nightingale/vendor/github.com/spf13/viper

• Install
• What is Viper?
• Why Viper?
• Putting Values into Viper
• • Establishing Defaults
  • Reading Config Files
  • Writing Config Files
  • Watching and re-reading config files
  • Reading Config from io.Reader
  • Setting Overrides
  • Registering and Using Aliases
  • Working with Environment Variables
  • • Env example
  • Working with Flags
  • • Flag interfaces
  • Remote Key/Value Store Support
  • Remote Key/Value Store Example - Unencrypted
  • • etcd
    • Consul
  • Remote Key/Value Store Example - Encrypted
  • Watching Changes in etcd - Unencrypted
• Getting Values From Viper
• • Accessing nested keys
  • Extract sub-tree
  • Unmarshaling
  • Marshalling to string
• Viper or Vipers?
• • Working with multiple vipers
• Q & A

Go configuration with fangs!

Many Go projects are built using Viper including:

• Hugo
• EMC RexRay
• Imgur’s Incus
• Nanobox/Nanopack
• Docker Notary
• BloomApi
• doctl
• Clairctl
• Mercure

Install

go get github.com/spf13/viper

What is Viper?

Viper is a complete configuration solution for Go applications including 12-Factor apps. It is designed
to work within an application, and can handle all types of configuration needs
and formats. It supports:

• setting defaults
• reading from JSON, TOML, YAML, HCL, envfile and Java properties config files
• live watching and re-reading of config files (optional)
• reading from environment variables
• reading from remote config systems (etcd or Consul), and watching changes
• reading from command line flags
• reading from buffer
• setting explicit values

Viper can be thought of as a registry for all of your applications configuration needs.

Why Viper?

When building a modern application, you don’t want to worry about
configuration file formats; you want to focus on building awesome software.
Viper is here to help with that.

Viper does the following for you:

1. Find, load, and unmarshal a configuration file in JSON, TOML, YAML, HCL, INI, envfile or Java properties formats.
2. Provide a mechanism to set default values for your different configuration options.
3. Provide a mechanism to set override values for options specified through command line flags.
4. Provide an alias system to easily rename parameters without breaking existing code.
5. Make it easy to tell the difference between when a user has provided a command line or config file which is the same as the default.

Viper uses the following precedence order. Each item takes precedence over the item below it:

• explicit call to Set
• flag
• env
• config
• key/value store
• default

Important: Viper configuration keys are case insensitive.
There are ongoing discussions about making that optional.

Putting Values into Viper

Establishing Defaults

A good configuration system will support default values. A default value is not
required for a key, but it’s useful in the event that a key hasn't been set via
config file, environment variable, remote configuration or flag.

Examples:

viper.SetDefault("ContentDir", "content")
viper.SetDefault("LayoutDir", "layouts")
viper.SetDefault("Taxonomies", map[string]string{"tag": "tags", "category": "categories"})

Reading Config Files

Viper requires minimal configuration so it knows where to look for config files.
Viper supports JSON, TOML, YAML, HCL, INI, envfile and Java Properties files. Viper can search multiple paths, but
currently a single Viper instance only supports a single configuration file.
Viper does not default to any configuration search paths leaving defaults decision
to an application.

Here is an example of how to use Viper to search for and read a configuration file.
None of the specific paths are required, but at least one path should be provided
where a configuration file is expected.

viper.SetConfigName("config") // name of config file (without extension)
viper.SetConfigType("yaml") // REQUIRED if the config file does not have the extension in the name
viper.AddConfigPath("/etc/appname/")   // path to look for the config file in
viper.AddConfigPath("$HOME/.appname")  // call multiple times to add many search paths
viper.AddConfigPath(".")               // optionally look for config in the working directory
err := viper.ReadInConfig() // Find and read the config file
if err != nil { // Handle errors reading the config file
	panic(fmt.Errorf("Fatal error config file: %s \n", err))
}

You can handle the specific case where no config file is found like this:

if err := viper.ReadInConfig(); err != nil {
    if _, ok := err.(viper.ConfigFileNotFoundError); ok {
        // Config file not found; ignore error if desired
    } else {
        // Config file was found but another error was produced
    }
}

// Config file found and successfully parsed

NOTE [since 1.6]: You can also have a file without an extension and specify the format programmaticaly. For those configuration files that lie in the home of the user without any extension like .bashrc

