[go: up one dir, main page]

Skip to content

jrapoport/chestnut

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

18 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

🌰  Chestnut

GitHub Workflow Status Go Report Card Codecov branch GitHub go.mod Go version GitHub

Buy Me A Coffee

Chestnut is encrypted storage for Go. The goal was an easy to use encrypted store with helpful features that was quick to set up, but highly flexible.

Chestnut is written in pure go and designed not to have strong opinions about things like storage, compression, hashing, secrets, or encryption. Chestnut is a storage chest, and not a datastore itself. As such, Chestnut must be backed by a storage solution.

Currently, Chestnut supports BBolt and NutsDB as backing storage.

Table of Contents

Getting Started

Installing

To start using Chestnut, install Go (version 1.11+) and run go get:

$ go get -u github.com/jrapoport/chestnut

Importing Chestnut

To use Chestnut as an encrypted store, import as:

import (
  "github.com/jrapoport/chestnut"
  "github.com/jrapoport/chestnut/encryptor/aes"
  "github.com/jrapoport/chestnut/encryptor/crypto"
  "github.com/jrapoport/chestnut/storage/nuts"
)

// use nutsdb for storage
store := nuts.NewStore(path)

// use AES256-CFB for encryption
opt := chestnut.WithAES(crypto.Key256, aes.CFB, mySecret)

cn := chestnut.NewChestnut(store, opt)
if err := cn.Open(); err != nil {
    return err
}

defer cn.Close()

Requirements

Chestnut has two requirements:

  1. Storage that supports the storage.Storage interface (with a lightweight adapter).
  2. Encryption which supports the crypto.Encryptor interface.

Storage

Chestnut will work seamlessly with any storage solution (or adapter) that supports thestorage.Storage interface.

Built-in

Currently, Chestnut has built-in support for BBolt and NutsDB.

BBolt

https://github.com/etcd-io/bbolt Chestnut has built-in support for using BBolt as a backing store.

To use bbolt for a backing store you can import Chestnut's bolt package and call bolt.NewStore():

import "github.com/jrapoport/chestnut/storage/bolt"

//use or create a bbolt backing store at path
store := bolt.NewStore(path)

// use bbolt for the storage chest
cn := chestnut.NewChestnut(store, ...)

NutsDB

https://github.com/nutsdb/nutsdb
Chestnut has built-in support for using NutsDB as a backing store.

To use nutsDB for a backing store you can import Chestnut's nuts package and call nuts.NewStore():

import "github.com/jrapoport/chestnut/storage/nuts"

//use or create a nutsdb backing store at path
store := nuts.NewStore(path)

// use nutsdb for the storage chest
cn := chestnut.NewChestnut(store, ...)

Planned

Other K/V stores like LevelDB.

GORM (probably not) Gorm is an ORM, so while it's not a datastore per se, it could be adapted to support sparse encryption and would mean automatic support for databases like mysql, sqlite, etc. However, most (if not all) of those DBs already support built-in encryption, so w/o some compelling use-case that's not already covered I don't see a lot of value-add.

Encryption

Chestnut supports several flavors of AES out of the box:

  • AES128-CFB, AES192-CFB, and AES256-CFB
  • AES128-CTR, AES192-CTR, and AES256-CTR
  • AES128-GCM, AES192-GCM, and AES256-GCM

You can add AES encryption to Chestnut by passing the chestnut.WithAES() option:

opt := chestnut.WithAES(crypto.Key256, aes.CFB, mySecret)

AES256-CTR

For encryption we recommend using AES256-CTR. We chose AES256-CTR based in part on this helpful analysis from Shawn Wang, PostgreSQL Database Core.

Custom Encryption

Chestnut supports drop-in custom encryption. A struct that supports the crypto.Encryptor interface can be used with the chestnut.WithEncryptor() option.

