cli-framework

CLI framework using ES Modules.

Overview

This project provides a Javascript framework for developing Command Line Interface (CLI) applications.

Key Features

• ES2015 module based
• Written in Typescript
• Minimal dependencies
• Support for building:
  • Single command CLIs with global arguments e.g. executable --<global_modifier_command_name> [argument]
  • Multiple command CLIs with sub-command based arguments e.g. executable <sub_command_name> [sub_command_arguments]
  • Multiple command CLIs with grouped sub-command based arguments e.g. executable <group_command_name> <member_sub_command_name> [member_sub_command_arguments]
  • A mix of the above!
• Support for both optional and positional arguments e.g. executable <sub_command_name> --<option_name>=<option_value> <positional_value>
• Support for multiple value options e.g. executable <sub_command_name> --<option_name>=<option_value_1> --<option_name>=<option_value_2>
• Support for multiple value positionals ("varargs") e.g. executable <sub_command_name> <positional_value_1> <positional_value_2>
• Dynamic plugin based import of commands using @flowscripter/esm-dynamic-plugins
• Core (but optional) commands for help, logging level, version management and plugin management.
• Core (but optional) services for colour output to stdout and stderr, user prompting and configuration management.
• Support for persisted configuration

Key Concepts

The key concepts are:

• A host application implements the CLI interface (a default NodeJS specific CLI implementation is provided).
• The CLI is responsible for:
  • maintaining a CommandRegistry and ensuring the Commands it provides are available to a Runner.
  • providing invocation arguments to the Runner which parses them and determines which Command to run.
  • maintaining a ServiceRegistry and ensuring the Services it provides are available in a Context.
  • running the specified Command providing it with the parsed arguments and a Context.
• Dynamic plugins (enabled by @flowscripter/esm-dynamic-plugins which makes use of ES2015 Dynamic Module Imports) are used to allow:
  • dynamic load and import of one or more CommandFactory implementations providing one or more Command implementations.
  • dynamic load and import of one or more ServiceFactory implementations providing one or more Service implementations.

The following high level class diagram illustrates these relationships:

Commands

All functionality for a CLI is implemented within one or more Commands. A Command declares:

• a name which is to be used in command line arguments to invoke it.
• a function to run when the command is invoked.

The sub-types of command are: GlobalCommand, GlobalModifierCommand, SubCommand and GroupCommand

Global Command

A GlobalCommand provides the ability to invoke functionality via a global argument and one optional value:

executable --<global_command_name>=[global_command_value]

Concrete examples:

myNetworkApp --help=connect
myNetworkApp --connect

A GlobalCommand also supports a short character alias which should always be an alphabetic ASCII character:

executable -<global_command_short_alias>=[global_command_value]

Concrete examples:

myNetworkApp -h=connect
myNetworkApp -c

Global Command Values

A GlobalCommand also supports the provision of a value consisting of:

• a type of either: number, boolean or string (the default).
• an optional set of valid value choices.
• an optional default value.
• whether the value is mandatory.

There are three ways in which global command values can be specified:

executable --<global_command_name>=<value>
executable --<global_command_name> <value>
executable -<global_command_short_alias> <value>

Concrete examples:

myNetworkApp --help=connect
myNetworkApp --help connect
myNetworkApp -h=connect

For boolean options, specifying the value as true is not required. All of the following set the value to true:

executable --<global_command_name>=true
executable --<global_command_name> true
executable --<global_command_name>
executable -<global_command_short_alias>=true
executable -<global_command_short_alias> true
executable -<global_command_short_alias>

Global Modifier Commands

A GlobalModifierCommand is invoked in a similar manner to a GlobalCommand.

Any number of GlobalModifierCommands can be specified as long as they are accompanied by a GlobalCommand, GroupCommand or SubCommand:

executable --<global_modifier_command_1_name> [global_modifier_command_1_arguments] \
           --<global_modifier_command_2_name> [global_modifier_command_2_arguments] \
           <sub_command_name> [sub_command_arguments]

A concrete example:

myNetworkApp --loglevel debug --config=config.json serve

where:

• loglevel is a global modifier command with a global command value of debug
• config is a global modifier command with a global command value of config.json
• serve is a sub-command.

Each GlobalModifierCommand will be executed before the single specified GlobalCommand, SubCommand or GroupCommand is executed. This behaviour allows GlobalModifierCommands to modify the context in which later commands run.

A GlobalModifierCommand defines a "run priority" which is used to determine the order of execution when multiple GlobalModifierCommands are specified.

Sub-Command

A SubCommand provides the ability to invoke functionality via specifying the command name followed by any number of option and positional arguments.

A sub-command is invoked as follows:

executable <sub_command_name> [sub_command_arguments]

A concrete example:

myNetworkApp serve --host=localhost

Sub-Command Arguments

Arguments for a SubCommand can take two forms: Option or Positional.

Common to both are the following features:

• a name which must consist of alphanumeric non-whitespace ASCII characters or _ and - characters. It cannot start with -.
• a type of either: number, boolean or string (the default).
• an optional set of valid value choices.