Writing Config Files

Reading from config files is useful, but at times you want to store all modifications made at run time.
For that, a bunch of commands are available, each with its own purpose:

• WriteConfig - writes the current viper configuration to the predefined path, if exists. Errors if no predefined path. Will overwrite the current config file, if it exists.
• SafeWriteConfig - writes the current viper configuration to the predefined path. Errors if no predefined path. Will not overwrite the current config file, if it exists.
• WriteConfigAs - writes the current viper configuration to the given filepath. Will overwrite the given file, if it exists.
• SafeWriteConfigAs - writes the current viper configuration to the given filepath. Will not overwrite the given file, if it exists.

As a rule of the thumb, everything marked with safe won't overwrite any file, but just create if not existent, whilst the default behavior is to create or truncate.

A small examples section:

viper.WriteConfig() // writes current config to predefined path set by 'viper.AddConfigPath()' and 'viper.SetConfigName'
viper.SafeWriteConfig()
viper.WriteConfigAs("/path/to/my/.config")
viper.SafeWriteConfigAs("/path/to/my/.config") // will error since it has already been written
viper.SafeWriteConfigAs("/path/to/my/.other_config")

Watching and re-reading config files

Viper supports the ability to have your application live read a config file while running.

Gone are the days of needing to restart a server to have a config take effect,
viper powered applications can read an update to a config file while running and
not miss a beat.

Simply tell the viper instance to watchConfig.
Optionally you can provide a function for Viper to run each time a change occurs.

Make sure you add all of the configPaths prior to calling WatchConfig()

viper.WatchConfig()
viper.OnConfigChange(func(e fsnotify.Event) {
	fmt.Println("Config file changed:", e.Name)
})

Reading Config from io.Reader

Viper predefines many configuration sources such as files, environment
variables, flags, and remote K/V store, but you are not bound to them. You can
also implement your own required configuration source and feed it to viper.

viper.SetConfigType("yaml") // or viper.SetConfigType("YAML")

// any approach to require this configuration into your program.
var yamlExample = []byte(`
Hacker: true
name: steve
hobbies:
- skateboarding
- snowboarding
- go
clothing:
  jacket: leather
  trousers: denim
age: 35
eyes : brown
beard: true
`)

viper.ReadConfig(bytes.NewBuffer(yamlExample))

viper.Get("name") // this would be "steve"

Setting Overrides

These could be from a command line flag, or from your own application logic.

viper.Set("Verbose", true)
viper.Set("LogFile", LogFile)

Registering and Using Aliases

Aliases permit a single value to be referenced by multiple keys

viper.RegisterAlias("loud", "Verbose")

viper.Set("verbose", true) // same result as next line
viper.Set("loud", true)   // same result as prior line

viper.GetBool("loud") // true
viper.GetBool("verbose") // true

Working with Environment Variables

Viper has full support for environment variables. This enables 12 factor
applications out of the box. There are five methods that exist to aid working
with ENV:

• AutomaticEnv()
• BindEnv(string...) : error
• SetEnvPrefix(string)
• SetEnvKeyReplacer(string...) *strings.Replacer
• AllowEmptyEnv(bool)

When working with ENV variables, it’s important to recognize that Viper
treats ENV variables as case sensitive.

Viper provides a mechanism to try to ensure that ENV variables are unique. By
using SetEnvPrefix, you can tell Viper to use a prefix while reading from
the environment variables. Both BindEnv and AutomaticEnv will use this
prefix.

BindEnv takes one or two parameters. The first parameter is the key name, the
second is the name of the environment variable. The name of the environment
variable is case sensitive. If the ENV variable name is not provided, then
Viper will automatically assume that the ENV variable matches the following format: prefix + "_" + the key name in ALL CAPS. When you explicitly provide the ENV variable name (the second parameter),
it does not automatically add the prefix. For example if the second parameter is "id",
Viper will look for the ENV variable "ID".

One important thing to recognize when working with ENV variables is that the
value will be read each time it is accessed. Viper does not fix the value when
the BindEnv is called.

AutomaticEnv is a powerful helper especially when combined with
SetEnvPrefix. When called, Viper will check for an environment variable any
time a viper.Get request is made. It will apply the following rules. It will
check for a environment variable with a name matching the key uppercased and
prefixed with the EnvPrefix if set.

