Marine Rust SDK
The marine-rs-sdk empowers developers to create services suitable for hosting on peers of the peer-to-peer network. Such services are constructed from one or more Wasm modules, which each are the result of Rust code compiled to the wasm32-wasi compile target, executable by the Marine runtime.
API
The procedural macros [marine]
and [marine_test]
are the two primary features provided by the SDK. The [marine]
macro can be applied to a function, external block or structure. The [marine_test]
macro, on the other hand, allows the use of the familiar cargo test
to execute tests over the actual Wasm module generated from the service code.
Function Export
Applying the [marine]
macro to a function results in its export, which means that it can be called from other modules or AIR scripts. For the function to be compatible with this macro, its arguments must be of the ftype
, which is defined as follows:
ftype
= bool
, u8
, u16
, u32
, u64
, i8
, i16
, i32
, i64
, f32
, f64
, String
ftype
= ftype
| Vec
<ftype
>
ftype
= ftype
| Record
<ftype
>
In other words, the arguments must be one of the types listed below:
one of the following Rust basic types:
bool
,u8
,u16
,u32
,u64
,i8
,i16
,i32
,i64
,f32
,f64
,String
a vector of elements of the above types
a vector composed of vectors of the above type, where recursion is acceptable, e.g. the type
Vec<Vec<Vec<u8>>>
is permissiblea record, where all fields are of the basic Rust types
a record, where all fields are of any above types or other records\
The return type of a function must follow the same rules, but currently only one return type is possible.
See the example below of an exposed function with a complex type signature and return value:
Function Export Requirements
wrap a target function with the
[marine]
macrofunction arguments must by of
ftype
the function return type also must be of
ftype
Function Import
The [marine]
macro can also wrap an extern
block. In this case, all functions declared in it are considered imported functions. If there are imported functions in some module, say, module A, then:
There should be another module, module B, that exports the same functions. The name of module B is indicated in the
link
macro (see examples below).Module B should be loaded to
Marine
by the moment the loading of module A starts. Module A cannot be loaded if at least one imported function is absent inMarine
.
See the examples below for wrapped extern
block usage:
Function import requirements
wrap an extern block with the function(s) to be imported with the
[marine]
macroall function(s) arguments must be of the
ftype
typethe return type of the function(s) must be
ftype
Structures
Finally, the [marine]
macro can wrap a struct
making possible to use it as a function argument or return type. Note that
only macro-wrapped structures can be used as function arguments and return types
all fields of the wrapped structure must be public and of the
ftype
.it is possible to have inner records in the macro-wrapped structure and to import wrapped structs from other crates
See the example below for wrapping struct
:
Structure passing requirements
wrap a structure with the
[marine]
macroall structure fields must be of the
ftype
the structure must be pointed to without preceding package import in a function signature, i.e
StructureName
but notpackage_name::module_name::StructureName
wrapped structs can be imported from crates
Call Parameters
There is a special API function marine_rs_sdk::get_call_parameters()
that returns an instance of the CallParameters
structure defined as follows:
CallParameters are especially useful in constructing authentication services:
MountedBinaryResult
Due to the inherent limitations of Wasm modules, such as a lack of sockets, it may be necessary for a module to interact with its host to bridge such gaps, e.g. use a https transport provider like curl. In order for a Wasm module to use a host's curl capabilities, we need to provide access to the binary, which at the code level is achieved through the Rust extern
block:
The above code creates a "curl adapter", i.e., a Wasm module that allows other Wasm modules to use the the curl_request
function, which calls the imported curl binary in this case, to make http calls. Please note that we are wrapping the extern
block with the [marine]
macro and introduce a Marine-native data structure MountedBinaryResult
as the linked-function return value.
Please not that if you want to use curl_request
with testing, see below, the curl call needs to be marked unsafe, e.g.:
since cargo does not access to the marine macro to handle unsafe.
MountedBinaryResult itself is a Marine-compatible struct containing a binary's return process code, error string and stdout and stderr as byte arrays:
MountedBinaryResult then can be used on a variety of match or conditional tests.
Testing
Since we are compiling to a wasm32-wasi target with ftype
constrains, the basic cargo test
is not all that useful or even usable for our purposes. To alleviate that limitation, Fluence has introduced the [marine-test]
macro that does a lot of the heavy lifting to allow developers to use cargo test
as intended. That is, [marine-test]
macro generates the necessary code to call Marine, one instance per test function, based on the Wasm module and associated configuration file so that the actual test function is run against the Wasm module not the native code.
To use the [marine-test]
macro please add marine-rs-sdk-test
crate to the [dev-dependencies]
section of Config.toml
:
Let's have a look at an implementation example:
We wrap a basic greeting function with the
[marine]
macro which results in the greeting.wasm moduleWe wrap our tests as usual with
[cfg(test)]
and import the marine test crate. Do not import super or the local crate.Instead, we apply the
[marine_test]
macro to each of the test functions by providing the path to the config file, e.g., Config.toml, and the directory containing the Wasm module we obtained after compiling our project withmarine build
. Moreover, we add the type of the test as an argument in the function signature. It is imperative that project build precedes the test runner otherwise the required Wasm file will be missing.The target of our tests is the
pub fn greeting
function. Since we are calling the function from the Wasm module we must prefix the function name with the module namespace --greeting
in this example case as specified in the function argument.
Now that we have our Wasm module and tests in place, we can proceed with cargo test --release.
