Skip to main content

Control, Scope And Visibility

In Aqua, the default namespace of a module is the file name and all declarations, i.e., data, services and functions, are public.

For example, the default.aqua file:

aqua
func foo() -> string:
<- "I am a visible foo func that compiles"
aqua
func foo() -> string:
<- "I am a visible foo func that compiles"

Which we compile with

sh
aqua -i aqua-scripts -o compiled-aqua
sh
aqua -i aqua-scripts -o compiled-aqua

to obtain Typescript wrapped AIR, default_foo.ts in the compiled-aqua directory:

typescript
import { FluenceClient, PeerIdB58 } from '@fluencelabs/fluence';
import { RequestFlowBuilder } from '@fluencelabs/fluence/dist/api.unstable';
import { RequestFlow } from '@fluencelabs/fluence/dist/internal/RequestFlow';
// Services
// Functions
export async function foo(client: FluenceClient, config?: {ttl?: number}): Promise<string> {
let request: RequestFlow;
const promise = new Promise<string>((resolve, reject) => {
const r = new RequestFlowBuilder()
.disableInjections()
.withRawScript(
`
(xor
(seq
(call %init_peer_id% ("getDataSrv" "-relay-") [] -relay-)
(xor
(call %init_peer_id% ("callbackSrv" "response") ["I am a visible foo func that compiles"])
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 1])
)
)
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 2])
)
`,
)
.configHandler((h) => {
h.on('getDataSrv', '-relay-', () => {
return client.relayPeerId!;
});
h.onEvent('callbackSrv', 'response', (args) => {
const [res] = args;
resolve(res);
});
h.onEvent('errorHandlingSrv', 'error', (args) => {
// assuming error is the single argument
const [err] = args;
reject(err);
});
})
.handleScriptError(reject)
.handleTimeout(() => {
reject('Request timed out for foo');
})
if(config && config.ttl) {
r.withTTL(config.ttl)
}
request = r.build();
});
await client.initiateFlow(request!);
return promise;
}
typescript
import { FluenceClient, PeerIdB58 } from '@fluencelabs/fluence';
import { RequestFlowBuilder } from '@fluencelabs/fluence/dist/api.unstable';
import { RequestFlow } from '@fluencelabs/fluence/dist/internal/RequestFlow';
// Services
// Functions
export async function foo(client: FluenceClient, config?: {ttl?: number}): Promise<string> {
let request: RequestFlow;
const promise = new Promise<string>((resolve, reject) => {
const r = new RequestFlowBuilder()
.disableInjections()
.withRawScript(
`
(xor
(seq
(call %init_peer_id% ("getDataSrv" "-relay-") [] -relay-)
(xor
(call %init_peer_id% ("callbackSrv" "response") ["I am a visible foo func that compiles"])
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 1])
)
)
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 2])
)
`,
)
.configHandler((h) => {
h.on('getDataSrv', '-relay-', () => {
return client.relayPeerId!;
});
h.onEvent('callbackSrv', 'response', (args) => {
const [res] = args;
resolve(res);
});
h.onEvent('errorHandlingSrv', 'error', (args) => {
// assuming error is the single argument
const [err] = args;
reject(err);
});
})
.handleScriptError(reject)
.handleTimeout(() => {
reject('Request timed out for foo');
})
if(config && config.ttl) {
r.withTTL(config.ttl)
}
request = r.build();
});
await client.initiateFlow(request!);
return promise;
}

Regardless of your output target, i.e. raw AIR or Typescript wrapped AIR, the default module namespace is default_foo and foo is the compiled function.

While this default approach is handy for single file, single module development, it makes for inefficient dependency management and unnecessary compilations for multi-module projects. The remainder of this section introduces the scoping and visibility concepts available in Aqua to effectively manage dependencies.

Managing Visibility With module and declare

By default, all declarations in a module, i.e., data, service and func, are public. With the module declaration, Aqua allows developers to create named modules and define membership visibility where the default visibility of module is private. That is, with the module declaration all module members are private and do not get compiled.

Let's create an export.aqua file like so:

aqua
module Export
func foo() -> string:
<- "I am Export foo"
aqua
module Export
func foo() -> string:
<- "I am Export foo"

When we compile export.aqua

sh
aqua -i aqua-scripts -o compiled-aqua
sh
aqua -i aqua-scripts -o compiled-aqua

nothing gets compiled as expected:

2021.09.02 11:31:41 [INFO] Aqua Compiler 0.2.1-219
2021.09.02 11:31:42 [INFO] Source /Users/user/aqua-scripts/export.aqua: compilation OK (nothing to emit)
2021.09.02 11:31:41 [INFO] Aqua Compiler 0.2.1-219
2021.09.02 11:31:42 [INFO] Source /Users/user/aqua-scripts/export.aqua: compilation OK (nothing to emit)

