smapping

Golang structs generic mapping.

Version Limit

To support nesting object conversion, the lowest Golang version supported is 1.12.0.
To support smapping.SQLScan, the lowest Golang version supported is 1.13.0.

Motivation At Glimpse.
Motivation Length.
Install.
Examples.
License.

At Glimpse

What?

A library to provide a mapped structure generically/dynamically.

Who?

Anyone who has to work with large structure.

Why?

Scalability and Flexibility.

When?

At the runtime.

Where?

In users code.

How?

By converting into smapping.Mapped which alias for map[string]interface{}, users can iterate the struct arbitarily with reflect package.

Motivation

Working with between struct, and json with Golang has various degree of difficulty. The thing that makes difficult is that sometimes we get arbitrary json or have to make json with arbitrary fields. Sometime we also need to have a different field names, extracting specific fields, working with same structure with different domain fields name etc.

In order to answer those flexibility, we map the object struct to the more general data structure as table/map.

Table/Map is the data structure which ubiquitous after list, which in turn table/map can be represented as list of pair values (In Golang we can't have it because there's no tuple data type, tuple is limited as return values).

Object can be represented as table/map dynamically just like in JavaScript/EcmaScript which object is behaving like table and in Lua with its metatable. By some extent we can represent the JSON as table too.

In this library, we provide the mechanism to smoothly map the object representation back-and-forth without having the boilerplate of type-checking one by one by hand. Type-checking by hand is certainly seems easier when the domain set is small, but it soon becomes unbearable as the structure and/or architecure dynamically changed because of newer insight and information. Hence in Who section mentioned this library is for anyone who has to work with large domain set.

Except for type smapping.Mapped as alias, we don't provide others type struct currently as each operation doesn't need to keep the internal state so each operation is transparent and almost functional (almost functional because we modify the struct fields values instead of returning the new struct itself, but this is only trade-off because Golang doesn't have type-parameter which known as generic).

Since v0.1.10, we added the MapEncoder and MapDecoder interfaces for users to have custom conversion for custom and self-defined struct.

Install

go get github.com/mashingan/smapping

Examples

Basic usage examples

Below example are basic representation how we can work with smapping. Several examples are converged into single runnable example for the ease of reusing the same structure definition and its various tags. Refer this example to get a glimpse of how to do things. Afterward, users can creatively use to accomplish what they're wanting to do with the provided flexibility.

package main

import (
	"encoding/json"
	"fmt"

	"github.com/mashingan/smapping"
)

type Source struct {
	Label   string `json:"label"`
	Info    string `json:"info"`
	Version int    `json:"version"`
}

type Sink struct {
	Label string
	Info  string
}

type HereticSink struct {
	NahLabel string `json:"label"`
	HahaInfo string `json:"info"`
	Version  string `json:"heretic_version"`
}

type DifferentOneField struct {
	Name    string `json:"name"`
	Label   string `json:"label"`
	Code    string `json:"code"`
	Private string `json:"private" api:"internal"`
}

func main() {
	source := Source{
		Label:   "source",
		Info:    "the origin",
		Version: 1,
	}
	fmt.Println("source:", source)
	mapped := smapping.MapFields(source)
	fmt.Println("mapped:", mapped)
	sink := Sink{}
	err := smapping.FillStruct(&sink, mapped)
	if err != nil {
		panic(err)
	}
	fmt.Println("sink:", sink)

	maptags := smapping.MapTags(source, "json")
	fmt.Println("maptags:", maptags)
	hereticsink := HereticSink{}
	err = smapping.FillStructByTags(&hereticsink, maptags, "json")
	if err != nil {
		panic(err)
	}
	fmt.Println("heretic sink:", hereticsink)

	fmt.Println("=============")
	recvjson := []byte(`{"name": "bella", "label": "balle", "code": "albel", "private": "allbe"}`)
	dof := DifferentOneField{}
	_ = json.Unmarshal(recvjson, &dof)
	fmt.Println("unmarshaled struct:", dof)

	marshaljson, _ := json.Marshal(dof)
	fmt.Println("marshal back:", string(marshaljson))