Supporting crypto.Encryptor interface is straightforward and mainly consists of vending the following two methods:

// Encrypt returns data encrypted with the secret.
Encrypt(plaintext []byte) (ciphertext []byte, err error)

// Decrypt returns data decrypted with the secret.
Decrypt(ciphertext []byte) (plaintext []byte, err error)

Chained Encryption

Chestnut also supports chained encryption which allows data to be arbitrarily transformed by a chain of Encryptors in a FIFO order.

A chain of crypto.Encryptors can be passed to Chestnut with the chestnut.WithEncryptorChain option:

opt := chestnut.WithEncryptorChain(
    encryptor.NewAESEncryptor(crypto.Key128, aes.CFB, secret1),
    encryptor.NewAESEncryptor(crypto.Key192, aes.CTR, secret2),
    encryptor.NewAESEncryptor(crypto.Key256, aes.GCM, secret3),
)

or by using a crypto.ChainEncryptor with the chestnut.WithEncryptor option:

encryptors := []crypto.Encryptor{
    encryptor.NewAESEncryptor(crypto.Key128, aes.CFB, secret1),
    encryptor.NewAESEncryptor(crypto.Key192, aes.CTR, secret2),
    encryptor.NewAESEncryptor(crypto.Key256, aes.GCM, secret3),
}
chain := crypto.NewChainEncryptor(encryptors...)
opt := chestnut.WithEncryptor(chain)

If you use both the chestnut.WithEncryptor and the chestnut.WithEncryptorChain options, the crypto.Encryptor from chestnut.WithEncryptor will be prepended* to the chain.

Sparse Encryption

Chestnut supports the sparse encryption of structs.

Sparse encryption is a transparent feature of saving structs with Chestnut.Save(), Chestnut.Load(), and Chestnut.Sparse(); or structs that support the value.Keyed interface with Chestnut.SaveKeyed(), Chestnut.LoadKeyed(), and Chestnut.SparseKeyed().

What is "sparse" encryption?

With sparse encryption, only struct fields marked as secure will be encrypted. The remaining "plaintext" fields are encoded and stored separately.

This allows you to load a "sparse" copy of the struct by calling Chestnut.Sparse() or Chestnut.SparseKeyed() (if you have a value.Keyed value) and examine the plaintext fields without the overhead of decryption. When a sparse struct is loaded, the contents of struct fields marked as secure are replaced by empty values.

Enabling Sparse Encryption

Chestnut uses struct tags to indicate which specific struct fields should be encrypted. To enable sparse encryption for a struct, add the secure tag option to the JSON tag of at least one struct field:

SecretKey string `json:",secure"` // 'secure' option (bare minimum)

like so:

type MySparseStruct struct {
    SecretValue string `json:"secret_value,secure"` // <-- add 'secure' here
    PublicValue string `json:"public_value"`
}

Using Sparse Encryption

Structs can be sparsely encrypted by calling Chestnut.Save(), or if the struct supports the value.Keyed interface, Chestnut.SaveKeyed(). Chestnut will automatically detect the secure tag and do the rest.

If no secure fields are found, Chestnut will encrypt the entire struct.

sparseObj := &MySparseStruct{
    SecretValue: "this is a secret",
    PublicValue: "this is public",
}

err := cn.Save("my-namespace",  []byte("my-key"), sparseObj)

When MySparseStruct is saved, Chestnut will detect the secure struct field and only encrypt those fields. Any remaining fields will be encoded as plaintext. In the case of MySparseStructthis means that SecretValuewill be encrypted prior to being encoded, and PublicValuewill not be encrypted.

Sparse Loading

A sparse struct can be loaded by calling Chestnut.Sparse(), or if the struct supports the value.Keyed interface, Chestnut.SparseKeyed(). When these methods are called to load a sparsely encrypted struct, a partially decoded struct will be returned, but the no decryption will occur. Secure fields will instead be replaced by empty values.

sparseObj := &MySparseStruct{}

err := cn.Sparse("my-namespace",  []byte("my-key"), sparseObj)

Examining the struct will reveal that the secure fields were replaced with empty values, and not decrypted.

