Skip to content

ulule/makroud

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Makroud

CircleCI Documentation License

A high level SQL Connector.

Introduction

Makroud is a high level SQL Connector that only support PostgreSQL at the moment.

It's an advanced mapper and/or a lightweight ORM that relies on reflection to generate queries. Using reflection has its flaws, type safety is not guaranteed and a panic is still possible, even if you are very careful and vigilant. However, development is super easy and straightforward since it doesn't rely on code generation.

Makroud isn't a migration tool and it doesn't inspect the database to define the application data model (since there is no code generation). It's really important to have Active Record that are synchronized with your data model in your database.

It also support simple associations (one, many) with preloading.

Under the hood, it relies on Loukoum for query generation. In addition, it's heavily inspired by Sqlx for its extended mapper and Sqalx to support nested transaction.

Installation

Using Go Modules

go get github.com/ulule/[email protected]

Usage

Create a Driver

A Driver is a high level abstraction of a database connection or a transaction. It's almost required everytime alongside a context.Context to manipulate rows.

driver, err := makroud.New(
	makroud.Host(cfg.Host),
	makroud.Port(cfg.Port),
	makroud.User(cfg.User),
	makroud.Password(cfg.Password),
	makroud.Database(cfg.Name),
	makroud.SSLMode(cfg.SSLMode),
	makroud.MaxOpenConnections(cfg.MaxOpenConnections),
	makroud.MaxIdleConnections(cfg.MaxIdleConnections),
)

Also, you can use a struct directly if you don't need to use functional options:

driver, err := makroud.NewWithOptions(&makroud.ClientOptions{
	Host:               cfg.Host,
	Port:               cfg.Port,
	User:               cfg.User,
	Password:           cfg.Password,
	Database:           cfg.Name,
	SSLMode:            cfg.SSLMode,
	MaxOpenConnections: cfg.MaxOpenConnections,
	MaxIdleConnections: cfg.MaxIdleConnections,
})

Advanced mapper

If a lightweight ORM doesn't fit your requirements and an advanced mapper is enough for your usecase: makroud does that very well.

You could either use primitive and compound types for your queries:

import "github.com/ulule/makroud"

stmt := `SELECT id FROM users WHERE email = '[email protected]'`
id := int64(0)

err := makroud.RawExec(ctx, driver, stmt, &id)
if err != nil {
	return err
}

stmt = `SELECT created_at FROM users WHERE id = 42`
created := time.Time{}

err := makroud.RawExec(ctx, driver, stmt, &created)
if err != nil {
	return err
}

stmt = `SELECT email FROM users WHERE id IN (1, 2, 3, 4)`
list := []string{}

err := makroud.RawExec(ctx, driver, stmt, &list)
if err != nil {
	return err
}

Or, define a struct that contains your database table (or view) columns:

import "github.com/ulule/makroud"
import "github.com/ulule/loukoum/v3"

type User struct {
	ID       int64  `mk:"id"`
	Email    string `mk:"email"`
	Password string `mk:"password"`
	Country  string `mk:"country"`
	Locale   string `mk:"locale"`
}

users := []User{}
stmt := loukoum.Select("*").
	From("users").
	Where(loukoum.Condition("id").In(1, 2, 3, 4))

err := makroud.Exec(ctx, driver, stmt, &users)
if err != nil {
	return err
}

Lightweight ORM

Define a Model

With the Active Record approach, you have to define a model that wraps your database table (or view) columns into properties.

Models are structs that contain basic go types, pointers, sql.Scanner, driver.Valuer or Model interface. All the fields of this struct will be columns in the database table.

An example

type User struct {
	// Columns
	ID        string `makroud:"column:id,pk:ulid"`
	Email     string `makroud:"column:email"`
	Password  string `makroud:"column:password"`
	Country   string `makroud:"column:country"`
	Locale    string `makroud:"column:locale"`
	ProfileID string `makroud:"column:profile_id,fk:profiles"`
	// Relationships
	Roles    []Role    `makroud:"relation:roles.user_id"`
	Profile  *Profile  `makroud:"relation:profile_id"`
}

func (User) TableName() string {
	return "users"
}

What does it means ?

First of all, you have to define a TableName method that returns the database table name (or view). Without that information, makroud cannot uses that struct as a Model.

