Home

Awesome

ropm

A package manager for the Roku platform.

build status coverage status monthly downloads npm version license Slack

Installation

  1. Install Node.JS from here
  2. install the ropm node module globally
    npm i ropm -g
    

Discovering packages

Click here to search npm for packages with the ropm keyword.

You can also search GitHub for ropm packages, but since GitHub doesn't support searching by keyword, you'll need to know what you're looking for.

Creating a package

See the Creating ropm Packages section for guidance about creating ropm packages.

How it works

ropm leverages NodeJS's npm module system behind the scenes. This means when you create packages, they should be pushed to an npm registry such as npm, GitHub packages, or even an on-premise registry.

The Roku project structure is fairly strict. There are a few rules:

  1. Components must be stored somewhere underneath pkg:/components/.
  2. Component names must be unique across the entire project
  3. Components each create their own scope
  4. All files found underneath pkg:/source/ are compiled into a single scope.
  5. All functions in a scope must have a unique name

This provides unique challenges for a Roku package manager, because file paths alone are not enough to prevent symbol collisions. ropm solves the naming collision problem by rewriting the names of all functions and components in an ropm module.

ropm will create a roku_modules/<ropm module name> folder into each corresponding folder of your project. For example, if a ropm module named "logger" has the following folders:

then ropm install will create the following folders in your project

Sanitizing module names

Most npm-style package registries allow many characters in package names that are not valid identifiers within a Roku application. As such, these names need to be sanitized. The following transformations will be applied to every module name.

While extremely unlikely, this does have the potential for name collisions. If collisions occur, you will need to define a custom prefix for one of the dependencies in question.

Prefixes

When module authors publish their modules, they should not include any baseline prefix or namespace in front of their component or function names. The prefixing will be handled by ropm itself.

ropm will scan every module for:

Then, ropm will append a prefix to each of the previously-mentioned items. For example, consider the FancyKeyboards package:

xml

<component name="SimpleKeyboard" extends="Keyboard">
<AdvancedKeyboard />

becomes

<component name="FancyKeyboards_SimpleKeyboard" extends="fancyKeyboards_Keyboard">
<FancyKeyboards_AdvancedKeyboard />

brs

sub SetKeyboardLanguage(language)
    WriteToRegistry("KeyboardLanguage", language)
end sub
sub WriteToRegistry(key, value)
    '...
end sub
CreateObject("RoSGNode", "SimpleKeyboard")
node.CreateChild("AdvancedKeyboard")

becomes

sub FancyKeyboards_SetKeyboardLanguage(language)
    FancyKeyboards_WriteToRegistry("KeyboardLanguage", language)
end sub
sub FancyKeyboards_WriteToRegistry(key, value)
    '...
end sub
CreateObject("RoSGNode", "FancyKeyboards_SimpleKeyboard")
node.CreateChild("FancyKeyboards_AdvancedKeyboard")

WARNING: ropm does not currently support rewriting components created with ifSGNodeChildren's update() call (see the docs here). If you are a ropm package author and need support for this, please raise an issue.

Special case

Leading underscore

Some languages use leading underscore to represent private or internal. Ropm detects leading underscores, and starts the prefixing process at the first non-underscore character. For example:

sub _SecretKeyboardSub1()
end sub

sub ________SecretKeyboardSub2()
end sub

becomes

sub _FancyKeyboards_SecretKeyboardSub1()
end sub

sub ________FancyKeyboards_SecretKeyboardSub2()
end sub

File paths

ropm will rewrite file paths as well. These file paths will be rewritten based on the final location. ropm looks for:

Some packages may wish to look for files in predefined locations in the host application's folder structure (like a config file that should be placed at exactly pkg:/config/loggerConfig.json for example). In these situations, you will need to trick ropm into not rewriting your file path. The trick is to restrict the first part of the string to ONLY contain "pkg:/" and nothing else. ropm looks for pkg:/ and at least one additional character, so standalone "pkg:/" paths will be ignored. Here are a few examples:

sub GetImagePath(imageName)

   'will be rewritten because we have content after 'pkg:/'
   image1 = "pkg:/images/" + imageName

   'will not be rewritten because the 'pkg:/' is isolated
   image2 = "pkg:/" + "images/" + imageName

end sub

Result:

 sub CatPhotoLib_GetImagePath(imageName)

    'will be rewritten because we have content after 'pkg:/'
    image1 = "pkg:/images/roku_modules/CatPhotoLib/" + imageName

    'will not be rewritten because the 'pkg:/' is isolated
    image2 = "pkg:/" + "images/" + imageName

 end sub

Tasks (m.top.functionName)

BrightScript Task objects have a special m.top.functionName property that specifies which function should be run during the task. ropm will find all instances of m.top.functionName = "<anything>" and add the prefix to the beginning of the string.

The syntax must be exactly m.top.functionName = followed by a string (i.e. m.top.functionName = "taskCallback"). ropm will skip the statement if anything other than a string is found to the right-hand-side of the equals sign. If you dynamically generate the value for m.top.functionName, or assign it in some other fashion, consider using the ROPM_PREFIX source literal instead.