SetEnvKeyReplacer allows you to use a strings.Replacer object to rewrite Env
keys to an extent. This is useful if you want to use - or something in your
Get() calls, but want your environmental variables to use _ delimiters. An
example of using it can be found in viper_test.go.

Alternatively, you can use EnvKeyReplacer with NewWithOptions factory function.
Unlike SetEnvKeyReplacer, it accepts a StringReplacer interface allowing you to write custom string replacing logic.

By default empty environment variables are considered unset and will fall back to
the next configuration source. To treat empty environment variables as set, use
the AllowEmptyEnv method.

Env example

SetEnvPrefix("spf") // will be uppercased automatically
BindEnv("id")

os.Setenv("SPF_ID", "13") // typically done outside of the app

id := Get("id") // 13

Working with Flags

Viper has the ability to bind to flags. Specifically, Viper supports Pflags
as used in the Cobra library.

Like BindEnv, the value is not set when the binding method is called, but when
it is accessed. This means you can bind as early as you want, even in an
init() function.

For individual flags, the BindPFlag() method provides this functionality.

Example:

serverCmd.Flags().Int("port", 1138, "Port to run Application server on")
viper.BindPFlag("port", serverCmd.Flags().Lookup("port"))

You can also bind an existing set of pflags (pflag.FlagSet):

Example:

pflag.Int("flagname", 1234, "help message for flagname")

pflag.Parse()
viper.BindPFlags(pflag.CommandLine)

i := viper.GetInt("flagname") // retrieve values from viper instead of pflag

The use of pflag in Viper does not preclude
the use of other packages that use the flag
package from the standard library. The pflag package can handle the flags
defined for the flag package by importing these flags. This is accomplished
by a calling a convenience function provided by the pflag package called
AddGoFlagSet().

Example:

package main

import (
	"flag"
	"github.com/spf13/pflag"
)

func main() {

	// using standard library "flag" package
	flag.Int("flagname", 1234, "help message for flagname")

	pflag.CommandLine.AddGoFlagSet(flag.CommandLine)
	pflag.Parse()
	viper.BindPFlags(pflag.CommandLine)

	i := viper.GetInt("flagname") // retrieve value from viper

	...
}

Flag interfaces

Viper provides two Go interfaces to bind other flag systems if you don’t use Pflags.

FlagValue represents a single flag. This is a very simple example on how to implement this interface:

type myFlag struct {}
func (f myFlag) HasChanged() bool { return false }
func (f myFlag) Name() string { return "my-flag-name" }
func (f myFlag) ValueString() string { return "my-flag-value" }
func (f myFlag) ValueType() string { return "string" }

Once your flag implements this interface, you can simply tell Viper to bind it:

viper.BindFlagValue("my-flag-name", myFlag{})

FlagValueSet represents a group of flags. This is a very simple example on how to implement this interface:

type myFlagSet struct {
	flags []myFlag
}

func (f myFlagSet) VisitAll(fn func(FlagValue)) {
	for _, flag := range flags {
		fn(flag)
	}
}

Once your flag set implements this interface, you can simply tell Viper to bind it:

fSet := myFlagSet{
	flags: []myFlag{myFlag{}, myFlag{}},
}
viper.BindFlagValues("my-flags", fSet)

Remote Key/Value Store Support

To enable remote support in Viper, do a blank import of the viper/remote
package:

import _ "github.com/spf13/viper/remote"

Viper will read a config string (as JSON, TOML, YAML, HCL or envfile) retrieved from a path
in a Key/Value store such as etcd or Consul. These values take precedence over
default values, but are overridden by configuration values retrieved from disk,
flags, or environment variables.

Viper uses crypt to retrieve
configuration from the K/V store, which means that you can store your
configuration values encrypted and have them automatically decrypted if you have
the correct gpg keyring. Encryption is optional.

You can use remote configuration in conjunction with local configuration, or
independently of it.

crypt has a command-line helper that you can use to put configurations in your
K/V store. crypt defaults to etcd on http://127.0.0.1:4001.

$ go get github.com/xordataexchange/crypt/bin/crypt
$ crypt set -plaintext /config/hugo.json /Users/hugo/settings/config.json

Confirm that your value was set:

$ crypt get -plaintext /config/hugo.json

See the crypt documentation for examples of how to set encrypted values, or
how to use Consul.

