This document details how to build CLI and REST interfaces for a module. Examples from various SDK modules are included. {synopsis}
- Building Modules Intro {prereq}
One of the main interfaces for an application is the command-line interface. This entrypoint adds commands from the application's modules to let end-users create messages and queries. The CLI files are typically found in the ./x/moduleName/client/cli folder.
Transactions are created by users to wrap messages that trigger state changes when they get included in a valid block. Transaction commands typically have their own tx.go file in the module ./x/moduleName/client/cli folder. The commands are specified in getter functions prefixed with GetCmd and include the name of the command.
Here is an example from the nameservice module:
This getter function creates the command for the Buy Name transaction. It does the following:
- Construct the command: Read the Cobra Documentation for details on how to create commands.
- Use: Specifies the format of a command-line entry users should type in order to invoke this command. In this case, the user uses
buy-nameas the name of the transaction command and provides thenamethe user wishes to buy and theamountthe user is willing to pay. - Args: The number of arguments the user provides, in this case exactly two:
nameandamount. - Short and Long: A description for the function is provided here. A
Shortdescription is expected, andLongcan be used to provide a more detailed description when a user uses the--helpflag to ask for more information. - RunE: Defines a function that can return an error, called when the command is executed. Using
Runwould do the same thing, but would not allow for errors to be returned.
- Use: Specifies the format of a command-line entry users should type in order to invoke this command. In this case, the user uses
RunEFunction Body: The function should be specified as aRunEto allow for errors to be returned. This function encapsulates all of the logic to create a new transaction that is ready to be relayed to nodes.- The function should first initialize a
TxBuilderwith the applicationcodec'sTxEncoder, as well as a newContextwith thecodecandAccountDecoder. These contexts contain all the information provided by the user and will be used to transfer this user-specific information between processes. To learn more about how contexts are used in a transaction, click here. - If applicable, the command's arguments are parsed. Here, the
amountgiven by the user is parsed into a denomination ofcoins. - If applicable, the
Contextis used to retrieve any parameters such as the transaction originator's address to be used in the transaction. Here, thefromaddress is retrieved by callingcliCtx.getFromAddress(). - A message is created using all parameters parsed from the command arguments and
Context. The constructor function of the specific message type is called directly. It is good practice to callValidateBasic()on the newly created message to run a sanity check and check for invalid arguments. - Depending on what the user wants, the transaction is either generated offline or signed and broadcasted to the preconfigured node using
GenerateOrBroadcastMsgs().
- The function should first initialize a
- Flags. Add any flags to the command. No flags were specified here, but all transaction commands have flags to provide additional information from the user (e.g. amount of fees they are willing to pay). These persistent transaction flags can be added to a higher-level command so that they apply to all transaction commands.
Finally, the module needs to have a GetTxCmd(), which aggregates all of the transaction commands of the module. Often, each command getter function has its own file in the module's cli folder, and a separate tx.go file contains GetTxCmd(). Application developers wishing to include the module's transactions will call this function to add them as subcommands in their CLI. Here is the auth GetTxCmd() function, which adds the Sign and MultiSign commands.
An application using this module likely adds auth module commands to its root TxCmd command by calling txCmd.AddCommand(authModuleClient.GetTxCmd()).