Note that using the release
flag vastly improves the import speed of the necessary Wasm modules.
The same macro also allows testing data flow between multiple services, so you do not need to deploy anything to the network and write an Aqua app just for basic testing. Let's look at an example:
We wrap the
test
function with themarine_test
macro by providing named service configurations with module locations. Based on its arguments the macro defines amarine_test_env
module with an interface to the services.We create new services. Each
ServiceInterface::new()
runs a new marine runtime with the service.We prepare data to pass to a service using structure definition from
marine_test_env
. The macro finds all structures used in the service interface functions and defines them in the corresponding submodule ofmarine_test_env
.We call a service function through the
ServiceInterface
object.It is possible to use the result of one service call as an argument for a different service call. The interface types with the same structure have the same rust type in
marine_test_env
.
In the test_on_mod.rs
tab we can see another option — applying marine_test
to a mod
. The macro just defines the marine_test_env
at the beginning of the module and then it can be used as usual everywhere inside the module.
The full example is here.
The marine_test
macro also gives access to the interface of internal modules which may be useful for setting up a test environment. This feature is designed to be used in situations when it is simpler to set up a service for a test through internal functions than through the service interface. To illustrate this feature we have rewritten the previous example:
We access the internal service interface to construct an interface structure. To do so, we use the following pattern:
marine_test_env::$service_name::modules::$module_name::$structure_name
.We access the internal service interface and directly call a function from one of the modules of this service. To do so, we use the following pattern:
$service_object.modules.$module_name.$function_name
.In the previous example, the same interface types had the same rust types. It is limited when using internal modules: the property is true only when structures are defined in internal modules of one service, or when structures are defined in service interfaces of different services. So, we need to construct the proper type to pass data to the internals of another module.
Testing sdk also has the interface for Cargo build scripts. Some IDEs can analyze files generated in build scripts, providing code completion and error highlighting for code generated in build scripts. But using it may be a little bit tricky because build scripts are not designed for such things.
Actions required to set up IDE:
CLion:
in the
Help -> Actions -> Experimental Futures
enableorg.rust.cargo.evaluate.build.scripts
refresh cargo project in order to update generated code: change
Cargo.toml
and build from IDE or pressRefresh Cargo Project
in Cargo tab.
VS Code:
install
rust-analyzer
pluginchange
Cargo.toml
to let plugin update code from generated files
The update will not work instantly: you should build service to wasm, and then trigger build.rs
run again, but for the native target.
And here is the example of using this:
We create a vector of pairs (service_name, service_description) to pass to the generator. The structure is the same with multi-service
marine_test
.We check if we build for a non-wasm target. As we build this marine service only for
wasm32-wasi
and tests are built for native target, we can generatemarine_test_env
only for tests. This is needed because our generator depends on the artifacts fromwasm32-wasi
build. We suggest using a separate crate for using build scripts for testing purposes. It is here for simplicity.We pass our services, a name of the file to generate, and a path to the build script file to the
marine_test_env
generator. Just always usefile!()
for the last argument. The generated file will be in the directory specified by theOUT_DIR
variable, which is set by cargo. The build script must not change any files outside of this directory.We set up condition to re-run the build script. It must be customized, a good choice is to re-run the build script when .wasm files or
Config.toml
are changed.We import the generated file with the
marine_test_env
definition to the project.Do not forget to add
marine-rs-sdk-test
to thebuild-dependencies
section ofCargo.toml
.
Features
The SDK has two useful features: logger
and debug
.
Logger
Using logging is a simple way to assist in debugging without deploying the module(s) to a peer-to-peer network node. The logger
feature allows you to use a special logger that is based at the top of the log crate.
To enable logging please specify the logger
feature of the Fluence SDK in Config.toml
and add the log crate:
The logger should be initialized before its usage. This can be done in the main
function as shown in the example below.
In addition to the standard log creation features, the Fluence logger allows the so-called target map to be configured during the initialization step. This allows you to filter out logs by logging_mask
, which can be set for each module in the service configuration. Let's consider an example:
Here, an array called TARGET_MAP
is defined and provided to a logger in the main
function of a module. Each entry of this array contains a string (a target) and a number that represents the bit position in the 64-bit mask logging_mask
. When you write a log message request log::info!
, its target must coincide with one of the strings (the targets) defined in the TARGET_MAP
array. The log will be printed if logging_mask
for the module has the corresponding target bit set.
REPL also uses the log crate to print logs from Wasm modules. Log messages will be printed ifRUST_LOG
environment variable is specified.
Debug
The application of the second feature is limited to obtaining some of the internal details of the IT execution. Normally, this feature should not be used by a backend developer. Here you can see example of such details for the greeting service compiled with the debug
feature:
The most important information these logs relates to the allocate
/deallocate
function calls. The sdk.allocate: 4
line corresponds to passing the 4-byte user
string to the Wasm module, with the memory allocated inside the module and the string is copied there. Whereas sdk.deallocate: 0x110080 8
refers to passing the 8-byte resulting string Hi, user
to the host side. Since all arguments and results are passed by value, deallocate
is called to delete unnecessary memory inside the Wasm module.
Module Manifest
The module_manifest!
macro embeds the Interface Type (IT), SDK and Rust project version as well as additional project and build information into Wasm module. For the macro to be usable, it needs to be imported and initialized in the main.rs file:
Using the Marine CLI, we can inspect a module's manifest with marine info
:
Last updated
Was this helpful?