If you want to use Wasm in your own app, here is how you can get this working
quickly and easily. First, check to make sure you fit the pre-requisites,
then integrate the x/wasm module as described below, and finally, you
can add custom messages and queries to your custom Go/SDK modules, exposing
them to any chain-specific contract.
The pre-requisites of integrating x/wasm into your custom app is to be using
a compatible version of the Cosmos SDK, and to accept some limits to the
hardware it runs on.
| wasmd | Cosmos SDK |
|---|---|
| v0.22 | v0.45.0 |
| v0.21 | v0.42.x |
| v0.20 | v0.42.x |
| v0.19 | v0.42.x |
| v0.18 | v0.42.x |
| v0.17 | v0.42.x |
| v0.16 | v0.42.x |
| v0.15 | v0.41.x |
| v0.14 | v0.40.x |
| v0.13 | v0.40.0-rc3 |
| v0.12 | v0.40.0-rc3 |
| v0.11 | v0.39.1 |
| v0.10 | v0.39.1 |
| v0.9 | v0.38.3 |
| v0.8 | v0.38.3 |
| v0.7 | v0.38.3 |
We currently only support Intel/AMD64 CPUs and OSX or Linux. For Linux, the standard build
commands work for glibc systems (Ubuntu, Debian, CentOS, etc). If you wish to compile
for a muslc based system (like alpine), you need to compile a static library wasmvm locally
and compile go with the muslc build tag. Or just use the Dockerfile,
which builds a static go binary in an alpine system.
This limit comes from the Rust dll we use to run the wasm code, which comes
from wasmvm. There are open issues
for adding ARM support, and
adding Windows support.
However, these issues are not high on the roadmap and unless you are championing
them, please count on the current limits for the near future.
The simplest way to try out CosmWasm is simply to run wasmd out of the box,
and focus on writing, uploading, and using your custom contracts. There is
plenty that can be done there, and lots to learn.
Once you are happy with it and want to use a custom Cosmos SDK app,
you may consider simply forking wasmd. I highly advise against this.
You should try one of the methods below.
The simplest way to use wasmd is just to import x/wasm and wire it up
in app.go. You now have access to the whole module and you custom modules
running side by side. (But the CosmWasm contracts will only have access
to bank and staking... more below on customization).
The requirement here is that you have imported the standard sdk modules
from the Cosmos SDK, and enabled them in app.go. If so, you can just look
at wasmd/app/app.go
for how to do so (just search there for lines with wasm).
wasmd also comes with a custom ante handler that adds the TX position in the block into the context
and passes it to the contracts. In order to support this feature you would need to add our custom
ante handler into the ante handler chain as in: app/ante.go
Sometimes, however, you will need to copy x/wasm into your app. This should
be in limited cases, and makes upgrading more difficult, so please take the
above path if possible. This is required if you have either disabled some key
SDK modules in your app (eg. using PoA not staking and need to disable those
callbacks and feature support), or if you have copied in the core x/* modules
from the Cosmos SDK into your application and customized them somehow.
In either case, your best approach is to copy the x/wasm module from the
latest release into your application. Your goal is to make minimal changes
in this module, and rather add your customizations in a separate module.
This is due to the fact that you will have to copy and customize x/wasm
from upstream on all future wasmd releases, and this should be as simple
as possible.
If, for example, you have forked the standard SDK libs, you just want to
change the imports (from eg. github.com/cosmos/cosmos-sdk/x/bank to
github.com/YOUR/APP/x/bank), and adjust any calls if there are compiler
errors due to differing APIs (maybe you use Decimals not Ints for currencies?).
By the end of this, you should be able to run the standard CosmWasm contracts in your application, alongside all your custom logic.
Once you have gotten this integration working and are happy with the flexibility it offers you, you will probably start wishing for deeper integration with your custom SDK modules. "It sure is nice to have custom tokens with a bonding curve from my native token, but I would love to trade them on the exchange I wrote as a Go module. Or maybe use them to add options to the exchange."
At this point, you need to dig down deeper and see how you can add this
power without forking either CosmWasm or wasmd.