Queries allow users to gather information about the application or network state; they are routed by the application and processed by the module in which they are defined. Query commands typically have their own query.go file in the module x/moduleName/client/cli folder. Like transaction commands, they are specified in getter functions and have the prefix GetCmdQuery. Here is an example of a query command from the nameservice module:
This query returns the address that owns a particular name. The getter function does the following:
codecandqueryRoute. In addition to taking in the applicationcodec, query command getters also take aqueryRouteused to construct a path Baseapp uses to route the query in the application.- Construct the command. Read the Cobra Documentation and the transaction command example above for more information. The user must type
whoisand provide thenamethey are querying for as the only argument. RunE. The function should be specified as aRunEto allow for errors to be returned. This function encapsulates all of the logic to create a new query that is ready to be relayed to nodes.- The function should first initialize a new
Contextwith the applicationcodec. - If applicable, the
Contextis used to retrieve any parameters (e.g. the query originator's address to be used in the query) and marshal them with the query parameter type, in preparation to be relayed to a node. There are noContextparameters in this case because the query does not involve any information about the user. - The
queryRouteis used to construct apathbaseappwill use to route the query to the appropriate querier. The expected format for a querypathis "queryCategory/queryRoute/queryType/arg1/arg2/...", wherequeryCategorycan bep2p,store,app, orcustom,queryRouteis the name of the module, andqueryTypeis the name of the query type defined within the module.baseappcan handle each type ofqueryCategoryby routing it to a module querier or retrieving results directly from stores and functions for querying peer nodes. Module queries arecustomtype queries (some SDK modules have exceptions, such asauthandgovmodule queries). - The
ContextQueryWithData()function is used to relay the query to a node and retrieve the response. It requires thepath. It returns the result and height of the query upon success or an error if the query fails. - The
codecis used to nmarshal the response and theContextis used to print the output back to the user.
- The function should first initialize a new
- Flags. Add any flags to the command.
Finally, the module also needs a GetQueryCmd, which aggregates all of the query commands of the module. Application developers wishing to include the module's queries will call this function to add them as subcommands in their CLI. Its structure is identical to the GetTxCmd command shown above.
Flags are entered by the user and allow for command customizations. Examples include the fees or gas prices users are willing to pay for their transactions.
The flags for a module are typically found in a flags.go file in the ./x/moduleName/client/cli folder. Module developers can create a list of possible flags including the value type, default value, and a description displayed if the user uses a help command. In each transaction getter function, they can add flags to the commands and, optionally, mark flags as required so that an error is thrown if the user does not provide values for them.
For full details on flags, visit the Cobra Documentation.
For example, the SDK ./client/flags package includes a PostCommands() function that adds necessary flags to transaction commands, such as the from flag to indicate which address the transaction originates from.
Here is an example of how to add a flag using the from flag from this function.
cmd.Flags().String(FlagFrom, "", "Name or address of private key with which to sign")The input provided for this flag - called FlagFrom is a string with the default value of "" if none is provided. If the user asks for a description of this flag, the description will be printed.
A flag can be marked as required so that an error is automatically thrown if the user does not provide a value:
cmd.MarkFlagRequired(FlagFrom)Since PostCommands() includes all of the basic flags required for a transaction command, module developers may choose not to add any of their own (specifying arguments instead may often be more appropriate). For a full list of what flags are included in the PostCommands() function, including which are required inputs from users, see the CLI documentation here.
Applications typically support web services that use HTTP requests (e.g. a web wallet like Lunie.io. Thus, application developers will also use REST Routes to route HTTP requests to the application's modules; these routes will be used by service providers. The module developer's responsibility is to define the REST client by defining routes for all possible requests and handlers for each of them. It's up to the module developer how to organize the REST interface files; there is typically a rest.go file found in the module's ./x/moduleName/client/rest folder.
To support HTTP requests, the module developer needs to define possible request types, how to handle them, and provide a way to register them with a provided router.
Request types, which define structured interactions from users, must be defined for all transaction requests. Users using this method to interact with an application will send HTTP Requests with the required fields in order to trigger state changes in the application. Conventionally, each request is named with the suffix Req, e.g. SendReq for a Send transaction. Each struct should include a base request baseReq, the name of the transaction, and all the arguments the user must provide for the transaction.
Here is an example of a request to buy a name from the nameservice module:
The BaseReq includes basic information that every request needs to have, similar to required flags in a CLI. All of these values, including GasPrices and AccountNumber, will be provided in the request body. The user will also need to specify the arguments Name and Amount fields in the body and Buyer will be provided by the user's address.