Remote Key/Value Store Example - Unencrypted

etcd

viper.AddRemoteProvider("etcd", "http://127.0.0.1:4001","/config/hugo.json")
viper.SetConfigType("json") // because there is no file extension in a stream of bytes, supported extensions are "json", "toml", "yaml", "yml", "properties", "props", "prop", "env", "dotenv"
err := viper.ReadRemoteConfig()

Consul

You need to set a key to Consul key/value storage with JSON value containing your desired config.
For example, create a Consul key/value store key MY_CONSUL_KEY with value:

{
    "port": 8080,
    "hostname": "myhostname.com"
}

viper.AddRemoteProvider("consul", "localhost:8500", "MY_CONSUL_KEY")
viper.SetConfigType("json") // Need to explicitly set this to json
err := viper.ReadRemoteConfig()

fmt.Println(viper.Get("port")) // 8080
fmt.Println(viper.Get("hostname")) // myhostname.com

Remote Key/Value Store Example - Encrypted

viper.AddSecureRemoteProvider("etcd","http://127.0.0.1:4001","/config/hugo.json","/etc/secrets/mykeyring.gpg")
viper.SetConfigType("json") // because there is no file extension in a stream of bytes,  supported extensions are "json", "toml", "yaml", "yml", "properties", "props", "prop", "env", "dotenv"
err := viper.ReadRemoteConfig()

Watching Changes in etcd - Unencrypted

// alternatively, you can create a new viper instance.
var runtime_viper = viper.New()

runtime_viper.AddRemoteProvider("etcd", "http://127.0.0.1:4001", "/config/hugo.yml")
runtime_viper.SetConfigType("yaml") // because there is no file extension in a stream of bytes, supported extensions are "json", "toml", "yaml", "yml", "properties", "props", "prop", "env", "dotenv"

// read from remote config the first time.
err := runtime_viper.ReadRemoteConfig()

// unmarshal config
runtime_viper.Unmarshal(&runtime_conf)

// open a goroutine to watch remote changes forever
go func(){
	for {
	    time.Sleep(time.Second * 5) // delay after each request

	    // currently, only tested with etcd support
	    err := runtime_viper.WatchRemoteConfig()
	    if err != nil {
	        log.Errorf("unable to read remote config: %v", err)
	        continue
	    }

	    // unmarshal new config into our runtime config struct. you can also use channel
	    // to implement a signal to notify the system of the changes
	    runtime_viper.Unmarshal(&runtime_conf)
	}
}()

Getting Values From Viper

In Viper, there are a few ways to get a value depending on the value’s type.
The following functions and methods exist:

• Get(key string) : interface{}
• GetBool(key string) : bool
• GetFloat64(key string) : float64
• GetInt(key string) : int
• GetIntSlice(key string) : []int
• GetString(key string) : string
• GetStringMap(key string) : map[string]interface{}
• GetStringMapString(key string) : map[string]string
• GetStringSlice(key string) : []string
• GetTime(key string) : time.Time
• GetDuration(key string) : time.Duration
• IsSet(key string) : bool
• AllSettings() : map[string]interface{}

One important thing to recognize is that each Get function will return a zero
value if it’s not found. To check if a given key exists, the IsSet() method
has been provided.

Example:

viper.GetString("logfile") // case-insensitive Setting & Getting
if viper.GetBool("verbose") {
    fmt.Println("verbose enabled")
}

Accessing nested keys

The accessor methods also accept formatted paths to deeply nested keys. For
example, if the following JSON file is loaded:

{
    "host": {
        "address": "localhost",
        "port": 5799
    },
    "datastore": {
        "metric": {
            "host": "127.0.0.1",
            "port": 3099
        },
        "warehouse": {
            "host": "198.0.0.1",
            "port": 2112
        }
    }
}

Viper can access a nested field by passing a . delimited path of keys:

GetString("datastore.metric.host") // (returns "127.0.0.1")

This obeys the precedence rules established above; the search for the path
will cascade through the remaining configuration registries until found.

For example, given this configuration file, both datastore.metric.host and
datastore.metric.port are already defined (and may be overridden). If in addition
datastore.metric.protocol was defined in the defaults, Viper would also find it.

However, if datastore.metric was overridden (by a flag, an environment variable,
the Set() method, …) with an immediate value, then all sub-keys of
datastore.metric become undefined, they are “shadowed” by the higher-priority
configuration level.