	// What we want actually "internal" instead of "private" field
	// we use the api tags on to make the json
	apijson, _ := json.Marshal(smapping.MapTagsWithDefault(dof, "api", "json"))
	fmt.Println("api marshal:", string(apijson))

	fmt.Println("=============")
	// This time is the reverse, we receive "internal" field when
	// we need to receive "private" field to match our json tag field
	respjson := []byte(`{"name": "bella", "label": "balle", "code": "albel", "internal": "allbe"}`)
	respdof := DifferentOneField{}
	_ = json.Unmarshal(respjson, &respdof)
	fmt.Println("unmarshal resp:", respdof)

	// to get that, we should put convert the json to Mapped first
	jsonmapped := smapping.Mapped{}
	_ = json.Unmarshal(respjson, &jsonmapped)
	// now we fill our struct respdof
	_ = smapping.FillStructByTags(&respdof, jsonmapped, "api")
	fmt.Println("full resp:", respdof)
	returnback, _ := json.Marshal(respdof)
	fmt.Println("marshal resp back:", string(returnback))
	// first we unmarshal respdof, we didn't get the "private" field
	// but after our mapping, we get "internal" field value and
	// simply marshaling back to `returnback`
}

Nested object example

This example illustrates how we map back-and-forth even with deep nested object structure. The ability to map nested objects is to creatively change its representation whether to flatten all tagged field name even though the inner struct representation is nested. Regardless of the usage (whether to flatten the representation) or just simply fetching and remapping into different domain name set, the ability to map the nested object is necessary.

type RefLevel3 struct {
	What string `json:"finally"`
}
type Level2 struct {
	*RefLevel3 `json:"ref_level3"`
}
type Level1 struct {
	Level2 `json:"level2"`
}
type TopLayer struct {
	Level1 `json:"level1"`
}
type MadNest struct {
	TopLayer `json:"top"`
}

var madnestStruct MadNest = MadNest{
	TopLayer: TopLayer{
		Level1: Level1{
			Level2: Level2{
				RefLevel3: &RefLevel3{
					What: "matryoska",
				},
			},
		},
	},
}

func main() {
	// since we're targeting the same MadNest, both of functions will yield
	// same result hence this unified example/test.
	var madnestObj MadNest
	var err error
	testByTags := true
	if testByTags {
		madnestMap := smapping.MapTags(madnestStruct, "json")
		err = smapping.FillStructByTags(&madnestObj, madnestMap, "json")
	} else {
		madnestMap := smapping.MapFields(madnestStruct)
		err = smapping.FillStruct(&madnestObj)
	}
	if err != nil {
		fmt.Printf("%s", err.Error())
		return
	}
	// the result should yield as intented value.
	if madnestObj.TopLayer.Level1.Level2.RefLevel3.What != "matryoska" {
		fmt.Printf("Error: expected \"matroska\" got \"%s\"", madnestObj.Level1.Level2.RefLevel3.What)
	}
}

SQLScan usage example

This example, we're using sqlite3 as the database, we add a convenience feature for any struct/type that implements Scan method as smapping.SQLScanner. Keep in mind this is quite different with sql.Scanner that's also requiring the type/struct to implement Scan method. The difference here, smapping.SQLScanner receiving variable arguments of interface{} as values' placeholder while sql.Scanner is only receive a single interface{} argument as source. smapping.SQLScan is working for Scan literally after we've gotten the *sql.Row or *sql.Rows.

package main

import (
	"database/sql"
	"encoding/json"
	"fmt"

