cosmos-sdk/docs/sdk/sdk-by-examples/simple-governance/simple-gov-module.md

Simple Governance Module

Module initialization

First, let us go into the module's folder and create a folder for our module.

cd x/
mkdir simple_governance
cd simple_governance
mkdir -p client/cli client/rest
touch client/cli/simple_governance.go client/rest/simple_governance.go errors.go handler.go handler_test.go keeper_keys.go keeper_test.go keeper.go test_common.go test_types.go types.go codec.go

Let us start by adding the files we will need. Your module's folder should look something like that:

x
└─── simple_governance
      ├─── client
      │     ├───  cli
      │     │     └─── simple_governance.go
      │     └─── rest
      │           └─── simple_governance.go
      ├─── errors.go
      ├─── handler.go
      ├─── keeper_keys.go
      ├─── keeper.go
      ├─── types.go
      └─── codec.go

Let us go into the detail of each of these files.

Types

File: x/simple_governance/types.go

In this file, we define the custom types for our module. This includes the types from the State section and the custom message types defined in the Messages section.

For each new type that is not a message, it is possible to add methods that make sense in the context of the application. In our case, we will implement an updateTally function to easily update the tally of a given proposal as vote messages come in.

Messages are a bit different. They implement the Message interface defined in the SDK's types folder. Here are the methods you need to implement when you define a custom message type:

• Type(): This function returns the name of our module's route. When messages are processed by the application, they are routed using the string returned by the Type() method.
• GetSignBytes(): Returns the byte representation of the message. It is used to sign the message.
• GetSigners(): Returns address(es) of the signer(s).
• ValidateBasic(): This function is used to discard obviously invalid messages. It is called at the beginning of runTx() in the baseapp file. If ValidateBasic() does not return nil, the app stops running the transaction.
• Get(): A basic getter, returns some property of the message.
• String(): Returns a human-readable version of the message

For our simple governance messages, this means:

• Type() will return "simpleGov"
• For SubmitProposalMsg, we need to make sure that the attributes are not empty and that the deposit is both valid and positive. Note that this is only basic validation, we will therefore not check in this method that the sender has sufficient funds to pay for the deposit
• For VoteMsg, we check that the address and option are valid and that the proposalID is not negative.
• As for other methods, less customization is required. You can check the code to see a standard way of implementing these.

Keeper

File: x/simple_governance/keeper.go

Short intro to keepers

Keepers are a module abstraction that handle reading/writing to the module store. This is a practical implementation of the Object Capability Model for Cosmos.

As module developers, we have to define keepers to interact with our module's store(s) not only from within our module, but also from other modules. When another module wants to access one of our module's store(s), a keeper for this store has to be passed to it at the application level. In practice, it will look like that:

// in app.go

// instanciate keepers
keeperA = moduleA.newKeeper(app.moduleAStoreKey)
keeperB = moduleB.newKeeper(app.moduleBStoreKey)

// pass instance of keeperA to handler of module B
app.Router().
        AddRoute("moduleA", moduleA.NewHandler(keeperA)).
        AddRoute("moduleB", moduleB.NewHandler(keeperB, keeperA))   // Here module B can access one of module A's store via the keeperA instance

KeeperA grants a set of capabilities to the handler of module B. When developing a module, it is good practice to think about the sensitivity of the different capabilities that can be granted through keepers. For example, some module may need to read and write to module A's main store, while others only need to read it. If a module has multiple stores, then some keepers could grant access to all of them, while others would only grant access to specific sub-stores. It is the job of the module developer to make sure it is easy for application developers to instanciate a keeper with the right capabilities. Of course, the handler of a module will most likely get an unrestricted instance of that module's keeper.

Store for our app

Before we delve into the keeper itself, let us see what objects we need to store in our governance sub-store, and how to index them.

• Proposals will be indexed by 'proposals'|<proposalID>.
• Votes (Yes, No, Abstain) will be indexed by 'proposals'|<proposalID>|'votes'|<voterAddress>.