*MySparseStruct{
    SecretValue: ""
    PublicValue: "this is public"
}

Only sparsely encrypted structs can be sparsely loaded
If Chestnut.Sparse() or Chestnut.SparseKeyed() is called on a struct that was not sparsely encrypted, the fully decrypted struct will be returned.

Decryption

A sparsely encrypted struct can be fully decrypted by calling Chestnut.Load(), or if the struct supports the value.Keyed interface, Chestnut.LoadKeyed(). When any of those methods are called on a sparsely encrypted struct, a fully decrpted copy of the struct is returned.

Secrets

Chestnut secrets are handled through the crypto.Secret interface. The crypto.Secret interface is designed to provide a high degree of flexibility around how you store, retrieve, and manage the secrets you use for encryption.

While Chestnut currently only comes with AES symmetric key encryption, the crypto.Secret interface can easily be adapted to support other forms of encryption like a private key-based crypto.Encryptor.

Chestnut currently provides three basic immplementations of the crypto.Secret interface which should cover most cases.

TextSecret

crypto.TextSecret provides a lightweight wrapper around a plaintext string.

textSecret := crypto.NewTextSecret("a-secret")

ManagedSecret

crypto.ManagedSecret provides a unique ID alongside a plaintext string secret. You can use this id to securely track the secret if you use external vaults or functionality like rollover.

managedSecret := crypto.NewManagedSecret("my-secret-id", "a-secret")

SecureSecret

crypto.SecureSecret provides a unique id for a secret alongside an openSecret() callback which returns a byte representation of the secret for encryption and decryption on SecureSecret.Open(). When crypto.SecureSecret calls openSecret() it will pass a copy of itself as a crypto.Secret. This allows for remote loading of the secret based on its id, or using a secure in-memory storage solution for the secret like memguarded.

openSecret := func(s crypto.Secret) []byte {
	// fetch the secret 
    mySecret := getMySecretFromTheVault(s.ID())
    return mySecret
}
secureSecret := crypto.NewSecureSecret("my-secret-id", openSecret)

Compression

Chestnut supports compression of the encoded data. Compression takes place prior to encryption.

Compression can be enabled through the chestnut.WithCompression option and passing it a supported compression format:

opt := chestnut.WithCompression(compress.Zstd)

Data compressed while chestnut.WithCompression is active with a supported compression format will continue to be correctly decompressed when read even if compression is no longer active (i.e. chestnut.WithCompression is no longer being used). This is not true with custom compression. Data compressed using custom compression cannot be decompressed if that custom compression is disabled.

Zstandard

Chestnut currently supports Zstandard compression out of the box with the compress.Zstd format option. To enable Zstandard compression, call chestnut.WithCompression passing compress.Zstd as the compression format:

opt := chestnut.WithCompression(compress.Zstd)

Please Note: I have no affiliation with Facebook (past or present) and just liked this compression format.

Custom Compression

If you wish to supply your own compression routines you can do so easily with the chestnut.WithCompressors option:

opt := chestnut.WithCompressors(myCompressorFn, myDecompressorFn)

Your two custom compression functions, a compressor compress.CompressorFunc, and a decompressor compress.DecompressorFunc must have the following format:

Compressor(data []byte) (compressed []byte, err error)

Decompressor(compressed []byte) (data []byte, err error)

Compression + Sparse Encryption

Enabling compression will not affect sparse encryption. Sparsely encrypted values compress their secure and plaintext encodings independently.

Operations

Chestnut supports all basic CRUD operations with a few extras.

All WRITE operations: Chestnut.Put(), Chestnut.Save(), & Chestnut.SaveKeyed(), will encrypt data prior to it being stored.

All READ operations: Chestnut.Get(), Chestnut.Load(), & Chestnut.LoadKeyed(), will decrypt data prior to it being returned.

All SPARSE operations: Chestnut.Sparse(), & Chestnut.SparseKeyed(), will not decrypt data prior to it being returned.

In all cases no record of the plaintext data is kept
(even with DebugLevel logging enabled).

