWith, contains, endsWith, notContains, equals, notEquals, !=, in, notIn, lt, lte, gt, gte, is, isNot
number	equals, notEquals, !=, in, notIn, lt, lte, gt, gte, is, isNot
date	is, isNot, before, after, dateIs, dateIsNot, dateBefore, dateAfter
boolean	is, isNot
string[]	in
number[]	in, between
date[]	between
geometry	st_contains, st_dwithin

Note: All types support the semantic match modesis/isNotor direct comparatorsIS/IS NOTwhen querying fornullornot nullvalues.

List of secondary types :

Name	equivalent
integer	number
float	number
even	number
odd	number
positive	number
negative	number
powerOfTwo	number
ascii	number
array	any[]
jwt	string
symbol	string
email	string
password	string
regex	string
ipAddress	string
slug	string
hexadecimal	string
date	date
timestamp	date
function	string
htmlElement	string
htmlEventAttribute	string
node	string
json	object
object	object

Available options for a property

Any of these can be passed into the options object for each function.

Name	Type	Description	Default value
key	string	Name of the property
type	Type	Type of the property
min	number	Date	Minimum value
max	number	Date	Maximum value
requiredFor	Method[]	property is required for the listed methods only	["PATCH", "PUT", "POST"]
isPrivate	boolean	Property is unsafe to send in the response	true
isTypeChecked	boolean	Type is checked during validation	false
isFilterable	boolean	property is filterable in a SELECT operation	true
operations	Operation[]	Property is used for the DML operations only	["SELECT", "INSERT", "UPDATE"]
sanitizer	((v: unknown) => unknown)	null	Custom sanitizer function if sanitize is true
normalizer	((v: unknown) => unknown)	null	Custom Normalizer function if normalize is true
validator	((v: unknown) => unknown)	null	validator function if validate is true

• Min and max parameters are not used for boolean type
• TypeCheck Parameter is not used for boolean, string and array types

Support

Environment	Version
Node.js	>= 22

Contributors

Antity.js is still in development and we would be glad to get all the help you can provide. To contribute please read contributor.md for detailed installation guide.

Stack

Purpose	Choice	Motivation
repository	Github	hosting for software development version control using Git
package manager	npm	default node.js package manager
language	TypeScript	static type checking along with the latest ECMAScript features
module bundler	Rollup	advanced module bundler for ES6 modules
unit testing	Jest	delightful testing with a focus on simplicity