Notice the quote mark on 'proposals' and 'votes'. They indicate that these are constant keywords. So, for example, the option casted by voter with address 0x01 on proposal 0101 will be stored at index 'proposals'|0101|'votes'|0x01.

These keywords are used to faciliate range queries. Range queries (TODO: Link to formal spec) allow developer to query a subspace of the store, and return an iterator. They are made possible by the nice properties of the IAVL+ tree that is used in the background. In practice, this means that it is possible to store and query a Key-Value pair in O(1), while still being able to iterate over a given subspace of Key-Value pairs. For example, we can query all the addresses that voted on a given proposal, along with their votes, by calling rangeQuery(SimpleGovStore, <proposalID|'addresses'>).

Keepers for our app

In our case, we only have one store to access, the SimpleGov store. We will need to set and get values inside this store via our keeper. However, these two actions do not have the same impact in terms of security. While there should no problem in granting read access to our store to other modules, write access is way more sensitive. So ideally application developers should be able to create either a governance mapper that can only get values from the store, or one that can both get and set values. To this end, we will introduce two keepers: Keeper and KeeperRead. When application developers create their application, they will be able to decide which of our module's keeper to use.

Now, let us try to think about which keeper from external modules our module's keepers need access to. Each proposal requires a deposit. This means our module needs to be able to both read and write to the module that handles tokens, which is the bank module. We also need to be able to determine the voting power of each voter based on their stake. To this end, we need read access to the store of the staking module. However, we don't need write access to this store. We should therefore indicate that in our module, and the application developer should be careful to only pass a read-only keeper of the staking module to our module's handler.

With all that in mind, we can define the structure of our Keeper:

    type Keeper struct {
        SimpleGov    sdk.StoreKey        // Key to our module's store
        cdc                 *codec.Codec         // Codec to encore/decode structs
        ck                  bank.Keeper         // Needed to handle deposits. This module onlyl requires read/writes to Atom balance
        sm                  stake.Keeper        // Needed to compute voting power. This module only needs read access to the staking store.
        codespace           sdk.CodespaceType   // Reserves space for error codes
    }

And the structure of our KeeperRead:

type KeeperRead struct {
    Keeper
}

KeeperRead will inherit all methods from Keeper, except those that we override. These will be the methods that perform writes to the store.

Functions and Methods

The first function we have to create is the constructor.

func NewKeeper(SimpleGov sdk.StoreKey, ck bank.Keeper, sm stake.Keeper, codespace sdk.CodespaceType) Keeper

This function is called from the main app.go file to instanciate a new Keeper. A similar function exits for KeeperRead.

func NewKeeperRead(SimpleGov sdk.StoreKey, ck bank.Keeper, sm stake.Keeper, codespace sdk.CodespaceType) KeeperRead

Depending on the needs of the application and its modules, either Keeper, KeeperRead, or both, will be instanciated at application level.

Note: Both the Keeper type name and NewKeeper() function's name are standard names used in every module. It is no requirement to follow this standard, but doing so can facilitate the life of application developers

Now, let us describe the methods we need for our module's Keeper. For the full implementation, please refer to keeper.go.

• GetProposal: Get a Proposal given a proposalID. Proposals need to be decoded from byte before they can be read.
• SetProposal: Set a Proposal at index 'proposals'|<proposalID>. Proposals need to be encoded to byte before they can be stored.
• NewProposalID: A function to generate a new unique proposalID.
• GetVote: Get a vote Option given a proposalID and a voterAddress.
• SetVote: Set a vote Option given a proposalID and a voterAddress.
• Proposal Queue methods: These methods implement a standard proposal queue to store Proposals on a First-In First-Out basis. It is used to tally the votes at the end of the voting period.

The last thing that needs to be done is to override certain methods for the KeeperRead type. KeeperRead should not have write access to the stores. Therefore, we will override the methods SetProposal(), SetVote() and NewProposalID(), as well as setProposalQueue() from the Proposal Queue's methods. For KeeperRead, these methods will just throw an error.