Basic Operations

Put

To save an encrypted value to a namespaced key in the storage chest, use the Chestnut.Put() function:

err := cn.Put("my-namespace", []byte("my-key"), []byte("plaintext"))

This will set the value of the "my-key" key to the encrypted ciphertext of "plaintext" in the my-namespace namespace. If a namespace does not exist, it will be automatically created.

If the key already exists, and the storage chest was initialized with the chestnut.OverwritesForbidden option, this call will fail with ErrForbidden.

To retrieve this value, we can use the Chestnut.Get() function:

Get

To retrieve a decrypted value from a namespaced key in the storage chest, we can use the Chestnut.Get() function:

plaintext, err := cn.Get("my-namespace", []byte("my-key"))

Delete

Use the Chestnut.Delete() function to delete a key from the store.

err := cn.Delete("my-namespace", []byte("my-key"))

Struct Operations

Chestnut provides several functions for working directly with structs. In addition to handling the marshalling, encoding and encryption of structs for you, these functions provide automatic support for the Chestnut struct field tag options secure and hash. SEE: Struct Field Tags for more detail.

Save

To encrypt and save a struct to the store we can use the Chestnut.Save() function:

err := cn.Save("my-namespace", []byte("my-key"), myStruct)

Chestnut will marshal and encrypt the encoded byte representation. If the struct supports the secure struct field tag option on one of its fields, Chestnut will automatically sparsely encrypt the struct. Other supported struct field tag options will also be applied.

Load

To retrieve the fully decrypted struct, we can use the Chestnut.Load() function:

err := cn.Load("my-namespace", []byte("my-key"), &myStruct)

Sparse

Chestnut.Sparse() loads the struct at key and returns the sparsely decoded result. Unlike Chestnut.Load(), it does not decrypt the encoded struct and secure fields are replaced with empty values. To retrieve a sparse value, we can use the Chestnut.Sparse() function:

err := cn.Sparse("my-namespace", []byte("my-key"), &myStruct)

If the struct was not saved as a sparsely encoded struct this has no effect and is equivalent to calling Chestnut.Load(). Structs must have been saved with secure fields to be loaded as sparse structs by Chestnut.Sparse().

When a sparse struct is returned, any fields marked as secure will be decoded
as nil or empty values. For more information, please see the section on sparse encryption.

Keyed Operations

Chestnut provides several convenience functions for working with struct values that support the value.Keyed interface. Keyed values can supply their own namespace and keys via calls to Keyed.Namespace() and Keyed.Keys(), respectively. Internally these functions are equivalent to calling Chestnut.Save(), Chestnut.Load(), and Chestnut.Sparse() with an explicit namespace and key.

SaveKeyed

To encrypt and store a struct that implements the value.Keyed interface to the store we can use the Chestnut.SaveKeyed() function:

err := cn.SaveKeyed(myKeyedStruct)

To save a keyed struct with Chestnut.SaveKeyed(), the struct must be initialized with namespace and key you want to save it to prior to calling Chestnut.LoadKeyed() in order to satisfy the value.Keyed interface:

ko := MyKeyedValue{ name: "my-namespase", key: "my key"}
err := cn.SaveKeyed(&ko)

For more information, please see Chestnut.Save()

LoadKeyed

To retrieve the fully decrypted struct that implements the value.Keyed interface, we can use the Chestnut.LoadKeyed() function:

err := cn.LoadKeyed(&myKeyedStruct)

To load a keyed struct with Chestnut.LoadKeyed(), the struct must be initialized with namespace and key you want to retrieve prior to calling Chestnut.LoadKeyed() in order to satisfy the value.Keyed interface:

ko := MyKeyedValue{ name: "my-namespase", key: "my key"}
err := cn.LoadKeyed(&ko)

For more information, please see Chestnut.Load()

SparseKeyed

To retrieve a sparsely encrypted struct that implements the value.Keyed interface, we can use the Chestnut.SparseKeyed() function:

err := cn.SparseKeyed(&myKeyedStruct)

To load a sparse keyed struct with Chestnut.SparseKeyed(), the struct must be initialized with namespace and key you want to retrieve prior to calling Chestnut.SparseKeyed() in order to satisfy the value.Keyed interface:

ko := MyKeyedValue{ name: "my-namespase", key: "my key"}
err := cn.SparseKeyed(&ko)

For more information, please see Chestnut.Sparse()

Extra Operations

Chestnut supports a few additional functions that you might find helpful. In the future more may be added assuming they can be reasonably supported by the storage.Storage interface and generally make sense to do so. If there is a specific function you'd like to see added, please feel free to open an issue request to discuss.

