Sylvia Framework

Sylvia is the old name meaning Spirit of The Wood.

Sylvia is the Roman goddess of the forest.

Sylvia is also a framework created to give you the abstraction-focused and scalable solution for building your CosmWasm Smart Contracts. Find your way into the forest of Cosmos ecosystem. We provide you with the toolset, so instead of focusing on the raw structure of your contract, you can create it in proper and idiomatic Rust and then just let cargo make sure that they are sound.

Learn more about sylvia in the book

Sylvia contract template

The Sylvia template streamlines the development of CosmWasm smart contracts by providing a project scaffold that adheres to best practices and leverages the Sylvia framework's powerful features. It's designed to help developers focus more on their contract's business logic rather than boilerplate code.

Learn more here: Sylvia Template on GitHub

The approach

CosmWasm ecosystem core provides the base building blocks for smart contracts - the cosmwasm-std for basic CW bindings, the cw-storage-plus for easier state management, and the cw-multi-test for testing them. Sylvia framework is built on top of them, so for creating contracts, you don't have to think about message structure, how their API is (de)serialized, or how to handle message dispatching. Instead, the API of your contract is a set of traits you implement on your contract type. The framework generates things like entry point structures, functions dispatching the messages, or even helpers for multitest. It allows for better control of interfaces, including validating their completeness in compile time.

Code generation

Sylvia macros generate code in the sv module. This means that every contract and interface macro call must be made in a separate module to avoid collisions between the generated modules.

Contract type

In Sylvia, we define our contracts as structures:

use cw_storage_plus::Item;
use cosmwasm_schema::cw_serde;
use sylvia::types::QueryCtx;
use sylvia::cw_std::ensure;

/// Our new contract type.
///
struct MyContract<'a> {
    pub counter: Item<'a, u64>,
}

/// Response type returned by the
/// query method.
/// 
#[cw_serde]
pub struct CounterResp {
    pub counter: u64,
}

#[entry_points]
#[contract]
#[sv::error(ContractError)]
impl MyContract<'_> {
    pub fn new() -> Self {
        Self {
            counter: Item::new("counter")
        }
    }

    #[sv::msg(instantiate)]
    pub fn instantiate(&self, ctx: InstantiateCtx, counter: u64) -> StdResult<Response> {
        self.counter.save(ctx.deps.storage, &counter)?;
        Ok(Response::new())
    }

    #[sv::msg(exec)]
    pub fn increment(&self, ctx: ExecCtx) -> Result<Response, ContractError> {
        let counter = self.counter.load(ctx.deps.storage)?;
        ensure!(counter < 10, ContractError::LimitReached);
        self.counter.save(ctx.deps.storage, &(counter + 1))?;
        Ok(Response::new())
    }

    #[sv::msg(query)]
    pub fn counter(&self, ctx: QueryCtx) -> StdResult<CounterResp> {
        self
            .counter
            .load(ctx.deps.storage)
            .map(|counter| CounterResp { counter })
    }
}

Sylvia will generate the following new structures:

pub mod sv {
    use super::*;

    struct InstantiateMsg {
        counter: u64,
    }

    enum ExecMsg {
        Increment {}
    }

    enum ContractExecMsg {
        MyContract(ExecMsg)
    }

    enum QueryMsg {
        Counter {}
    }

    enum ContractQueryMsg {
        MyContract(QueryMsg)
    }

    // [...]
}

pub mod entry_points {
    use super::*;

    #[sylvia::cw_std::entry_point]
    pub fn instantiate(
        deps: sylvia::cw_std::DepsMut,
        env: sylvia::cw_std::Env,
        info: sylvia::cw_std::MessageInfo,
        msg: InstantiateMsg,
    ) -> Result<sylvia::cw_std::Response, StdError> {
        msg.dispatch(&MyContract::new(), (deps, env, info))
            .map_err(Into::into)
    }

    // [...]
}

entry_points macro generates instantiate, execute, query and sudo entry points. All those methods call dispatch on the msg received and run proper logic defined for the sent variant of the message.

