# README

frostfs-sdk-go

Go implementation of FrostFS SDK. It contains high-level version-independent wrappers for structures from frostfs-api-go as well as helper functions for simplifying node/dApp implementations.

Repository structure

accounting

Contains fixed-point Decimal type for performing balance calculations.

eacl

Contains Extended ACL types for fine-grained access control. There is also a reference implementation of checking algorithm which is used in FrostFS node.

checksum

Contains Checksum type encapsulating checksum as well as it's kind. Currently Sha256 and Tillich-Zemor hashsum are in use.

owner

owner.ID type represents single account interacting with FrostFS. In v2 version of protocol it is just raw bytes behing base58-encoded address in Neo blockchain. Note that for historical reasons it contains version prefix and checksum in addition to script-hash.

token

Contains Bearer token type with several FrostFS-specific methods.

ns

In FrostFS there are 2 types of name resolution: DNS and NNS. NNS stands for Neo Name Service is just a contract deployed on a Neo blockchain. Basically, NNS is just a DNS-on-chain which can be used for resolving container nice-names as well as any other name in dApps. See our CoreDNS plugin for the example of how NNS can be integrated in DNS.

session

To help lightweight clients interact with FrostFS without sacrificing trust, FrostFS has a concept of session token. It is signed by client and allows any node with which a session is established to perform certain actions on behalf of the user.

client

Contains client for working with FrostFS.

var prmInit client.PrmInit
prmInit.SetDefaultPrivateKey(key) // private key for request signing
prmInit.ResolveFrostFSFailures() // enable erroneous status parsing

var c client.Client
c.Init(prmInit)

var prmDial client.PrmDial
prmDial.SetServerURI("grpcs://localhost:40005") // endpoint address

err := c.Dial(prmDial)
if err != nil {
    return
}
    
ctx, cancel := context.WithTimeout(context.Background(), 5 * time.Second)
defer cancel()

var prm client.PrmBalanceGet
prm.SetAccount(acc)

res, err := c.BalanceGet(ctx, prm)
if err != nil {
    return
}

fmt.Printf("Balance for %s: %v\n", acc, res.Amount())

Response status

In FrostFS every operation can fail on multiple levels, so a single error doesn't suffice, e.g. consider a case when object was put on 4 out of 5 replicas. Thus, all request execution details are contained in Status returned from every RPC call. dApp can inspect them if needed and perform any desired action. In the case above we may want to report these details to the user as well as retry an operation, possibly with different parameters. Status wire-format is extendable and each node can report any set of details it wants. The set of reserved status codes can be found in FrostFS API. There is also a client.PrmInit.ResolveFrostFSFailures() to seamlessly convert erroneous statuses into Go error type.

policy

Contains helpers allowing conversion of placing policy from/to JSON representation and SQL-like human-readable language.

p, _ := policy.Parse(`
    REP 2
    SELECT 6 FROM F
    FILTER StorageType EQ SSD AS F`)

// Convert parsed policy back to human-readable text and print.
println(strings.Join(policy.Encode(p), "\n"))

netmap

Contains CRUSH-like implementation of container node selection algorithm. Relevant details are described in this paper http://ceur-ws.org/Vol-2344/short10.pdf . Note that it can be outdated in some details.

netmap/json_tests subfolder contains language-agnostic tests for selection algorithm.

import (
    "github.com/TrueCloudLab/frostfs-sdk-go/netmap"
    "github.com/TrueCloudLab/frostfs-sdk-go/object"
)

func placementNodes(addr *object.Address, p *netmap.PlacementPolicy, frostfsNodes []netmap.NodeInfo) {
    // Convert list of nodes in FrostFS API format to the intermediate representation.
    nodes := netmap.NodesFromInfo(nodes)

    // Create new netmap (errors are skipped for the sake of clarity). 
    nm, _ := NewNetmap(nodes)

    // Calculate nodes of container.
    cn, _ := nm.GetContainerNodes(p, addr.ContainerID().ToV2().GetValue())

    // Return list of nodes for each replica to place object on in the order of priority.
    return nm.GetPlacementVectors(cn, addr.ObjectID().ToV2().GetValue())
}