This is perhaps the easiest part. Let's say your native exchange module
wants to call into a token that lives as a CosmWasm module. You need to
pass the wasm.Keeper into your exchange.Keeper. If you know the format
for sending messages and querying the contract (exported as json schema
from each contract), and have a way of configuring addresses of supported
token contracts, your exchange code can simply call wasm.Keeper.Execute
with a properly formatted message to move funds, or wasm.Keeper.SmartQuery
to check balances.
If you look at the unit tests in x/wasm/internal/keeper,
it should be pretty straight forward.
If you want to let the contracts access your native modules, the first
step is to define a set of Messages and Queries that you want to expose,
and then add them as CosmosMsg::Custom and QueryRequest::Custom
variants. You can see an example of the bindings for Terra.
Once you have those bindings, use them to build a simple contact using much of the API. Don't worry too much about the details, this should be usable, but mainly you will want to upload it to your chain and use for integration tests with your native Cosmos SDK modules. Once that is solid, then add more and more complex contracts.
You will then likely want to add a mocks package so you can provide
mocks for the functionality of your native modules when unit testing
the contracts (provide static data for exchange rates when your contracts
query it). You can see an example of mocks for Terra contracts.
What these three steps provide is basically a chain-specific extension to the CosmWasm contract SDK. Any CosmWasm contract can import you library (bindings and mocks) and easily get started using your custom, chain-specific extensions just as easily as using the standard CosmWasm interfaces. What is left is actually wiring them up in your chain so they work as desired.
Note, in order to ensure that no one tries to run the contracts on an unsupported chain,
you will want to include a requires_XYZ directive in your bindings library, this will
mean that only blockchain apps that explicitly declare their support for the XYZ extensions
(please rename XYZ to your project name) will allow the contract to be uploaded, and others
get error messages upon upload, not while running a critical feature later on.
You just need to add a line like this
to your binding library to add the requirement to any contract that imports your bindings lib.
Before I show how this works, I want to remind you, if you have copied x/wasm,
please do not make these changes to x/wasm.
We will add a new module, eg. x/contracts, that will contain custom
bindings between CosmWasm contracts and your native modules. There are two entry points
for you to use. The first is
CustomQuerier,
which allows you to handle your custom queries. The second is
CustomEncoder
which allows you to convert the CosmosMsg::Custom(YourMessage) types to []sdk.Msg to be dispatched.
Writing stubs for these is rather simple. You can look at the reflect_test.go file to see this in action.
In particular, here we define a CustomQuerier,
and here we define a CustomHandler.
This code is responsible to take json.RawMessage from the raw bytes serialized from your custom types in rust and parse it into
Go structs. Then take these go structs and properly convert them for your custom SDK modules.
You can look at the implementations for the staking module to see how to build these for non-trivial
cases, including passing in the Keeper via a closure. Here we
encode staking messages.
Note that withdraw returns 2 messages, which is an option you can use if needed to translate into native messages.
When we handle staking queries
we take in a Keeper in the closure and dispatch the custom QueryRequest from the contract to the native Keeper interface,
then encodes a response. When defining the return types, note that for proper parsing in the Rust contract, you
should properly name the JSON fields and use the omitempty keyword if Rust expects Option<T>. You must also use
omitempty and pointers for all fields that correspond to a Rust enum, so exactly one field is serialized.
Once you have writen and tested these custom callbacks for your module, you need to enable it in your application.
The first step is to write an integration test with a contract compiled with your custom SDK to ensure it works properly,
then you need to configure this in app.go.
For the test cases, you must
define the supported feature set
to include your custom name (remember requires_XYZ above?). Then, when creating TestInput,
you can pass in your custom encoder and querier.
Run a few tests with your compiled contract, ideally exercising the majority of the interfaces to ensure that all parsing between the contract and
the SDK is implemented properly.
Once you have tested this and are happy with the results, you can wire it up in app.go.
Just edit the default NewKeeper constructor
to have the proper supportedFeatures and pass in the CustomEncoder and CustomQuerier as the last two arguments to NewKeeper.
Now you can compile your chain and upload your custom contracts on it.