Then, you have to define your model columns using struct tags:

  • column(string): Define column name.
  • pk(bool|string): Define column as a primary key, it accepts the following argument:
    • true: Uses internal db mechanism to define primary key value
    • ulid: Generate a ULID to define primary key value
    • uuid-v1: Generate a UUID V1 to define primary key value
    • uuid-v4: Generate a UUID V4 to define primary key value
  • default(bool): On insert, if model has a zero value, it will use the db default value.
  • fk(string): Define column as a foreign key, reference table must be provided.
  • relation(string): Define which column to use for preload. The column must be prefixed by the table name if it's not the model table name (However, the prefix is optional if the table name is the same as the model). See Preload section for further information.
  • -(bool): Ignore this field.

NOTE: Tags of type bool can be set as key:true or just key for implicit true.

NOTE: Tags must be separated by a comma (tagA, tagB, tagC).

Keep in mind that a model requires one primary key (and just one). It's a known limitation that only one primary key can be specified and it can't be a composite key.

After that, you can define optional relationships (or associations) that can be preloaded later. The preload mechanism, which enables you to fetch relationships from database, support these types:

  • Model
  • *Model
  • []Model
  • []*Model
  • *[]Model
  • *[]*Model

NOTE: You could either use makroud or mk as tag identifier.

Conventions

ID as Primary Key

By default, if the pk tag is undefined, makroud will use the field named ID as primary key with this configuration: pk:db

type User struct {
	ID   string `makroud:"column:id"`   // Field named ID will be used as a primary key by default.
	Name string `makroud:"column:name"`
}
Snake Case Column Name

By default, if the column tag is undefined, makroud will transform field name to lower snake case as column name.

type User struct {
	ID   string `makroud:"pk"` // Column name is `id`
	Name string `makroud:""`   // Column name is `name`
}
Preload relationships

By default, if the relation tag is undefined, makroud will infer the column name to use for the preload mechanism.

Local foreign key:

Let's define a user with a profile:

type User struct {
	ID       string  `makroud:"column:id,pk"`
	Email    string  `makroud:"column:email"`
	PID      string  `makroud:"column:profile_id,fk:profiles"`
	Profile  *Profile
}

func (User) TableName() string {
	return "users"
}

type Profile struct {
	ID         string  `makroud:"column:id,pk:ulid"`
	FirstName  string  `makroud:"column:first_name"`
	LastName   string  `makroud:"column:last_name"`
	Enabled    bool    `makroud:"column:enabled"`
}

func (Profile) TableName() string {
	return "profiles"
}

Since the field Profile in the User has no relation tag, makroud will try to find, in the first pass, the field with the name ProfileID (FieldName + ID) in User model. It's mandatory that this field is also a foreign key to the profiles table.

Unfortunately for us, User model has no such field. So, makroud will try to find, in the second and final pass, the first field that is a foreign key to the profiles table. In our example, it will use the field PID.

Remote foreign key:

Let's define a user with a profile:

type User struct {
	ID       string  `makroud:"column:id,pk"`
	Email    string  `makroud:"column:email"`
	Profile  *Profile
}

func (User) TableName() string {
	return "users"
}

type Profile struct {
	ID         string  `makroud:"column:id,pk:ulid"`
	FirstName  string  `makroud:"column:first_name"`
	LastName   string  `makroud:"column:last_name"`
	Enabled    bool    `makroud:"column:enabled"`
	UID        string  `makroud:"column:user_id,fk:users"`
}

func (Profile) TableName() string {
	return "profiles"
}

Since the field Profile in the User has no relation tag, makroud will try to find, in the first pass, the field with the name UserID (ModelName + ID) in Profile model. It's mandatory that this field is also a foreign key to the users table.

Unfortunately for us, Profile model has no such field. So, makroud will try to find, in the second and final pass, the first field that is a foreign key to the users table. In our example, it will use the field UID.

CreatedAt tracking

For models having a CreatedAt field, it will be set to current time when the record is first created.

type User struct {
	ID        string    `makroud:"column:id,pk"`
	Name      string    `makroud:"column:name"`
	CreatedAt time.Time `makroud:"column:created_at"`
}

You can override the default field name and/or column name by adding this method:

func (User) CreatedKey() string {
	return "created"
}
UpdatedAt tracking

For models having a UpdatedAt field, it will be set to current time when records are updated.

type User struct {
	ID        string    `makroud:"column:id,pk"`
	Name      string    `makroud:"column:name"`
	UpdatedAt time.Time `makroud:"column:updated_at"`
}

You can override the default field name and/or column name by adding this method:

func (User) UpdatedKey() string {
	return "updated"
}
DeletedAt tracking

For models having a DeletedAt field, it will be set to current time when records are archived.

type User struct {
	ID        string      `makroud:"column:id,pk"`
	Name      string      `makroud:"column:name"`
	DeletedAt pq.NullTime `makroud:"column:deleted_at"`
}

You can override the default field name and/or column name by adding this method:

func (User) DeletedKey() string {
	return "deleted"
}

Operations

For the following sections, we assume that you have a context.Context and a makroud.Driver instance.