Has

You can check to see if a key exists by calling Chestnut.Has(). If the key is found, it will return true, otherwise false. If an error occured, Chestnut.Has() will return false along with the error.

has, err := cn.Has("my-namespace", []byte("my-key"))

List

To get a list of all the keys for a namespace you can call Chestnut.List():

keys, err := cn.List("my-namespace")

ListAll

To get a mapped list of all keys in the store organized by namespace you can call Chestnut.ListAll():

keymap, err := cn.ListAll()

Export

To export the storage chest to another path you can call Chestnut.Export():

err := cn.Export("/a/path/someplace")

Chestnut cannot be exported to its current location. If you call Chestnut.Export() and pass the path to Chestnut's current location an error will be returned.

Struct Field Tags

Chestnut currently supports two extensions to the `json` struct field tag as options: secure when added to the `json` tag, secure marks the field for sparse encryption. hash when added to the `json` tag, hash marks a string field for hashing.

These options will be automatically detected and applied when the struct is saved with Chestnut.Save(), or Chestnut.SaveKeyed().

NOTE: The order in which the tag options appear is unimportant.

// these are equivalent

`json:"my_value,secure,omitempty"`

`json:"my_value,omitempty,secure"`

In the future, Chestnut will also support its own struct field tag. `cn`.

Secure

When the secure option is added to a `json` struct field tag, the struct field is marked for sparse encryption. If Chestnut detects a secure option on a struct field tag, only those fields marked with secure will be encrypted. If no secure fields are found, Chestnut will encrypt the entire struct.

To mark a struct field as secure, just add secure as an option to a `json` struct field tag (like omitempty). The following are some examples of how the secure option can be added to the `json` struct field tag:

type MySecureStruct struct {
    ValueA     int      `json:",secure"`           // *will* be encrypted
    ValueB     struct{} `json:"value_b,secure"`    // *will* be encrypted
    ValueC     string   `json:",omitempty,secure"` // *will* be encrypted
    PlaintextA string                              // will *not* be encrypted
    PlaintextB int      `json:""`                  // will *not* be encrypted
    PlaintextC int      `json:"-"`                 // will *not* be encrypted
    privateA   int      `json:",secure"`           // will *not* be encrypted
}

Fields marked with secure are encrypted hierarchically, meaning if you have:

package main

type MyStructA struct {
	ValueA string `json:"value_a,secure"`    // *will* be encrypted
}

type MyStructB struct {
	MyStructA                                // will *not* be encrypted
	ValueB    string `json:"value_b"`        // will *not* be encrypted
}

type MyStructC struct {
	MyStructA                                // will *not* be encrypted
	ValueC    string `json:"value_c"`        // will *not* be encrypted
}

type MyStructD struct {
	ValueD string    `json:"value_d,secure"` // *will* be encrypted
	Embed1 MyStructA                         // will *not* be encrypted
	Embed2 MyStructB                         // will *not* be encrypted
	Embed3 MyStructB `json:"embed_3,secure"` // *will* be encrypted
}

var myStruct = &MyStructD{
	ValueD: "foo",
	Embed1: MyStructA{
		ValueA: "bar",
	},
	Embed2: MyStructB{
		MyStructA: MyStructA{
			ValueA: "quack",
		},
		ValueB: "baz",
	},
	Embed3: MyStructB{
		MyStructA: MyStructA{
			ValueA: "foobar",
		},
		ValueB: "bonk",
	},
}

myStruct will be encrypted by Chestnut as:

*MyStructD {
  ValueD: ****
  Embed1: main.MyStructA{
      ValueA: ****
  },
  Embed2: main.MyStructB{
      MyStructA: main.MyStructA{
      	ValueA: ****
      },
      ValueB: ****
  },
  Embed3: ****
}

where '****' represents an encrypted value.

