Go CSV

The GoCSV package aims to provide easy serialization and deserialization functions to use CSV in Golang

API and techniques inspired from https://godoc.org/gopkg.in/mgo.v2

Installation

go get -u github.com/gocarina/gocsv

Full example

Consider the following CSV file

client_id,client_name,client_age
1,Jose,42
2,Daniel,26
3,Vincent,32

Easy binding in Go!

package main

import (
	"fmt"
	"os"

	"github.com/gocarina/gocsv"
)

type NotUsed struct {
	Name string
}

type Client struct { // Our example struct, you can use "-" to ignore a field
	Id            string `csv:"client_id"`
	Name          string `csv:"client_name"`
	Age           string `csv:"client_age"`
	NotUsedString string `csv:"-"`
	NotUsedStruct NotUsed `csv:"-"` 
}

func main() {
	clientsFile, err := os.OpenFile("clients.csv", os.O_RDWR|os.O_CREATE, os.ModePerm)
	if err != nil {
		panic(err)
	}
	defer clientsFile.Close()

	clients := []*Client{}

	if err := gocsv.UnmarshalFile(clientsFile, &clients); err != nil { // Load clients from file
		panic(err)
	}
	for _, client := range clients {
		fmt.Println("Hello", client.Name)
	}

	if _, err := clientsFile.Seek(0, 0); err != nil { // Go to the start of the file
		panic(err)
	}

	clients = append(clients, &Client{Id: "12", Name: "John", Age: "21"}) // Add clients
	clients = append(clients, &Client{Id: "13", Name: "Fred"})
	clients = append(clients, &Client{Id: "14", Name: "James", Age: "32"})
	clients = append(clients, &Client{Id: "15", Name: "Danny"})
	csvContent, err := gocsv.MarshalString(&clients) // Get all clients as CSV string
	//err = gocsv.MarshalFile(&clients, clientsFile) // Use this to save the CSV back to the file
	if err != nil {
		panic(err)
	}
	fmt.Println(csvContent) // Display all clients as CSV string

}

Customizable Converters

type DateTime struct {
	time.Time
}

// Convert the internal date as CSV string
func (date *DateTime) MarshalCSV() (string, error) {
	return date.Time.Format("20060102"), nil
}

// You could also use the standard Stringer interface 
func (date *DateTime) String() (string) {
	return date.String() // Redundant, just for example
}

// Convert the CSV string as internal date
func (date *DateTime) UnmarshalCSV(csv string) (err error) {
	date.Time, err = time.Parse("20060102", csv)
	return err
}

type Client struct { // Our example struct with a custom type (DateTime)
	Id       string   `csv:"id"`
	Name     string   `csv:"name"`
	Employed DateTime `csv:"employed"`
}

Nested structs

When decoding nested pointer structs, GoCSV allocates the pointers needed to reach each field. An omitempty field still omits an empty leaf pointer value; it does not prevent allocation of the enclosing structs.

By default, the fields of nested structs are prefixed with the parent field's name. For example:

type Foo struct {
	B Bar
	X int `csv:"x"`
}

type Bar struct {
	A int `csv:"a"`
	B int `csv:"b"`
}

produces columns B.a, B.b, x.

Use the csv:"." tag to inline the nested struct's fields without that prefix; for example:

type Foo struct {
	B Bar `csv:"."`
	X int `csv:"x"`
}

type Bar struct {
	A int `csv:"a"`
	B int `csv:"b"`
}

produces columns a, b, x.

Sanitizing values before decoding

Use a custom Decoder with UnmarshalDecoder to normalize values before they are converted to struct fields. Parse the CSV first, keep its header row, and map only the data columns that need cleaning. For example, the runnable sanitizing example maps N/A to NaN only in the Height column; quoted names, Unicode, other columns and ordinary conversion errors keep their usual behavior.

The example uses csv.Reader.ReadAll, so it holds all records in memory. For row-by-row processing, implement a SimpleDecoder whose GetCSVRow maps each data row, then use UnmarshalDecoderToChan or UnmarshalDecoderToCallback. Neither approach requires changing the process-wide SetCSVReader configuration.

Customizable CSV Reader / Writer

func main() {
        ...
	
        gocsv.SetCSVReader(func(in io.Reader) gocsv.CSVReader {
            r := csv.NewReader(in)
            r.Comma = '|'
            return r // Allows use pipe as delimiter
        })	
	
        ...
	
        gocsv.SetCSVReader(func(in io.Reader) gocsv.CSVReader {
            r := csv.NewReader(in)
            r.LazyQuotes = true
            r.Comma = '.'
            return r // Allows use dot as delimiter and use quotes in CSV
        })
	
        ...
	
        gocsv.SetCSVReader(func(in io.Reader) gocsv.CSVReader {
            //return csv.NewReader(in)
            return gocsv.LazyCSVReader(in) // Allows use of quotes in CSV
        })

        ...

        gocsv.UnmarshalFile(file, &clients)

        ...

        gocsv.SetCSVWriter(func(out io.Writer) *gocsv.SafeCSVWriter {
            writer := csv.NewWriter(out)
            writer.Comma = '|'
            return gocsv.NewSafeCSVWriter(writer)
        })

        ...

        gocsv.MarshalFile(&clients, file)

        ...
}