What is essential - the field in the InstantiateMsg (and other messages) gets the same name as the function argument.

The ExecMsg is the primary one you may use to send messages to the contract. The ContractExecMsg is only an additional abstraction layer that would matter later when we define traits for our contract. Thanks to the entry_point macro it is already being used in the generated entry point and we don't have to do it manually.

What you might notice - we can still use StdResult (so StdError) if we don't need ContractError in a particular function. What is important is that the returned result type has to implement Into<ContractError>, where ContractError is a contract error type - it will all be commonized in the generated dispatching function (so entry points have to return ContractError as its error variant).

Interfaces

One of the fundamental ideas of the Sylvia framework is the interface, allowing the grouping of messages into their semantical groups. Let's define a Sylvia interface:

pub mod group {
    use super::*;
    use sylvia::interface;
    use sylvia::types::ExecCtx;
    use sylvia::cw_std::StdError;

    #[cw_serde]
    pub struct IsMemberResp {
        pub is_member: bool,
    }

    #[interface]
    pub trait Group {
        type Error: From<StdError>;

        #[sv::msg(exec)]
        fn add_member(&self, ctx: ExecCtx, member: String) -> Result<Response, Self::Error>;

        #[sv::msg(query)]
        fn is_member(&self, ctx: QueryCtx, member: String) -> Result<IsMemberResp, Self::Error>;
    }
}

Then we need to implement the trait on the contract type:

use sylvia::cw_std::{Empty, Addr};
use cw_storage_plus::{Map, Item};

pub struct MyContract<'a> {
    counter: Item<'a, u64>,
    // New field added - remember to initialize it in `new`
    members: Map<'a, &'a Addr, Empty>,
}

impl group::Group for MyContract<'_> {
    type Error = ContractError;

    fn add_member(&self, ctx: ExecCtx, member: String) -> Result<Response, ContractError> {
        let member = ctx.deps.api.addr_validate(&member)?;
        self.members.save(ctx.deps.storage, &member, &Empty {})?;
        Ok(Response::new())
    }

    fn is_member(&self, ctx: QueryCtx, member: String) -> Result<group::IsMemberResp, ContractError> {
        let is_member = self.members.has(ctx.deps.storage, &Addr::unchecked(&member));
        let resp = group::IsMemberResp {
            is_member,
        };

        Ok(resp)
    }
}

#[contract]
#[sv::messages(group as Group)]
impl MyContract<'_> {
    // Nothing changed here
}

First, note that I defined the interface trait in its separate module with a name matching the trait name, but written "snake_case" instead of CamelCase. Here I have the group module for the Group trait, but the CrossStaking trait should be placed in its own cross_staking module (note the underscore). This is a requirement right now - Sylvia generates all the messages and boilerplate in this module and will try to access them through this module. If the interface's name is a camel-case version of the last module path's segment, the as InterfaceName can be omitted. F.e. #[sv::messages(cw1 as Cw1)] can be reduced to #[sv::messages(cw1)]

Then there is the Error type embedded in the trait - it is also needed there, and the trait bound here has to be at least From<StdError>, as Sylvia might generate code returning the StdError in deserialization/dispatching implementation. The trait can be more strict - this is the minimum.

Finally, the implementation block has an additional #[sv::messages(module as Identifier)] attribute. Sylvia needs it to generate the dispatching properly - there is the limitation that every macro has access only to its local scope. In particular - we cannot see all traits implemented by a type and their implementation from the #[contract] crate.

To solve this issue, we put this #[sv::messages(...)] attribute pointing to Sylvia what is the module name where the interface is defined, and giving a unique name for this interface (it would be used in generated code to provide proper enum variant).