Please see Sparse Encryption for more information.

Hash

When the hash option is added to a `json` struct field tag of a string field, the string field is marked for hashing. If Chestnut detects a hash option on a string field, the string value of the field will be replaced with its hash.

If the hash option is applied to a struct field that is not type string, it is ignored.

To hash a string field of a struct, just add hash as an option to a `json` struct field tag (like omitempty). The following are some examples of how the hash option can be added to the `json` struct field tag:

type MyHashStruct struct {
    ValueA     string   `json:",hash"`           // *will* be hashed
    ValueB     string   `json:"value_b,hash"`    // *will* be hashed
    ValueC     string   `json:",omitempty,hash"` // *will* be hashed
    ValueD     string   `json:",hash,omitempty"` // *will* be hashed
    ...
    Count      int      `json:"count,hash"`      // will *not* be hashed
}

Taking the above struct as an example:

var myHashStruct = &MyHashStruct {
    ValueA: "value a",
    ValueB  "value b",
    ValueC  "value c",
    ValueD  "value d",
    ...
    Count   42,
}

myHashStruct will be encoded as:

*main.MyHashStruct {
    ValueA: "sha256:BEBA1D9847D6E595D8DD6832DEE5432916C6F7AE438BC9A99C5BAFDD0E93793E"
    ValueB  "sha256:2A53D83488A34E898436908A7064276859FFF56F69D16E2F61573057EDEBFB64"
    ValueC  "sha256:80C92DE321A1CB8AEA3025890DE39A5BAA95A91DF022B10E1A71BEABB8BCC1BE"
    ValueD  "sha256:8080378350428FABDE1724D9B920D613B8920A71151F1A9AF37EB4AF43628AE4"
	...
    Count   42
}

SHA256

Chestnut currently supports SHA256 for hashing. In the future the hash option may be extending to include an algorithm name e.g.:

`json:"some_field,hash=sha3-256"`

in which which case hash would continue to default to sha256.

Hash Prefix

Hashed struct fields will have the algorithm used to hash the value pre-pended to the hash string:

sha256:BEBA1D9847D6E595D8DD6832DEE5432916C6F7AE438BC9A99C5BAFDD0E93793E

Chestnut uses the [hash alogorithm name]: prefix to know that it has already hashed the value, and it should hash it again when the struct is saved.

IMPORTANT! Changing or removing the hash prefix will cause Chestnut to rehash the value of the struct field the next time the struct is saved.

Multiple Tags

Chestnut supports the combining of tag options. You are free to mark a struct field as both secure and hash:

type MyCombinedStruct struct {
    ValueA     string   `json:"value_a,secure,hash"` // will be hashed *AND* encrypted 
    ...
}

As with other tag options, the order in which they appear is unimportant.

However, the order in which Chestnut applies them is fixed. A field marked with both secure and hash will be first be hashed, and then encrypted. This order of operations cannot be changed for obvious reasons.

Disable Overwrites

Chestnut supports the disabling of overwrites via the chestnut.OverwritesForbidden option.

cn := chestnut.NewChestnut(store, encryptor chestnut.OverwritesForbidden())

When this option is set, once a value has been saved to a namespaced key, successive calls to save a value to the same key will fail with ErrForbidden.

The key must be explicitly deleted before a new call to save a value for the same key will succeed.

Keystore

Chestnut includes an implementation of IPFS compliant keystore which can be found here.