Never-prefixed functions

Due to their special handling within the Roku architecture, the following functions will never be prefixed:

ROPM_PREFIX source literal

ropm provides a source literal called ROPM_PREFIX which, during ropm install, gets replaced with a string literal containing your module's prefix. This enables developers to decide the best way to construct their own function name strings. Consider the following exmaple.

Your ropm module code:

sub init()
    m.top.functionName = getFunctionName()
end sub
function getFunctionName()
    if isLoggedIn()
        return ROPM_PREFIX + "initLoggedIn"
    else
        return ROPM_PREFIX + "initLoggedOut"
    end if
end function

After ropm install (the module prefix is "logger"):

sub init()
    m.top.functionName = getFunctionName()
end sub
function getFunctionName()
    if isLoggedIn()
        return "logger_" + "initLoggedIn"
    else
        return "logger_" + "initLoggedOut"
    end if
end function

Renaming modules

By default, ropm will install modules with their default names from the registry. For example, if you run ropm install roku-promise, then the ropm package name will be rokupromise. But what if you wanted to reference it as promise in your project? You can accomplish this by leveraging the flexibility of the package.json dependencies section. Here's the command to install roku-promise with an alternate name:

ropm install promise@npm:roku-promise

This will install the roku-promise library from the npmjs.com registry and call it promise.

Here's the resulting package.json dependencies section:

{
    "dependencies": {
        "promise": "npm:roku-promise@1.2.3"
    }
}

Semantic versioning

It is highly recommended that ropm package authors strictly adhere to the rules of Semantic Versioning. This will provide the most stability and consistency for consumers of your package, as well as provide the highest performance and smallest possible package size.

Version narrowing

Whenever ropm encounters a project that directly or indirectly requires multiple versions of a ropm module, ropm will attempt to minimize the number of versions of that package.

For example, consider a project that has these dependencies:

ropm will only install 1.4.0 since semantic versioning states that the only differences between 1.1.0 and 1.4.0 are new features and bugfixes, and should not contain breaking changes.

Here's another example:

ropm will only install 1.4.0 and 2.3.4 since those are the highest versions within the same major range.

Prerelease versions

Due to their unstable nature, prerelease versions of packages have special treatment in ropm. Each prerelease version will be considered a standalone package. Example:

Both versions will be copied to the project. We do not recommend publishing packages that depend on prerelease versions of a package.

Version prefixing

As previously mentioned, all packages will have prefixes applied to functions and components. All direct dependencies of a project (i.e. the packages listed in your app's package.json dependencies section) will use the exact dependency name listed. However, in order to resolve version conflicts and maintain consistency, any indirect dependencies (i.e. the dependencies of your dependencies) will be prefixed using the major version.

For example, consider the following dependency graph:

The prefixes will be as follows:

Disabling module prefixing

If you need to disable ropm's module prefixing, you can do this on a per-module basis by adding the names of of the desired modules to the ropm.noprefix key in package.json. Here's an example:

{
    "dependencies": {
        "roku-logger": "1.0.0"
    },
    "ropm": {
        "noprefix": [
            "roku-logger"
        ]
    }
}

Be sure you're using the npm alias of the package, not the original package name. The alias is the string on the left-hand-side of the dependency. For example:

{
    "dependencies": {
        "p": "npm:roku-promise@1.0.0"
    },
    "ropm": {
        "noprefix": [
            "p"
        ]
    }
}

In this example, the actual name of the package is "roku-promise", but we are using the npm alias "p". So "p" is what you should add to ropm.noprefix.

Here's another example:

{
    "dependencies": {
        "roku-promise": "1.0.0",
        "r": "npm:roku-requests@1.0.0"
    },
    "ropm": {
        "noprefix": [
            "roku-promise",
            "r"
        ]
    }
}

The npm aliases in this example is "roku-promise" and "r".

Do not use ropm.noprefix in published packages

Ropm will reject installing any ropm package that has the ropm.noprefix key in its package.json, so package authors should NOT use ropm.noprefix.

Do not change the code within roku_modules

The files and folders within the roku_modules folders should not be altered at all, as these changes could be erased by future ropm install commands. If there are issues with a ropm module you are using, consider reaching out to the module publisher to have them fix and release a new version.

Don't commit roku_modules

The roku_modules folders that ropm creates should not be commited to your project repository. Instead, developers should follow the practice of running ropm install anytime they fetch code from a repository. Here's how to ignore roku_modules in your .gitignore file:

.gitignore

roku_modules

The algorithm

Running ropm install executes the following operations for each package:

  1. For each folder in the ropm package's rootDir folder, delete ${rootDir}/<folder_name>/roku_modules/<package_name> and then copy all of the files from that corresponding folder.
  2. Rewrite function names, function calls, callfunc statements, and component names to have a prefix. (this prevents naming collisions)
  3. Rewrite in-code file references to point to their new locations

Configuring ropm

You can configure certain characteristics of ropm by specifying ropm options in the package.json.

rootDir

If you wish to install all ropm dependencies in a different location, then you should specify the rootDir ropm option in package.json. Here's an example

{
    "name": "your-app-name",
    "ropm": {
        //tell ropm that the files for this module reside in the "./src" folder
        "rootDir": "src"
    }
    ...
}

CLI commands

install

Install one or more packages locally in both node_modules and roku_modules. This also updates the local package.json dependencies section.