Macro attributes

struct MyMsg;
impl CustomMsg for MyMsg {}

struct MyQuery;
impl CustomQuery for MyMsg {}

#[entry_point]
#[contract]
#[sv::error(ContractError)]
#[sv::messages(interface as Interface)]
#[sv::messages(interface as InterfaceWithCustomType: custom(msg, query))]
#[sv::custom(msg=MyMsg, query=MyQuery)]
#[sv::msg_attr(exec, PartialOrd)]
#[sv::override_entry_point(sudo=crate::entry_points::sudo(crate::SudoMsg))]
impl MyContract {
    // ...
    #[sv::msg(query)]
    #[sv::attr(serde(rename(serialize = "CustomQueryMsg")))]
    fn query_msg(&self, _ctx: QueryCtx) -> StdResult<Response> {
        // ...
    }
}

• sv::error is used by both contract and entry_point macros. It is necessary in case a custom error is being used by your contract. If omitted generated code will use StdError.

• sv::messages is the attribute for the contract macro. Its purpose is to inform Sylvia about interfaces implemented for the contract. If the implemented interface does not use a default Empty message response for query and/or exec then the : custom(query), : custom(msg) or : custom(msg, query) should be indicated.

• sv::override_entry_point - refer to the Overriding entry points section.

• sv::custom allows to define CustomMsg and CustomQuery for the contract. By default generated code will return Response<Empty> and will use Deps<Empty> and DepsMut<Empty>.

• sv::msg_attr forwards any attribute to the message's type.

• sv::attr forwards any attribute to the enum's variant.

Usage in external crates

What is important is the possibility of using generated code in the external code. First, let's start with generating the documentation of the crate:

cargo doc --document-private-items --open

This generates and opens documentation of the crate, including all generated structures. --document-private-item is optional, but it will generate documentation of non-public modules which is sometimes useful.

Going through the doc, you will see that all messages are generated in their structs/traits modules. To send messages to the contract, we can just use them:

use sylvia::cw_std::{WasmMsg, to_json_binary};

fn some_handler(my_contract_addr: String) -> StdResult<Response> {
    let msg = my_contract_crate::sv::ExecMsg::Increment {};
    let msg = WasmMsg::ExecMsg {
        contract_addr: my_contract_addr,
        msg: to_json_binary(&msg)?,
        funds: vec![],
    }

    let resp = Response::new()
        .add_message(msg);
    Ok(resp)
}

We can use messages from traits in a similar way:

let msg = my_contract_crate::group::QueryMsg::IsMember {
    member: addr,
};

let is_member: my_contract_crate::group::IsMemberResp =
    deps.querier.query_wasm_smart(my_contract_addr, &msg)?;

It is important not to confuse the generated ContractExecMsg/ContractQueryMsg with ExecMsg/QueryMsg - the former is generated only for contract, not for interfaces, and is not meant to be used to send messages to the contract - their purpose is for proper messages dispatching only, and should not be used besides the entry points.

Query helpers

To make querying more user-friendly Sylvia provides users with sylvia::types::BoundQuerier and sylvia::types::Remote helpers. The latter is meant to store the address of some remote contract. For each query method in the contract, Sylvia will add a method in a generated sv::Querier trait. The sv::Querier is then implemented for sylvia::types::BoundQuerier so the user can call the method.

Let's modify the query from the previous paragraph. Currently, it will look as follows:

let is_member = Remote::<OtherContractType>::new(remote_addr)
    .querier(&ctx.deps.querier)
    .is_member(addr)?;

Your contract might communicate with some other contract regularly. In such a case you might want to store it as a field in your Contract:

pub struct MyContract<'a> {
    counter: Item<'a, u64>,
    members: Map<'a, &'a Addr, Empty>,
    remote: Item<'a, Remote<'static, OtherContractType>>,
}

#[sv::msg(exec)]
pub fn evaluate_member(&self, ctx: ExecCtx, ...) -> StdResult<Response> {
    let is_member = self
        .remote
        .load(ctx.deps.storage)?
        .querier(&ctx.deps.querier)
        .is_member(addr)?;
}