Insert

For a simple insert, you can save a model like this:

func CreateUser(ctx context.Context, driver makroud.Driver, name string) (*User, error) {
	user := &User{
		Name: name,
	}

	err := makroud.Save(ctx, driver, user)
	if err != nil {
		return nil, err
	}

	return user, nil
}

Or for more complex statements, use a Loukoum InsertBuilder alongside the model.

import "github.com/ulule/loukoum/v3"

func CreateUser(ctx context.Context, driver makroud.Driver, name string) (*User, error) {
	user := &User{
		Name: name,
	}

	stmt := loukoum.Insert("users").
		Set(loukoum.Pair("name", user.Name)).
		Returning("id, created_at, updated_at")

	err := makroud.Exec(ctx, driver, stmt, user)
	if err != nil {
		return nil, err
	}

	return user, nil
}

Update

For a simple update, asumming your model have a primary key defined, you can save it by executing:

func UpdateUser(ctx context.Context, driver makroud.Driver, user *User, name string) error {
	user.Name = name
	return makroud.Save(ctx, driver, user)
}

Or for more complex statements, use a Loukoum UpdateBuilder alongside the model.

import "github.com/ulule/loukoum/v3"

func UpdateUser(ctx context.Context, driver makroud.Driver, user *User, name string) error {
	user.Name = name

	stmt := loukoum.Update("users").
		Set(
			loukoum.Pair("updated_at", loukoum.Raw("NOW()")),
			loukoum.Pair("name", user.Name),
		).
		Where(loukoum.Condition("id").Equal(user.ID)).
		Returning("updated_at")

	err := makroud.Exec(ctx, driver, stmt, user)
	if err != nil {
		return nil, err
	}

	return user, nil
}

Delete

For a simple delete (using a DELETE statement), asumming your model have a primary key defined, you can delete it using:

func DeleteUser(ctx context.Context, driver makroud.Driver, user *User) error {
	return makroud.Delete(ctx, driver, user)
}

Or for more complex statements, use a Loukoum DeleteBuilder alongside the model.

import "github.com/ulule/loukoum/v3"

func DeleteUser(ctx context.Context, driver makroud.Driver, user *User) error {
	stmt := loukoum.Delete("users").Where(loukoum.Condition("id").Equal(user.ID))

	err := makroud.Exec(ctx, driver, stmt, user)
	if err != nil {
		return nil, err
	}

	return user, nil
}

Archive

Archive executes an UPDATE on DeletedAt field on given value.

func ArchiveUser(ctx context.Context, driver makroud.Driver, user *User) error {
	return makroud.Archive(ctx, driver, user)
}

NOTE: If the model has no DeletedAt field, an error is returned.

Or for more complex statements, use a Loukoum UpdateBuilder alongside the model.

import "github.com/ulule/loukoum/v3"

func ArchiveUser(ctx context.Context, driver makroud.Driver, user *User) error {
	user.Name = name

	stmt := loukoum.Update("users").
		Set(
			loukoum.Pair("deleted_at", loukoum.Raw("NOW()")),
			loukoum.Pair("name", ""),
		).
		Where(loukoum.Condition("id").Equal(user.ID)).
		Returning("deleted_at")

	err := makroud.Exec(ctx, driver, stmt, user)
	if err != nil {
		return nil, err
	}

	return user, nil
}

Query

Because querying data is a bit more complex than just writing and/or deleting stuff. By using Loukoum components, you can either execute simple query:

import "github.com/ulule/loukoum/v3"

func GetUserByID(ctx context.Context, driver makroud.Driver, id string) (*User, error) {
	user := &User{}
	err := makroud.Select(ctx, driver, user, loukoum.Condition("id").Equal(id))
	if err != nil {
		return nil, err
	}

	return user, nil
}

func ListMessagesByUserID(ctx context.Context, driver makroud.Driver,
	userID string, page int) ([]*Message, error) {

	messages := []*Message{}
	err := makroud.Select(ctx, driver, &messages,
		loukoum.Condition("user_id").Equal(id),
		loukoum.Order("created_at", loukoum.Desc),
		loukoum.Limit(50),
		loukoum.Offset(50 * (page - 1)),
	)
	if err != nil {
		return nil, err
	}

	return messages, nil
}

Or execute more complex statements:

import "github.com/ulule/loukoum/v3"

func FindStaffComments(ctx context.Context, driver makroud.Driver) ([]*Comment, error) {
	comments := []*Comment{}

	stmt := loukoum.Select("id", "email", "status", "user_id", "message", "created_at").
		From("comments").
		Where(loukoum.Condition("deleted_at").IsNull(true)).
		Where(
			loukoum.Condition("user_id").In(
				loukoum.Select("id").
					From("users").
					Where(loukoum.Condition("role").Equal("staff")),
			),
		)

	err := makroud.Exec(ctx, driver, stmt, &comments)
	if err != nil {
		return nil, err
	}

	return comments, nil
}

Also, it supports query without Model.