Importing Keystore

Using the Keystore is straight forward:

package main

import (
	"github.com/jrapoport/chestnut"
	"github.com/jrapoport/chestnut/encryptor/aes"
	"github.com/jrapoport/chestnut/encryptor/crypto"
	"github.com/jrapoport/chestnut/keystore"
	"github.com/jrapoport/chestnut/storage/nuts"
)

// use nutsdb
store := nuts.NewStore(path)

// use a simple text secret
textSecret := crypto.TextSecret("i-am-a-good-secret")

// use AES256-CFB encryption
opt := chestnut.WithAES(crypto.Key256, aes.CFB, textSecret)

// open the keystore with nutsdb and the aes encryptor
ks := keystore.NewKeystore(store, opt)
if err := ks.Open(); err != nil {
    return err
}

A complete example of the Chestnut Keystore can be found here.

Important Note

package main

import  (
    "github.com/ipfs/go-ipfs/keystore"
    "github.com/libp2p/go-libp2p/core/crypto"
)

If you want to work with the Keystore, please make make sure you are importing go-ipfs and go-libp2p-core, and NOT importing go-ipfs-keystore and go-libp2p-crypto β€” which are DEPRECATED, out of date, a/o archived, etc.

Logging

Chestnut supports logging via the log.Logger interface and the chestnut.WithLogger() option. The log.Logger interface conforms to Logrus, Zap, and the standard Go logger (with the log.Std adapter), for example:

opt := chestnut.WithLogger(myLogger)

Logrus Logger

Chestnut supports Logrus logger.

*logrus.Logger and *logrus.Entry both work with Chestnut's log.Logger interface:

logger := logrus.New() // *logrus.Logger

opt := chestnut.WithLogger(logger)

or

logger := logrus.New()
logger = logger.WithField("hello", "world") // *logrus.Entry

opt := chestnut.WithLogger(logger)

In addition to the chestnut.WithLogger() option, you can use the convenience option, chestnut.WithLogrusLogger():

opt := chestnut.WithLogrusLogger(log.InfoLevel)

chestnut.WithLogrusLogger() will return a new *logrus.Entry set to the normalized log level you requested. This is equivalent to calling logrus.New() followed by logrus.SetLevel().

Zap Logger

Chestnut supports Zap logger.

*zap.SugaredLogger works with Chestnut's log.Logger interface:

logger :=zap.NewProduction().Sugar() // *zap.SugaredLogger

opt := chestnut.WithLogger(logger)

In addition to the chestnut.WithLogger() option, you can use the convenience option, chestnut.WithZapLogger():

opt := chestnut.WithZapLogger(log.InfoLevel)

chestnut.WithZapLogger() will return a new *zap.SugaredLogger set to the normalized log level you requested. This is equivalent to calling zap.NewProduction(), followed by zap.Core().Enabled(), and finally, zap.Sugar().

Standard Logger

Chestnut supports the Go standard logger.

We provide a lightweight wrapper for Go's standard logger *log.Logger which supports Chestnut's log.Logger interface.

import (
    "log"
    "os"

    "github.com/jrapoport/chestnut"
    cnlog "github.com/jrapoport/chestnut/log"
)

logger := cnlog.NewStdLogger(log.InfoLevel os.Stderr, "", log.LstdFlags)
opt := chestnut.WithLogger(logger)

chestnut.WithStdLogger() will return a new *log.stdLogger set to the normalized log level you requested. This is equivalent to calling log.NewStdLogger() with the specified log level.

Storage

Lastly, the Chestnut stores also accept matching options for logging: storage.WithLogger, storage.WithLogrusLogger, storage.WithZapLogger, and storage.WithStdLogger.

// enable logging 
opt := storage.WithLogger(myLogger)
// use nutsdb
store := nuts.NewStore(path, opt)

Enabling logging for the backing store was intentionally kept separate for additional flexibility. This allows you to log Chestnut operations without automatically incurring the noise of store operations and vice versa.

Examples

Run any example with make <example-dir>

$ make sparse

Known Issues

  • Because we use JSON encoding for structs, time.Time will lose resolution when encoded:

    IN:  2021-01-09 17:29:36.349522 -0800 PST m=+0.002746093
    OUT: 2021-01-09 17:29:36.349522 -0800 PST
    
    

Misc

JSON encoding

We use the jsoniter JSON encoder internally.