Executor message builder

Sylvia defines the ExecutorBuilder type, which can be accessed through Remote::executor. It's generic over the contract type and exposes execute methods from the contract and every interface implemented on it through an auto-generated Executor traits. Execute messages of other contracts can be built with Remote as well by calling executor method. It returns a message builder that implements auto-generated Executor traits of all Sylvia contracts. Methods defined in the Executor traits constructs an execute message, which variant corresponds to the method name. The message is then wrapped in the WasmMsg, and returned once ExecutorBuilder::build() method is called.

use sylvia::types::Remote;
use other_contract::contract::OtherContract;
use other_contract::contract::sv::Executor;

let some_exec_msg: WasmMsg = Remote::<OtherContract>::new(remote_addr)
    .executor()
    .some_exec_method()?
    .build();

Using unsupported entry points

If there's a need for an entry point that is not implemented in Sylvia, you can implement it manually using the #[entry_point] macro. As an example, let's see how to implement replies for messages:

use sylvia::cw_std::{DepsMut, Env, Reply, Response};

#[contract]
#[entry_point]
#[sv::error(ContractError)]
#[sv::messages(group as Group)]
impl MyContract<'_> {
    fn reply(&self, deps: DepsMut, env: Env, reply: Reply) -> Result<Response, ContractError> {
        todo!()
    }
    // [...]
}

#[entry_point]
fn reply(deps: DepsMut, env: Env, reply: Reply) -> Result<Response, ContractError> {
    &MyContract::new().reply(deps, env, reply)
}

It is important to create an entry function in the contract type - this way, it gains access to all the state accessors defined on the type.

Overriding entry points

There is a way to override an entry point or to add a custom-defined one. Let's consider the following code:

#[cw_serde]
pub enum UserExecMsg {
    IncreaseByOne {},
}

pub fn increase_by_one(ctx: ExecCtx) -> StdResult<Response> {
    crate::COUNTER.update(ctx.deps.storage, |count| -> Result<u32, StdError> {
        Ok(count + 1)
    })?;
    Ok(Response::new())
}

#[cw_serde]
pub enum CustomExecMsg {
    ContractExec(crate::ContractExecMsg),
    CustomExec(UserExecMsg),
}

impl CustomExecMsg {
    pub fn dispatch(self, ctx: (DepsMut, Env, MessageInfo)) -> StdResult<Response> {
        match self {
            CustomExecMsg::ContractExec(msg) => {
                msg.dispatch(&crate::contract::Contract::new(), ctx)
            }
            CustomExecMsg::CustomExec(_) => increase_by_one(ctx.into()),
        }
    }
}

#[entry_point]
pub fn execute(
    deps: DepsMut,
    env: Env,
    info: MessageInfo,
    msg: CustomExecMsg,
) -> StdResult<Response> {
    msg.dispatch((deps, env, info))
}

It is possible to define a custom exec message that will dispatch over one generated by your contract and one defined by you. To use this custom entry point with contract macro you can add the sv::override_entry_point(...) attribute.

#[contract]
#[sv::override_entry_point(exec=crate::entry_points::execute(crate::exec::CustomExecMsg))]
#[sv::override_entry_point(sudo=crate::entry_points::sudo(crate::SudoMsg))]
impl Contract {
    // ...
}

It is possible to override all message types like that. Next to the entry point path, you will also have to provide the type of your custom message. It is required to deserialize the message in the multitest helpers.

Multitest

Sylvia also generates some helpers for testing contracts - it is hidden behind the mt feature flag, which has to be enabled.

It is important to ensure no mt flag is set when the contract is built in wasm target because of some dependencies it uses, which are not buildable on Wasm. The recommendation is to add an extra sylvia entry with mt enabled in the dev-dependencies, and also add the mt feature on your contract, which enables mt utilities in other contract tests. An example Cargo.toml:

[package]
name = "my-contract"
version = "0.1.0"
edition = "2021"

[lib]
crate-type = ["cdylib", "rlib"]

[features]
library = []
mt = ["sylvia/mt"]

[dependencies]
sylvia = "0.10.0"

# [...]

[dev-dependencies]
sylvia = { version = "0.10.0", features = ["mt"] }

And the example code:

#[cfg(test)]
mod tests {
    use super::*;
    use sylvia::multitest::App;

    #[test]
    fn counter_test() {
        let app = App::default();

        let owner = "owner";

        let code_id = contract::CodeId::store_code(&app);

        let contract = code_id.instantiate(3)
            .with_label("My contract")
            .call(&owner)
            .unwrap();

        let counter = contract.counter().unwrap();
        assert_eq!(counter, contract::CounterResp { counter: 3});

        contract.increment().call(&owner).unwrap();

        let counter = contract.counter().unwrap();
        assert_eq!(counter, contract::CounterResp { counter: 4});
    }
}

Note the contract module I am using here - it is a slight change that doesn't match the previous code - I assume here that all the contract code sits in the contract module to make sure it is clear where the used type lies. So if I use contract::something, it is something in the module of the original contract (most probably Sylvia-generated).

First of all - we do not use cw-multi-test app directly. Instead, we use the sylvia wrapper over it. It contains the original multi-test App internally, but it does it in an internally mutable manner which makes it possible to avoid passing it everywhere around. It adds some overhead, but it should not matter for testing code.

We are first using the CodeId type generated for every single Sylvia contract separately. Its purpose is to abstract storing the contract in the blockchain. It makes sure to create the contract object and pass it to the multitest.

A contract's CodeId type has one particularly interesting function - the instantiate, which calls an instantiation function. It takes the same arguments as an instantiation function in the contract, except for the context that Sylvia's utilities would provide.

The function doesn't instantiate contract immediately - instead, it returns what is called InstantiationProxy. We decided that we don't want to force users to set all the metadata - admin, label, and funds to send with every instantiation call, as in the vast majority of cases, they are irrelevant. Instead, the InstantiationProxy provides with_label, with_funds, and with_amin functions, which set those meta fields in the builder pattern style.

When the instantiation is ready, we call the call function, passing the message sender - we could add another with_sender function, but we decided that as the sender has to be passed every single time, we can save some keystrokes on that.

The thing is similar when it comes to execution messages. The biggest difference is that we don't call it on the CodeId, but on instantiated contracts instead. We also have fewer fields to set on that - the proxy for execution provides only the with_funds function.

All the instantiation and execution functions return the Result<cw_multi_test::AppResponse, ContractError> type, where ContractError is an error type of the contract.

Interface items in multitest

Trait declaring all the interface methods is directly implemented on the contracts Proxy type.

use contract::mt::Group;

#[test]
fn member_test() {
    let app = App::default();

    let owner = "owner";
    let member = "john";

    let code_id = contract::mt::CodeId::store_code(&app);

    let contract = code_id.instantiate(0)
        .with_label("My contract")
        .call(&owner);

    contract
        .add_member(member.to_owned())
        .call(&owner);

    let resp = contract
        .is_member(member.to_owned())

    assert_eq!(resp, group::IsMemberResp { is_member: true });
}

Generics

Interface

Defining associated types on an interface is as simple as defining them on a regular trait.

#[interface]
pub trait Generic {
    type Error: From<StdError>;
    type ExecParam: CustomMsg;
    type QueryParam: CustomMsg;
    type RetType: CustomMsg;

    #[sv::msg(exec)]
    fn generic_exec(
        &self,
        ctx: ExecCtx,
        msgs: Vec<CosmosMsg<Self::ExecParam>>,
    ) -> Result<Response, Self::Error>;

    #[sv::msg(query)]
    fn generic_query(&self, ctx: QueryCtx, param: Self::QueryParam) -> Result<Self::RetType, Self::Error>;
}

Generic contract

Generics in a contract might be either used as generic field types or as generic parameters of return types in the messages. When Sylvia generates the messages' enums, only generics used in respective methods will be part of a given generated message type.