Note: If you look at the code, you'll notice that the context ctx is a parameter of many of the methods. The context ctx provides useful information on the current state such as the current block height and allows the keeper k to access the KVStore. You can check all the methods of ctx here.

Handler

File: x/simple_governance/handler.go

Constructor and core handlers

Handlers implement the core logic of the state-machine. When a transaction is routed from the app to the module, it is run by the handler function.

In practice, one handler will be implemented for each message of the module. In our case, we have two message types. We will therefore need two handler functions. We will also need a constructor function to route the message to the correct handler:

func NewHandler(k Keeper) sdk.Handler {
    return func(ctx sdk.Context, msg sdk.Msg) sdk.Result {
        switch msg := msg.(type) {
        case SubmitProposalMsg:
            return handleSubmitProposalMsg(ctx, k, msg)
        case VoteMsg:
            return handleVoteMsg(ctx, k, msg)
        default:
            errMsg := "Unrecognized gov Msg type: " + reflect.TypeOf(msg).Name()
            return sdk.ErrUnknownRequest(errMsg).Result()
        }
    }
}

The messages are routed to the appropriate handler depending on their type. For our simple governance module, we only have two handlers, that correspond to our two message types. They have similar signatures:

func handleSubmitProposalMsg(ctx sdk.Context, k Keeper, msg SubmitProposalMsg) sdk.Result

Let us take a look at the parameters of this function:

• The context ctx to access the stores.
• The keeper k allows the handler to read and write from the different stores, including the module's store (SimpleGovernance in our case) and all the stores from other modules that the keeper k has been granted an access to (stake and bank in our case).
• The message msg that holds all the information provided by the sender of the transaction.

The function returns a Result that is returned to the application. It contains several useful information such as the amount of Gas for this transaction and wether the message was succesfully processed or not. At this point, we exit the boundaries of our simple governance module and go back to root application level. The Result will differ from application to application. You can check the sdk.Result type directly here for more info.

BeginBlocker and EndBlocker

In contrast to most smart-contracts platform, it is possible to perform automatic (i.e. not triggered by a transaction sent by an end-user) execution of logic in Cosmos-SDK applications.

This automatic execution of code takes place in the BeginBlock and EndBlock functions that are called at the beginning and at the end of every block. They are powerful tools, but it is important for application developers to be careful with them. For example, it is crutial that developers control the amount of computing that happens in these functions, as expensive computation could delay the block time, and never-ending loop freeze the chain altogether.

BeginBlock and EndBlock are composable functions, meaning that each module can implement its own BeginBlock and EndBlock logic. When needed, BeginBlock and EndBlock logic is implemented in the module's handler. Here is the standard way to proceed for EndBlock (BeginBlock follows the exact same pattern):

func NewEndBlocker(k Keeper) sdk.EndBlocker {
    return func(ctx sdk.Context, req abci.RequestEndBlock) (res abci.ResponseEndBlock) {
        err := checkProposal(ctx, k)
        if err != nil {
            panic(err)
        }
        return
    }
}

Do not forget that each module need to declare its BeginBlock and EndBlock constructors at application level. See the Application - Bridging it all together.

For the purpose of our simple governance application, we will use EndBlock to automatically tally the results of the vote. Here are the different steps that will be performed:

1. Get the oldest proposal from the ProposalProcessingQueue
2. Check if the CurrentBlock is the block at which the voting period for this proposal ends. If Yes, go to 3.. If no, exit.
3. Check if proposal is accepted or rejected. Update the proposal status.
4. Pop the proposal from the ProposalProcessingQueue and go back to 1.

Let us perform a quick safety analysis on this process.