Lastly, if there exists a key that matches the delimited key path, its value
will be returned instead. E.g.

{
    "datastore.metric.host": "0.0.0.0",
    "host": {
        "address": "localhost",
        "port": 5799
    },
    "datastore": {
        "metric": {
            "host": "127.0.0.1",
            "port": 3099
        },
        "warehouse": {
            "host": "198.0.0.1",
            "port": 2112
        }
    }
}

GetString("datastore.metric.host") // returns "0.0.0.0"

Extract sub-tree

Extract sub-tree from Viper.

For example, viper represents:

app:
  cache1:
    max-items: 100
    item-size: 64
  cache2:
    max-items: 200
    item-size: 80

After executing:

subv := viper.Sub("app.cache1")

subv represents:

max-items: 100
item-size: 64

Suppose we have:

func NewCache(cfg *Viper) *Cache {...}

which creates a cache based on config information formatted as subv.
Now it’s easy to create these 2 caches separately as:

cfg1 := viper.Sub("app.cache1")
cache1 := NewCache(cfg1)

cfg2 := viper.Sub("app.cache2")
cache2 := NewCache(cfg2)

Unmarshaling

You also have the option of Unmarshaling all or a specific value to a struct, map,
etc.

There are two methods to do this:

• Unmarshal(rawVal interface{}) : error
• UnmarshalKey(key string, rawVal interface{}) : error

Example:

type config struct {
	Port int
	Name string
	PathMap string `mapstructure:"path_map"`
}

var C config

err := viper.Unmarshal(&C)
if err != nil {
	t.Fatalf("unable to decode into struct, %v", err)
}

If you want to unmarshal configuration where the keys themselves contain dot (the default key delimiter),
you have to change the delimiter:

v := viper.NewWithOptions(viper.KeyDelimiter("::"))

v.SetDefault("chart::values", map[string]interface{}{
    "ingress": map[string]interface{}{
        "annotations": map[string]interface{}{
            "traefik.frontend.rule.type":                 "PathPrefix",
            "traefik.ingress.kubernetes.io/ssl-redirect": "true",
        },
    },
})

type config struct {
	Chart struct{
        Values map[string]interface{}
    }
}

var C config

v.Unmarshal(&C)

Viper also supports unmarshaling into embedded structs:

/*
Example config:

module:
    enabled: true
    token: 89h3f98hbwf987h3f98wenf89ehf
*/
type config struct {
	Module struct {
		Enabled bool

		moduleConfig `mapstructure:",squash"`
	}
}

// moduleConfig could be in a module specific package
type moduleConfig struct {
	Token string
}

var C config

err := viper.Unmarshal(&C)
if err != nil {
	t.Fatalf("unable to decode into struct, %v", err)
}

Viper uses github.com/mitchellh/mapstructure under the hood for unmarshaling values which uses mapstructure tags by default.

Marshalling to string

You may need to marshal all the settings held in viper into a string rather than write them to a file.
You can use your favorite format's marshaller with the config returned by AllSettings().

import (
    yaml "gopkg.in/yaml.v2"
    // ...
) 

func yamlStringSettings() string {
    c := viper.AllSettings()
    bs, err := yaml.Marshal(c)
    if err != nil {
        log.Fatalf("unable to marshal config to YAML: %v", err)
    }
    return string(bs)
}

Viper or Vipers?

Viper comes ready to use out of the box. There is no configuration or
initialization needed to begin using Viper. Since most applications will want
to use a single central repository for their configuration, the viper package
provides this. It is similar to a singleton.

In all of the examples above, they demonstrate using viper in its singleton
style approach.

Working with multiple vipers

You can also create many different vipers for use in your application. Each will
have its own unique set of configurations and values. Each can read from a
different config file, key value store, etc. All of the functions that viper
package supports are mirrored as methods on a viper.

Example:

x := viper.New()
y := viper.New()

x.SetDefault("ContentDir", "content")
y.SetDefault("ContentDir", "foobar")

//...

When working with multiple vipers, it is up to the user to keep track of the
different vipers.

Q & A

Q: Why is it called “Viper”?

A: Viper is designed to be a companion
to Cobra. While both can operate completely
independently, together they make a powerful pair to handle much of your
application foundation needs.

Q: Why is it called “Cobra”?

A: Is there a better name for a commander?