func FindUserIDWithStaffRole(ctx context.Context, driver makroud.Driver) ([]string, error) {
	list := []string{}

	stmt := loukoum.Select("id").
		From("users").
		Where(loukoum.Condition("role").Equal("staff"))

	err := makroud.Exec(ctx, driver, stmt, &list)
	if err != nil {
		return nil, err
	}

	return list, nil
}

Transaction

Sometimes, you need to execute queries and/or commands inside a transaction block, that bundles multiple steps into a single, all-or-nothing operation.

This is achieved by declaring a lambda function. If this function returns an error, the transaction rollbacks automatically. Otherwise, the transaction will be committed.

func SetupUsers(ctx context.Context, driver makroud.Driver) error {
	return makroud.Transaction(ctx, driver, nil, func(tx makroud.Driver) error {

		err := makroud.Save(ctx, tx, &User{
			Name: "Benjamin",
		})
		if err != nil {
			return err
		}

		err = makroud.Save(ctx, tx, &User{
			Name: "Taha",
		})
		if err != nil {
			return err
		}

		return nil
	})
}

And other times, transactions with an isolation level.

func Withdraw(ctx context.Context, driver makroud.Driver) error {
	return makroud.Transaction(ctx, driver, makroud.LevelSerializable,
		func(tx makroud.Driver) error {
			// Withdraw operation...
			return nil
		},
	)
}

Or even, nested transaction with the option SavepointEnabled.

func AcceptOffer(ctx context.Context, driver makroud.Driver) error {
	return makroud.Transaction(ctx, driver, nil, func(tx1 makroud.Driver) error {
		//
		// Execute several operations.
		//
		err := makroud.Transaction(ctx, tx1, nil, func(tx2 makroud.Driver) error {
			//
			// Execute complex operations that may succeed...
			//
			return err
		})
		if err != nil {
			//
			// Execute fallback operations if an error has occurred...
			//
			return nil
		}
		//
		// Execute normal operations otherwise...
		//
		return nil
	})
}

Preload

On models having associations, you can execute a preload to fetch these relationships from the database.

Let's define a user with a profile:

type User struct {
	ID       string   `makroud:"column:id,pk"`
	Email    string   `makroud:"column:email"`
	Profile  *Profile `makroud:"relation:profiles.user_id"`
}

func (User) TableName() string {
	return "users"
}

type Profile struct {
	ID         string  `makroud:"column:id,pk:ulid"`
	FirstName  string  `makroud:"column:first_name"`
	LastName   string  `makroud:"column:last_name"`
	UserID     string  `makroud:"column:user_id,fk:users"`
	Enabled    bool    `makroud:"column:enabled"`
}

func (Profile) TableName() string {
	return "profiles"
}

Once you obtain a user record, you can preload its profile by executing:

err := makroud.Preload(ctx, driver, &user, makroud.WithPreloadField("Profile"))

Or, if preloading requires specific conditions, you can use a callback like this:

import "github.com/ulule/loukoum/v3/builder"

err := makroud.Preload(ctx, driver, &user,
	makroud.WithPreloadCallback("Profile", func(query builder.Select) builder.Select {
		return query.Where(loukoum.Condition("enabled").Equal(true))
	}),
)

If there is no error and if the user record has a profile, then you should have the Profile value loaded.

Development

Docker

The test suite is running on PostgreSQL. We use Docker to create a running container using scripts/database.

Testing

To run the test suite, simply execute:

scripts/test

Also, you can execute the linter with:

scripts/lint

Notes

If you have to examine rows generated from unit test, you can prevent the test suite to cleanup by using:

DB_KEEP=true scripts/test

Then, you can access the database with:

scripts/database --client

Random

Because sometimes it's hard to think of a good test fixture, using generators can save your productivity.

This website was a great help to write unit test: http://www.fantasynamegenerators.com

License

This is Free Software, released under the MIT License.

Contributing

Don't hesitate ;)