Options

An option argument also provides for:

• a short character alias for the option which should always be an alphabetic ASCII character.

• whether the option is mandatory.

• an optional default value.

• whether the option can be specified more than once e.g.:

  executable <sub_command_name> --<option_name_1>=foo --<option_name_1>=bar

There are four ways in which options can be specified:

executable <sub_command_name> --<option_name>=<option_value>
executable <sub_command_name> --<option_name> <option_value>
executable <sub_command_name> -<option_short_alias>=<option_value>
executable <sub_command_name> -<option_short_alias> <option_value>

For boolean options, specifying the value as true is not required. All of the following set the value to true:

executable <sub_command_name> --<boolean_option_name>=true
executable <sub_command_name> --<boolean_option_name> true
executable <sub_command_name> --<boolean_option_name>
executable <sub_command_name> -<boolean_option_short_alias>=true
executable <sub_command_name> -<boolean_option_short_alias> true
executable <sub_command_name> -<boolean_option_short_alias>

Positionals

A positional argument is specified by a value which appears at the correct position in the list of SubCommand arguments:

 executable <sub_command_name> <positional_1_value> <positional_2_value>

A concrete example:

myHelloWorldApp say hello

where:

• say is a sub-command.
• hello is the value for the first positional argument.

A positional argument also provides for "varargs" support (both optional and multiple) which allows for zero, one or more entries:

NOTE: Only one "varargs" positional can be defined and it must be the last positional expected for the command.

If "varargs" optional is set for positional_1, these are valid:

 executable <sub_command_name>
 executable <sub_command_name> <positional_1_value_1>

If "varargs" multiple is set for positional_1, these are valid:

 executable <sub_command_name> <positional_1_value_1>
 executable <sub_command_name> <positional_1_value_1> <positional_1_value_2>

If "varargs" optional AND multiple is set for positional_1, these are valid:

 executable <sub_command_name>
 executable <sub_command_name> <positional_1_value_1>
 executable <sub_command_name> <positional_1_value_1> <positional_1_value_2> <positional_1_value_3>

Group Command

A GroupCommand allows multiple member SubCommands to be grouped under a single named group. The name of the GroupCommand is specified before the desired member SubCommand in one of two ways:

executable <group_command_name> <member_sub_command_name> [member_sub_command_arguments]
executable <group_command_name>:<member_sub_command_name> [member_sub_command_arguments]

Concrete examples:

myNetworkApp utils ping --host=localhost
myNetworkApp utils:ping --host=localhost

A GroupCommand also provides for a command to be invoked BEFORE the specified sub-command. However, a GroupCommand does not support any arguments itself (apart from the member SubCommand name and its arguments).

Runner

Core CLI behaviour is provided by a Runner implementation which is responsible for parsing invocation arguments, determining which Command to run and then running it.

Default Command

The provided default implementation of Runner (DefaultRunner) supports specification of a default command which should be run if no command names are parsed on the command line. In this scenario, any arguments provided will be parsed as possible arguments for the default command as well as potential Global Modifier Commands.

Logic Overview

The following activity diagram illustrates the DefaultRunner logic:

Parser

The Runner defers to a Parser implementation which performs the actual argument parsing.

The following parsing rules apply for the provided DefaultParser implementation:

Arguments Must Follow Command

All arguments for a command are expected to FOLLOW the command i.e. this is NOT valid:

executable <sub_command_argument> <sub_command_name>

Arbitrary Option Order

The order of options for a particular command is not important i.e. these are equivalent:

executable <sub_command_name> --<option_1_name> <option_1_value> --<option_2_name> <option_2_value>
executable <sub_command_name> --<option_2_name> <option_2_value> --<option_1_name> <option_1_value>

Arbitrary Command Order

The order of commands is not important i.e. these are equivalent:

executable <sub_command_name> [sub_command_arguments] --<modifier_command_1> [modifier_command_1_arguments] \
           --<modifier_command_2> [modifier_command_2_arguments]
executable --<modifier_command_1> [modifier_command_1_arguments] <sub_command_name> [sub_command_arguments] \
           --<modifier_command_2> [modifier_command_2_arguments]

No Command Interleaving

Arguments for commands cannot be interleaved with other commands i.e. this is NOT valid:

executable --<modifier_command_name> <sub_command_name> [sub_command_arguments] [modifier_command_arguments]

Single Command

Apart from global modifier commands, there is expected to be only one command specified i.e. these are NOT valid:

executable <sub_command_1_name> [sub_command_1_arguments] <sub_command_2_name> [sub_command_2_arguments]
executable --<global_command_name> <sub_command_name> [sub_command_arguments]

Group Command

A group command name must always be following immediately by a container sub-command name i.e. these are NOT valid:

executable <member_sub_command_name> <group_command_name>
executable <group_command_name> <global_command_name> <member_sub_command_name>

Leading Arguments and Unused Trailing Arguments

Any leading arguments which appear BEFORE an identified command name are retained. Any trailing arguments which appear after an identified name and are not consumed when parsing the command arguments are also retained.