	"github.com/mashingan/smapping"
	_ "github.com/mattn/go-sqlite3"
)

type book struct {
	Author author `json:"author"`
}

type author struct {
	Num  int            `json:"num"`
	ID   sql.NullString `json:"id"`
	Name sql.NullString `json:"name"`
}

func (a author) MarshalJSON() ([]byte, error) {
	mapres := map[string]interface{}{}
	if !a.ID.Valid {
		//if a.ID == nil || !a.ID.Valid {
		mapres["id"] = nil
	} else {
		mapres["id"] = a.ID.String
	}
	//if a.Name == nil || !a.Name.Valid {
	if !a.Name.Valid {
		mapres["name"] = nil
	} else {
		mapres["name"] = a.Name.String
	}
	mapres["num"] = a.Num
	return json.Marshal(mapres)
}

func getAuthor(db *sql.DB, id string) author {
	res := author{}
	err := db.QueryRow("select * from author where id = ?", id).
		Scan(&res.Num, &res.ID, &res.Name)
	if err != nil {
		panic(err)
	}
	return res
}

func getAuthor12(db *sql.DB, id string) author {
	result := author{}
	fields := []string{"num", "id", "name"}
	err := smapping.SQLScan(
		db.QueryRow("select * from author where id = ?", id),
		&result,
		"json",
		fields...)
	if err != nil {
		panic(err)
	}
	return result
}

func getAuthor13(db *sql.DB, id string) author {
	result := author{}
	fields := []string{"num", "name"}
	err := smapping.SQLScan(
		db.QueryRow("select num, name from author where id = ?", id),
		&result,
		"json",
		fields...)
	if err != nil {
		panic(err)
	}
	return result
}

func getAllAuthor(db *sql.DB) []author {
	result := []author{}
	rows, err := db.Query("select * from author")
	if err != nil {
		panic(err)
	}
	for rows.Next() {
		res := author{}
		if err := smapping.SQLScan(rows, &res, "json"); err != nil {
			fmt.Println("error scan:", err)
			break
		}
		result = append(result, res)
	}
	return result
}

func main() {
	db, err := sql.Open("sqlite3", "./dummy.db")
	if err != nil {
		panic(err)
	}
	defer db.Close()
	_, err = db.Exec(`
drop table if exists author;
create table author(num integer primary key autoincrement, id text, name text);
insert into author(id, name) values
('id1', 'name1'),
('this-nil', null);`)
	if err != nil {
		panic(err)
	}
	//auth1 := author{ID: &sql.NullString{String: "id1"}}
	auth1 := author{ID: sql.NullString{String: "id1"}}
	auth1 = getAuthor(db, auth1.ID.String)
	fmt.Println("auth1:", auth1)
	jsonbyte, _ := json.Marshal(auth1)
	fmt.Println("json auth1:", string(jsonbyte))
	b1 := book{Author: auth1}
	fmt.Println(b1)
	jbook1, _ := json.Marshal(b1)
	fmt.Println("json book1:", string(jbook1))
	auth2 := getAuthor(db, "this-nil")
	fmt.Println("auth2:", auth2)
	jbyte, _ := json.Marshal(auth2)
	fmt.Println("json auth2:", string(jbyte))
	b2 := book{Author: auth2}
	fmt.Println("book2:", b2)
	jbook2, _ := json.Marshal(b2)
	fmt.Println("json book2:", string(jbook2))
	fmt.Println("author12:", getAuthor12(db, auth1.ID.String))
	fmt.Println("author13:", getAuthor13(db, auth1.ID.String))
	fmt.Println("all author1:", getAllAuthor(db))
}

Omit fields example

Often we need to reuse the same object with exception a field or two. With smapping it's possible to generate map with custom tag. However having different tag would be too much of manual work.
In this example, we'll see how to exclude using the delete keyword.

package main

import (
	"github.com/mashingan/smapping"
)

type Struct struct {
	Field1       int    `json:"field1"`
	Field2       string `json:"field2"`
	RequestOnly  string `json:"input"`
	ResponseOnly string `jsoN:"output"`
}

func main() {
	s := Struct{
		Field1:       5,
		Field2:       "555",
		RequestOnly:  "vanish later",
		ResponseOnly: "still available",
	}

	m := smapping.MapTags(s, "json")
	_, ok := m["input"]
	if !ok {
		panic("key 'input' should be still available")
	}
	delete(m, "input")
	_, ok = m["input"]
	if ok {
		panic("key 'input' should be not available")
	}
}

LICENSE

MIT

mashingan/smapping