jet/wiki/Generator.md
2019-06-27 19:55:21 +02:00

9.2 KiB
Raw Blame History

Before we can write SQL queries in Go we have to generate necessary Go files. To generate files we need running database instance containing already defined database schema.

This files can be generated in two ways:

1) Generating from command line

Install jetgen to GOPATH bin folder. This will allow generating jet files from the command line.

go install github.com/go-jet/jet/cmd/jetgen

Make sure GOPATH bin folder is added to the PATH environment variable.

Test jetgen can be found in the PATH.

jetgen -h
Usage of jetgen:
  -host string
    	Database host path (Example: localhost)
  -port string
    	Database port
  -user string
        Database user
  -password string
    	The users password
  -dbname string
    	name of the database
  -schema string
    	Database schema name. (default "public")
  -path string
    	Destination dir for generated files.
  -sslmode string
    	Whether or not to use SSL(optional) (default "disable")
  -params string
    	Additional connection string parameters(optional)

Now to generate sample database:

jetgen -host=localhost -port=5432 -user=jet -password=jet -dbname=jetdb -schema dvds -path ./gen
Connecting to postgres database: host=localhost port=5432 user=jet password=jet dbname=jetdb sslmode=disable 
Retrieving schema information...
	FOUND 15  table(s),  1  enum(s)
Cleaning up destination directory...
Generating table sql builder files...
Generating table model files...
Generating enum sql builder files...
Generating enum model files...
Done

2) Generating from code

import "github.com/go-jet/jet/generator/postgresgen"

...

err = postgresgen.Generate("./gen", postgresgen.DBConnection{
    Host:       "localhost",
    Port:       "5432",
    User:       "jet",
    Password:   "jet",
    DBName:     "jetdb",
    SchemaName: "dvds",
    SslMode:    "disable",
})

In both ways, generator will:

  • connect to postgres database and retrieve information about the tables and enums of dvds schema
  • delete everything in destination folder ./gen,
  • generate sql builder and model Go files for each schema tables and enums into destination folder ./gen.

Generated files folder structure will look like this:

|-- gen                               # destination folder
|   `-- jetdb                         # database name
|       `-- dvds                      # schema name
|           |-- enum                  # sql builder folder for enums
|           |   |-- mpaa_rating.go
|           |-- table                 # sql builder folder for tables
|               |-- actor.go
|               |-- address.go
|               |-- film.go
                ...
|           |-- model                 # Plain Old Data for every enum and table
|           |   |-- actor.go
|           |   |-- address.go
|           |   |-- film.go
                ...

Table and enums from database schema are used as a template to generate two types of Go files:

  • SQL builder files: Files used to write type safe SQL statements in Go (enum and table package)
  • Model files: Files used to store result from database queries (model package)

SQL builder files

Part of the sample sql builder file for table film.

package table

import (
	"github.com/go-jet/jet"
)
var Film = newFilmTable()

type FilmTable struct {
	jet.Table

	//Columns
	FilmID          jet.ColumnInteger
	Title           jet.ColumnString
	Description     jet.ColumnString
	ReleaseYear     jet.ColumnInteger
	LanguageID      jet.ColumnInteger
	RentalDuration  jet.ColumnInteger
	RentalRate      jet.ColumnFloat
	Length          jet.ColumnInteger
	ReplacementCost jet.ColumnFloat
	Rating          jet.ColumnString
	LastUpdate      jet.ColumnTimestamp
	SpecialFeatures jet.ColumnString
	Fulltext        jet.ColumnString

	AllColumns jet.ColumnList
}

// creates new FilmTable with assigned alias
func (a *FilmTable) AS(alias string) *FilmTable {
	aliasTable := newFilmTable()

	aliasTable.Table.AS(alias)

	return aliasTable
}

Table name and column names are camelized inside sql builder type. AllColumns is used as shorthand notation for list of all columns FilmID, Title, Description,....

Mappings of database types to sql builder column types are following:

Database type(postgres) Sql builder column type
boolean ColumnBool
smallint, integer, bigint ColumnInteger
real, numeric, decimal, double precision ColumnFloat
date ColumnDate
timestamp without time zone ColumnTimestamp
timestamp with time zone ColumnTimestampz
time without time zone ColumnTime
time with time zone ColumnTimez
enums, text, character, character varying
bytea, uuid
and all remaining types ColumnString

Sql builder file for enum type mpaa_rating:

package enum

import "github.com/go-jet/jet"

var MpaaRating = &struct {
	G    jet.StringExpression
	PG   jet.StringExpression
	PG13 jet.StringExpression
	R    jet.StringExpression
	NC17 jet.StringExpression
}{
	G:    jet.NewEnumValue("G"),
	PG:   jet.NewEnumValue("PG"),
	PG13: jet.NewEnumValue("PG-13"),
	R:    jet.NewEnumValue("R"),
	NC17: jet.NewEnumValue("NC-17"),
}

Model files

Sample model file for table film:

package model

import (
	"time"
)

type Film struct {
	FilmID          int32 `sql:"primary_key"`
	Title           string
	Description     *string
	ReleaseYear     *int32
	LanguageID      int16
	RentalDuration  int16
	RentalRate      float64
	Length          *int16
	ReplacementCost float64
	Rating          *MpaaRating
	LastUpdate      time.Time
	SpecialFeatures *string
	Fulltext        string
}

For every column of table film there is appropriate field in model type.
Fields corresponding to primary key columns are tagged with sql:"primary_key". This tag is used during query execution to group row results into desired arbitrary structure. See more at TODO:
Fields are pointer types, if they relate to column that can be NULL.

Mappings of database types to Go types:

Database type(postgres) Go type
boolean bool
smallint int16
integer int32
bigint int64
real float32
numeric, decimal, double precision float64
date, timestamp, time(with or without timezone) time.Time
bytea []byte
uuid uuid.UUID
enum enum name
text, character, character varying,
and all remaining types string

Part of the sample model file for enum mpaa_rating:

package model

import "errors"

type MpaaRating string

const (
	MpaaRating_G    MpaaRating = "G"
	MpaaRating_PG   MpaaRating = "PG"
	MpaaRating_PG13 MpaaRating = "PG-13"
	MpaaRating_R    MpaaRating = "R"
	MpaaRating_NC17 MpaaRating = "NC-17"
)

For a reference SQL table definition of table film:

CREATE TABLE dvds.film (
    film_id integer DEFAULT nextval('dvds.film_film_id_seq'::regclass) NOT NULL,
    title character varying(255) NOT NULL,
    description text,
    release_year dvds.year,
    language_id smallint NOT NULL,
    rental_duration smallint DEFAULT 3 NOT NULL,
    rental_rate numeric(4,2) DEFAULT 4.99 NOT NULL,
    length smallint,
    replacement_cost numeric(5,2) DEFAULT 19.99 NOT NULL,
    rating dvds.mpaa_rating DEFAULT 'G'::dvds.mpaa_rating,
    last_update timestamp without time zone DEFAULT now() NOT NULL,
    special_features text[],
    fulltext tsvector NOT NULL
);