• The loop will not run forever because the number of proposals in ProposalProcessingQueue is finite
• The computation should not be too expensive because tallying of individual proposals is not expensive and the number of proposals is expected be relatively low. That is because proposals require a Deposit to be accepted. MinDeposit should be high enough so that we don't have too many Proposals in the queue.
• In the eventuality that the application becomes so successful that the ProposalProcessingQueue ends up containing so many proposals that the blockchain starts slowing down, the module should be modified to mitigate the situation. One clever way of doing it is to cap the number of iteration per individual EndBlock at MaxIteration. This way, tallying will be spread over many blocks if the number of proposals is too important and block time should remain stable. This would require to modify the current check if (CurrentBlock == Proposal.SubmitBlock + VotingPeriod) to if (CurrentBlock > Proposal.SubmitBlock + VotingPeriod) AND (Proposal.Status == ProposalStatusActive).

Codec

File: x/simple_governance/codec.go

The codec.go file allows developers to register the concrete message types of their module into the codec. In our case, we have two messages to declare:

func RegisterCodec(cdc *codec.Codec) {
    cdc.RegisterConcrete(SubmitProposalMsg{}, "simple_governance/SubmitProposalMsg", nil)
    cdc.RegisterConcrete(VoteMsg{}, "simple_governance/VoteMsg", nil)
}

Don't forget to call this function in app.go (see Application - Bridging it all together) for more).

Errors

File: x/simple_governance/errors.go

The error.go file allows us to define custom error messages for our module. Declaring errors should be relatively similar in all modules. You can look in the error.go file directly for a concrete example. The code is self-explanatory.

Note that the errors of our module inherit from the sdk.Error interface and therefore possess the method Result(). This method is useful when there is an error in the handler and an error has to be returned in place of an actual result.

Command-Line Interface

File: x/simple_governance/client/cli/simple_governance.go

Go in the cli folder and create a simple_governance.go file. This is where we will define the commands for our module.

The CLI builds on top of Cobra. Here is the schema to build a command on top of Cobra:

    // Declare flags
    const(
        Flag = "flag"
        ...
    )

    // Main command function. One function for each command.
    func Command(codec *codec.Codec) *cobra.Command {
        // Create the command to return
        command := &cobra.Command{
            Use: "actual command",
            Short: "Short description",
            Run: func(cmd *cobra.Command, args []string) error {
                // Actual function to run when command is used
            },
        }

        // Add flags to the command
        command.Flags().<Type>(FlagNameConstant, <example_value>, "<Description>")

        return command
    }

Rest API

File: x/simple_governance/client/rest/simple_governance.goo

The Rest Server, also called Light-Client Daemon (LCD), provides support for HTTP queries.

---

USER INTERFACE <=======> REST SERVER <=======> FULL-NODE

---

It allows end-users that do not want to run full-nodes themselves to interract with the chain. The LCD can be configured to perform Light-Client verification via the flag --trust-node, which can be set to true or false.

• If light-client verification is enabled, the Rest Server acts as a light-client and needs to be run on the end-user's machine. It allows them to interract with the chain in a trustless way without having to store the whole chain locally.

• If light-client verification is disabled, the Rest Server acts as a simple relayer for HTTP calls. In this setting, the Rest server needs not be run on the end-user's machine. Instead, it will probably be run by the same entity that operates the full-node the server connects to. This mode is useful if end-users trust the full-node operator and do not want to store anything locally.

Now, let us define endpoints that will be available for users to query through HTTP requests. These endpoints will be defined in a simple_governance.go file stored in the rest folder.

Method	URL	Description
GET	/proposals	Range query to get all submitted proposals
POST	/proposals	Submit a new proposal
GET	/proposals/{id}	Returns a proposal given its ID
GET	/proposals/{id}/votes	Range query to get all the votes casted on a given proposal
POST	/proposals/{id}/votes	Cast a vote on a given proposal
GET	/proposals/{id}/votes/{address}	Returns the vote of a given address on a given proposal

It is the job of module developers to provide sensible endpoints so that front-end developers and service providers can properly interact with it.

Additionaly, here is a link for REST APIs best practices.