You can further check the output directory, compiled-aqua, in our case, for the lack of output files. Consequently, foo cannot be imported from other files. For example:

aqua
-- import.aqua
import "export.aqua"
func wrapped_foo() -> string:
res <- foo()
<- res
aqua
-- import.aqua
import "export.aqua"
func wrapped_foo() -> string:
res <- foo()
<- res

Results in compile failure since foo is not visible to import.aqua:

sh
6 func wrapped_foo() -> string:
7 res <- foo()
^^^==
Undefined arrow, available: HOST_PEER_ID, INIT_PEER_ID, nil, LAST_ERROR
8 <- res
sh
6 func wrapped_foo() -> string:
7 res <- foo()
^^^==
Undefined arrow, available: HOST_PEER_ID, INIT_PEER_ID, nil, LAST_ERROR
8 <- res

We can use declares to create visibility for a module namespace for consuming modules. For example,

aqua
-- export.aqua
module Export declares foo
func bar() -> string:
<- " I am MyFooBar bar"
func foo() -> string:
res <- bar()
<- res
aqua
-- export.aqua
module Export declares foo
func bar() -> string:
<- " I am MyFooBar bar"
func foo() -> string:
res <- bar()
<- res

in and by itself does not result in compiled Aqua:

sh
aqua -i aqua-scripts -o compiled-aqua -a
Aqua JS: node /Users/user/.nvm/versions/node/v14.16.0/lib/node_modules/@fluencelabs/aqua/aqua.js -i aqua-scripts -o compiled-aqua -a
Aqua JS:
Aqua JS: 2021.09.08 13:36:17 [INFO] Aqua Compiler 0.3.0-222
2021.09.08 13:36:21 [INFO] Source /Users/user/aqua-scripts/export.aqua: compilation OK (nothing to emit)
sh
aqua -i aqua-scripts -o compiled-aqua -a
Aqua JS: node /Users/user/.nvm/versions/node/v14.16.0/lib/node_modules/@fluencelabs/aqua/aqua.js -i aqua-scripts -o compiled-aqua -a
Aqua JS:
Aqua JS: 2021.09.08 13:36:17 [INFO] Aqua Compiler 0.3.0-222
2021.09.08 13:36:21 [INFO] Source /Users/user/aqua-scripts/export.aqua: compilation OK (nothing to emit)

But once we link from another module, e.g.:

aqua
import foo from "export.aqua"
func foo_wrapper() -> string:
res <- foo()
<- res
aqua
import foo from "export.aqua"
func foo_wrapper() -> string:
res <- foo()
<- res

We get the appropriate result:

2021.09.08 13:40:17 [INFO] Source /Users/user/aqua-scripts/export.aqua: compilation OK (nothing to emit)
2021.09.08 13:40:17 [INFO] Result /Users/user/compiled-aqua/import.ts: compilation OK (1 functions)
2021.09.08 13:40:17 [INFO] Source /Users/user/aqua-scripts/export.aqua: compilation OK (nothing to emit)
2021.09.08 13:40:17 [INFO] Result /Users/user/compiled-aqua/import.ts: compilation OK (1 functions)

in form of import.ts:

typescript
// compiled-aqua/import.ts
import { FluencePeer } from '@fluencelabs/fluence';
import {
ResultCodes,
RequestFlow,
RequestFlowBuilder,
CallParams,
} from '@fluencelabs/fluence/dist/internal/compilerSupport/v1';
// Services
// Functions
export function foo_wrapper(config?: {ttl?: number}) : Promise<string>;
export function foo_wrapper(peer: FluencePeer, config?: {ttl?: number}) : Promise<string>;
export function foo_wrapper(...args) {
let peer: FluencePeer;
let config;
if (args[0] instanceof FluencePeer) {
peer = args[0];
config = args[1];
} else {
peer = FluencePeer.default;
config = args[0];
}
let request: RequestFlow;
const promise = new Promise<string>((resolve, reject) => {
const r = new RequestFlowBuilder()
.disableInjections()
.withRawScript(
`
(xor
(seq
(call %init_peer_id% ("getDataSrv" "-relay-") [] -relay-)
(xor
(call %init_peer_id% ("callbackSrv" "response") [" I am MyFooBar bar"])
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 1])
)
)
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 2])
)
`,
)
.configHandler((h) => {
h.on('getDataSrv', '-relay-', () => {
return peer.connectionInfo.connectedRelay ;
});
h.onEvent('callbackSrv', 'response', (args) => {
const [res] = args;
resolve(res);
});
h.onEvent('errorHandlingSrv', 'error', (args) => {
const [err] = args;
reject(err);
});
})
.handleScriptError(reject)
.handleTimeout(() => {
reject('Request timed out for foo_wrapper');
})
if(config && config.ttl) {
r.withTTL(config.ttl)
}
request = r.build();
});
peer.internals.initiateFlow(request!);
return promise;
}
typescript
// compiled-aqua/import.ts
import { FluencePeer } from '@fluencelabs/fluence';
import {
ResultCodes,
RequestFlow,
RequestFlowBuilder,
CallParams,
} from '@fluencelabs/fluence/dist/internal/compilerSupport/v1';
// Services
// Functions
export function foo_wrapper(config?: {ttl?: number}) : Promise<string>;
export function foo_wrapper(peer: FluencePeer, config?: {ttl?: number}) : Promise<string>;
export function foo_wrapper(...args) {
let peer: FluencePeer;
let config;
if (args[0] instanceof FluencePeer) {
peer = args[0];
config = args[1];
} else {
peer = FluencePeer.default;
config = args[0];
}
let request: RequestFlow;
const promise = new Promise<string>((resolve, reject) => {
const r = new RequestFlowBuilder()
.disableInjections()
.withRawScript(
`
(xor
(seq
(call %init_peer_id% ("getDataSrv" "-relay-") [] -relay-)
(xor
(call %init_peer_id% ("callbackSrv" "response") [" I am MyFooBar bar"])
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 1])
)
)
(call %init_peer_id% ("errorHandlingSrv" "error") [%last_error% 2])
)
`,
)
.configHandler((h) => {
h.on('getDataSrv', '-relay-', () => {
return peer.connectionInfo.connectedRelay ;
});
h.onEvent('callbackSrv', 'response', (args) => {
const [res] = args;
resolve(res);
});
h.onEvent('errorHandlingSrv', 'error', (args) => {
const [err] = args;
reject(err);
});
})
.handleScriptError(reject)
.handleTimeout(() => {
reject('Request timed out for foo_wrapper');
})
if(config && config.ttl) {
r.withTTL(config.ttl)
}
request = r.build();
});
peer.internals.initiateFlow(request!);
return promise;
}

Of course, if we change import.aqua to include the private bar:

aqua
import bar from "export.aqua"
func bar_wrapper() -> string:
res <- bar()
<- res
aqua
import bar from "export.aqua"
func bar_wrapper() -> string:
res <- bar()
<- res

We get the expected error:

aqua
import bar from "export.aqua"
^^^===================
Imported file declares [foo], no bar declared. Try adding `declares *` to that file.
aqua
import bar from "export.aqua"
^^^===================
Imported file declares [foo], no bar declared. Try adding `declares *` to that file.

As indicated in the error message, declares * makes all members of the namespace public, although we can be quite fine-grained and use a comma separated list of members we want to be visible, such as declares foo, bar.

Scoping Inclusion With use and import

We already encountered the import statement earlier. Using import with the file name, e.g., import "export.aqua", imports all visible, i.e., public, members from the dependency. We can manage import granularity with the from modifier, e.g., import foo from "file.aqua", to limit our imports and subsequent compilation outputs. Moreover, we can alias imported declarations with the as modifier, e.g.,import foo as HighFoo, bar as LowBar from "export_file.aqua".

In addition to import, we also have the use keyword available to link and scope. The difference betweenuse and import is that use brings in module namespaces declared in the referenced source file. For example:

aqua
-- export.aqua
module ExportModule declares foo
func foo() -> string:
<- "I am a foo fighter"
aqua
-- export.aqua
module ExportModule declares foo
func foo() -> string:
<- "I am a foo fighter"

declares the ExportModule namespace and makes foo visible. We can now bring foo into scope by means of its module namespace ExportModule in our import file without having to (re-) declare anything:

aqua
-- import.aqua
use "export.aqua"
func foo -> string:
res <- ExportModule.foo()
<- res
aqua
-- import.aqua
use "export.aqua"
func foo -> string:
res <- ExportModule.foo()
<- res

This example already illustrates the power of use as we now can declare a local foo function rather than the foo_wrapper we used earlier. use provides very clear namespace separation that is fully enforced at the compiler level allowing developers to build, update and extend complex code bases with clear namespace separation by default.

The default behavior for use is to use the dependent filename if no module declaration was provided. Moreover, we can use the as modifier to change the module namespace. Continuing with the above example:

aqua
-- import.aqua
use "export.aqua" as RenamedExport
func foo() -> string:
-- res <- ExportModule.foo() --< this fails
res <- RenamedExport.foo()
<- res
aqua
-- import.aqua
use "export.aqua" as RenamedExport
func foo() -> string:
-- res <- ExportModule.foo() --< this fails
res <- RenamedExport.foo()
<- res