BaseReq is a type defined in the SDK that encapsulates much of the transaction configurations similar to CLI command flags. Users must provide the information in the body of their requests.
Fromindicates which account the transaction originates from. This account is used to sign the transaction.Memosends a memo along with the transaction.ChainIDspecifies the unique identifier of the blockchain the transaction pertains to.AccountNumberis an identifier for the account.Sequenceis the value of a counter measuring how many transactions have been sent from the account. It is used to prevent replay attacks.Gasrefers to how much gas, which represents computational resources, Tx consumes. Gas is dependent on the transaction and is not precisely calculated until execution, but can be estimated by providing auto as the value forGas.GasAdjustmentcan be used to scale gas up in order to avoid underestimating. For example, users can specify their gas adjustment as 1.5 to use 1.5 times the estimated gas.GasPricesspecifies how much the user is willing pay per unit of gas, which can be one or multiple denominations of tokens. For example, --gas-prices=0.025uatom, 0.025upho means the user is willing to pay 0.025uatom AND 0.025upho per unit of gas.Feesspecifies how much in fees the user is willing to pay in total. Note that the user only needs to provide eithergas-pricesorfees, but not both, because they can be derived from each other.Simulateinstructs the application to ignore gas and simulate the transaction running without broadcasting.
Request handlers must be defined for both transaction and query requests. Handlers' arguments include a reference to the application's codec and the Context created in the user interaction.
Here is an example of a request handler for the nameservice module buyNameReq request (the same one shown above):
The request handler can be broken down as follows:
- Parse Request: The request handler first attempts to parse the request, and then run
SanitizeandValidateBasicon the underlyingBaseReqto check the validity of the request. Next, it attempts to parse the argumentsBuyerandAmountto the typesAccountAddressandCoinsrespectively. - Message: Then, a message of the type
MsgBuyName(defined by the module developer to trigger the state changes for this transaction) is created from the values and another sanity check,ValidateBasicis run on it. - Generate Transaction: Finally, the HTTP
ResponseWriter, applicationcodec,Context, requestBaseReq, and message is passed toWriteGenerateStdTxResponseto further process the request.
To read more about how a transaction is generated, visit the transactions documentation here.
The application CLI entrypoint will have a RegisterRoutes function in its main.go file, which calls the registerRoutes functions of each module utilized by the application. Module developers need to implement registerRoutes for their modules so that applications are able to route messages and queries to their corresponding handlers and queriers.
The router used by the SDK is Gorilla Mux. The router is initialized with the Gorilla Mux NewRouter() function. Then, the router's HandleFunc function can then be used to route urls with the defined request handlers and the HTTP method (e.g. "POST", "GET") as a route matcher. It is recommended to prefix every route with the name of the module to avoid collisions with other modules that have the same query or transaction names.
Here is a registerRoutes function with one query route example from the nameservice tutorial:
func RegisterRoutes(cliCtx client.Context, r *mux.Router, cdc *codec.LegacyAmino, storeName string) {
// ResolveName Query
r.HandleFunc(fmt.Sprintf("/%s/names/{%s}", storeName, restName), resolveNameHandler(cdc, cliCtx, storeName)).Methods("GET")
}A few things to note:
- The router
rhas already been initialized by the application and is passed in here as an argument - this function is able to add on the nameservice module's routes onto any application's router. The application must also provide aContextthat the querier will need to process user requests and the applicationcodecfor encoding and decoding application-specific types. "/%s/names/{%s}", storeName, restNameis the url for the HTTP request.storeNameis the name of the module,restNameis a variable provided by the user to specify what kind of query they are making.resolveNameHandleris the query request handler defined by the module developer. It also takes the applicationcodecandContextpassed in from the user side, as well as thestoreName."GET"is the HTTP Request method. As to be expected, queries are typically GET requests. Transactions are typically POST and PUT requests.
Read about the recommended module structure {hide}