Examples:

ropm install roku-promise
ropm install module1 module2 module3

clean

Clean out all ropm_modules folders found in rootDir

Example:

ropm clean

copy

Similar to install, but does not fetch missing dependencies from the registry. This command should be faster than install as long as all the necessary dependencies are already downloaded. Example:

ropm copy

uninstall

Uninstall one or more packages from both node_modules and roku_modules. This also updates the local package.json dependencies section

Examples:

ropm uninstall roku-promise
ropm uninstall module1 module2 module3

Creating ropm Packages

Overview

Here is some overview information to help ropm package authors get started:

How to create a ropm package

The ropm package system leverages the npm package system from Node.js. Simply follow these instructions from npm on how to create a package.

Steps:

  1. Create a new package.json in your project (you can run ropm init to have it help build one)

  2. Add "ropm" to the keywords portion of the package.json. Without this tag, ropm will completely ignore your package when installed in an application folder.

  3. Ensure that all files are contained within a folder (preferably source/, /components, images/, and fonts/). Files at the root of a ropm package will be ignored.

    Here's a simple package.json showing how to add the ropm keyword:

    {
        "name": "pretty-list",
        "version": "0.0.1",
        "description": "",
        "keywords": ["ropm"]
        //...additional package.json properties
    }
    
  4. publish your package to a registry using the instructions from your registry of choice (npm, GitHub Packages)

Changing where the module's files are copied from (As a package author)

By default, ropm will copy every file from the root of your module (the folder where package.json resides), with a few exceptions:

These files will always be excluded (not copied):

All folders named roku_modules that are found in a ropm module will be ignored. This is due to the fact that modules should not be publishing their own copies of their ropm modules. ropm will handle this for them. So as a package author, be sure to exclude all folders named roku_modules during your publishing process.

We recommend excluding extraneous files when you create your package using npm's files property. In certain situations, you may want to store your files in a subdirectory. In this situation, you should use the ropm.packageRootDir property to specify the relative path to the root of your module. This pattern is useful when you want to publish documentation or other extraneous files with your package, but don't want those files to be included when published to a Roku device.

Here's an example (NOTE: comments are included here for explanation purposes but are invalid within an package.json)

{
    "name": "quick-list"
    //when npm creates the package, only include these files
    "files": [
        "src/**/*",
        "instructions/**/*"
    ],
    "ropm": {
        //tell ropm that the files for this module reside in the "./src" folder
        "packageRootDir": "src"
    }
}

Syntax parsing using BrighterScript

ropm uses the BrighterScript parser to find and apply the prefixes to ropm modules. While fairly comprehensive, there are still a few BrightScript syntax bugs that BrighterScript does not yet support, and will therefore cause your package to fail during install. Please open an issue if you encounter any syntax issues that prevent your package from working properly.

Finding function references is a package-wide operation

HINT: do not give local variables the same name as any function in your package

Function reference replacement is package-wide and does not operate on a per-scope basis, meaning ropm could add prefixes to local variables that share a name with any function across your entire package. We recommend that you do not give local variables the same name as any function your package.

Component interface handling

SceneGraph components can declare interface functions which will be callable via the callFunc function on a node. For example, the following LoggerComponent exposes a function called doSomething:

<?xml version="1.0" encoding="utf-8" ?>
<component name="LoggerComponent">
    <script uri="LoggerComponent.brs" />
    <interface>
        <function name="doSomething"/>
    </interface>
</component>

ropm will not rename functions referenced by component interfaces because prefixing those functions would change the public API of declared components. This does introduce a small risk for function name collisions, but those risks can be avoided if you adhere to the following guidelines:

rootDir versus packageRootDir

BrightScript in XML CDATA blocks is unsupported

It is considered bad practice to insert BrightScript code into <![CDATA[ xml script blocks, and as such, ropm does not support CDATA blocks.

Any BrightScript code found in CDATA blocks will be ignored by the ropm prefixing logic, so use at your own risk (or peril!).

Handling observeField

ropm will auto-detect most common observeField function calls. In order to prevent naming conflicts, please do not use the name observeField for custom object functions.

Here are the requirements for having ropm prefix your observeField string function names.

  1. Use a single string literal for the function name. For example, m.top.observeField(fieldName, "callbackFunction")
  2. The observeField call must be on a single line. For example, this call would remain unprefixed:
    m.top.observeField(getFieldName({
        componentName: "something"
    }, "callbackFunction")
    

Sponsors

image

Changelog

Click here to view the changelog.