Example of usage:

pub struct GenericContract<
    InstantiateParam,
    ExecParam,
    FieldType,
> {
    _field: Item<'static, FieldType>,
    _phantom: std::marker::PhantomData<(
        InstantiateParam,
        ExecParam,
    )>,
}

#[contract]
impl<InstantiateParam, ExecParam, FieldType>
    GenericContract<InstantiateParam, ExecParam, FieldType>
where
    for<'msg_de> InstantiateParam: CustomMsg + Deserialize<'msg_de> + 'msg_de,
    ExecParam: CustomMsg + DeserializeOwned + 'static,
    FieldType: 'static,
{
    pub const fn new() -> Self {
        Self {
            _field: Item::new("field"),
            _phantom: std::marker::PhantomData,
        }
    }

    #[sv::msg(instantiate)]
    pub fn instantiate(
        &self,
        _ctx: InstantiateCtx,
        _msg: InstantiateParam,
    ) -> StdResult<Response> {
        Ok(Response::new())
    }

    #[sv::msg(exec)]
    pub fn contract_execute(
        &self,
        _ctx: ExecCtx,
        _msg: ExecParam,
    ) -> StdResult<Response> {
        Ok(Response::new())
    }
}

Generics in entry_points

Entry points have to be generated with concrete types. Using the entry_points macro on the generic contract we have to specify the types that have to be used. We do that with entry_points(generics<..>):

#[cfg_attr(not(feature = "library"), entry_points(generics<SvCustomMsg, SvCustomMsg, SvCustomMsg>))]
#[contract]
impl<InstantiateParam, ExecParam, FieldType>
    GenericContract<InstantiateParam, ExecParam, FieldType>
where
    for<'msg_de> InstantiateParam: CustomMsg + Deserialize<'msg_de> + 'msg_de,
    ExecParam: CustomMsg + DeserializeOwned + 'static,
    FieldType: 'static,
{
    ...
}

The contract might define a generic type in place of a custom message and query. In such case we have to inform entry_points macro using custom:

#[cfg_attr(not(feature = "library"), entry_points(generics<SvCustomMsg, SvCustomMsg, SvCustomMsg>, custom(msg=SvCustomMsg, query=SvCustomQuery))]
#[contract]
#[sv::custom(msg=MsgT, query=QueryT)]
impl<InstantiateParam, ExecParam, FieldType, MsgT, QueryT>
    GenericContract<InstantiateParam, ExecParam, FieldType, MsgT, QueryT>
where
    for<'msg_de> InstantiateParam: CustomMsg + Deserialize<'msg_de> + 'msg_de,
    ExecParam: CustomMsg + DeserializeOwned + 'static,
    FieldType: 'static,
{
    ...
}

Generating schema

Sylvia is designed to generate all the code that cosmwasm-schema relies on - this makes it very easy to generate schema for the contract. Just add a bin/schema.rs module, which would be recognized as a binary, and add a simple main function there:

use cosmwasm_schema::write_api;

use my_contract_crate::contract::{ContractExecMsg, ContractQueryMsg, InstantiateMsg};

fn main() {
    write_api! {
        instantiate: InstantiateMsg,
        execute: ContractExecMsg,
        query: ContractQueryMsg,
    }
}

Road map

Sylvia is in the adoption stage right now, but we are still working on more and more features for you. Here is a rough roadmap for the coming months:

• Replies - Sylvia still needs support for essential CosmWasm messages, which are replies. We want to make them smart, so expressing the correlation between the sent message and the executed handler is more direct and not hidden in the reply dispatcher.
• Migrations - Another important message we don't support, but the reason is similar to replies - we want them to be smart. We want to give you a nice way to provide upgrading Api for your contract, which would take care of its versioning.
• IBC - we want to give you a nice IBC Api too! However, expect it to be a while - we must first understand the best patterns here.
• Better tooling support - The biggest Sylvia issue is that the code it generates is not trivial, and not all the tooling handles it well. We are working on improving user experience in that regard.

Troubleshooting

For more descriptive error messages, consider using the nightly toolchain (add +nightly argument for cargo)

• Missing messages from an interface on your contract - You may be missing the #[sv::messages(interface as Interface)] attribute.