Once a command has been identified and parsed any retained arguments are considered unused and a warning is output.

If a command is NOT identified any retained arguments are considered potential arguments for a default command if it has been configured. This behaviour means the following are all equivalent:

executable <default_command_argument> --<modifier_command_name> <modifier_command_argument>
executable --<modifier_command_name> <modifier_command_argument> <default_command_argument>
executable --<modifier_command_1_name> <modifier_command_1_argument> <default_command_argument> \
           --<modifier_command_2_name> <modifier_command_2_argument>

Core Support

The following core services and commands are provided in the project and used within the provided BaseCLI and NodeCLI.

Core Services

Core service interfaces are provided together with a default implementation for printing output to the user, requesting input from the user and reading/writing configuration.

Printer Service

Two printer services are registered in the context under the IDs:

• STDOUT_PRINTER_SERVICE
• STDERR_PRINTER_SERVICE

Both of these provide the ability for commands to:

• output text at debug, info, warn and error levels together with a threshold filter
• add optional success, failure, alert or information icons
• enable or disable colour output and also print bold text
• show and hide a spinner

Prompter Service

The prompter service is registered in the context under the ID PROMPTER_SERVICE.

This provides the ability for commands to prompt for user input of a number, boolean, string, password or to select from a list of choices.

Configuration Service

The configuration service is registered in the context under the ID CONFIGURATION_SERVICE.

This service provides the ability to:

• read and write configuration data from a YAML file in the default location of $HOME/.<context.cliConfig.name>.yaml
• get and set config for commands and services

Note that in the default location, the name used will be stripped of all non-alphanumeric characters (except _ and -).

Core Commands

Core commands are provided to manage the output of the CLI, the location of configuration and to provide help to the user.

The dependencies these commands have on core services are outlined below.

Log Level Command

The LogLevel command overrides the default level threshold (INFO) of the printer services.

NOTE: This requires two Printer services in the context registered with the IDs STDOUT_PRINTER_SERVICE and STDERR_PRINTER_SERVICE.

No Color and Color Commands

The NoColor and Color commands override the auto-detected colour state of the printer services.

NOTE: These require two Printer services in the context registered with the IDs STDOUT_PRINTER_SERVICE and STDERR_PRINTER_SERVICE.

Config Command

The Config command overrides the default location ($HOME/.<context.cliConfig.name>.yaml) of the YAML configuration file.

NOTE: This requires a Printer service in the context registered with the ID STDERR_PRINTER_SERVICE and a Configuration service registered with the ID CONFIGURATION_SERVICE.

Version Command

The VersionCommand simply outputs the CLI application version to stdout.

NOTE: This requires a Printer service in the context registered with the ID STDOUT_PRINTER_SERVICE.

Usage Command

The UsageCommand provides a very simple help output. It is configured as the default command for the BaseCLI and NodeCLI.

NOTE: This requires a Printer service in the context registered with the ID STDOUT_PRINTER_SERVICE and it also requires provision of a HelpCommand.

Help Command

The HelpCommand outputs either generic help for the CLI (listing all commands available) or specific help for a command if it was specified as an argument.

NOTE: This requires a Printer service in the context registered with the ID STDOUT_PRINTER_SERVICE.

Plugin Support

Dynamic discovery and registration of @flowscripter/esm-dynamic-plugins plugins providing Command and Service implementations is supported.

Plugin Registry Service

The PluginRegistryService is responsible for instantiating a @flowscripter/esm-dynamic-plugins PluginManager which can then be used to discover available command and service plugins and add these to the ServiceRegistry and CommandRegistry.

Plugin Command

The PluginCommand is a group command providing AddCommand and RemoveCommand. These implement a VERY BASIC NPM functionality assuming that plugins are bundled packages i.e. there is no need to install their declared dependencies. Because of this the following is the case:

• No support is provided for multiple versions of the same package.
• All packages are installed at the top level or at scoped level i.e. there are no installations below other packages to accommodate multiple versions.

As well as this the following limitations currently exist:

• The installed set of packages is assumed to always be in a valid state and to have not been modified manually or by another process!
• The only dist-tag supported is latest.
• Package checksums are not verified!
• Git URLS are not supported.

NOTE: If this basic implementation doesn't suffice you can always use npm or yarn to install plugin packages.

Node CLI

The BaseCLI expects to be provided a CLIConfig object containing the CLI application's name, version and description together with streams to use for stdout and stderr.

The BaseNodeCLI implementation is a simple extension to the BaseCLI which provides the name, version and description from the current project's package.json and uses the NodeJS provided process.stdout and process.stderr streams.

SimpleSingleCommandNodeCLI, SimpleMultiCommandNodeCLI and AdvancedMultiCommandNodeCLI provide ready to go CLI classes with increasing levels of complexity. These can be used by simply implementing a Command interface and passing this into the constructor.

Example Projects

ts-example-cli is a demo CLI Typescript application based on this framework.

js-example-cli is a demo CLI Javascript application based on this framework.

ts-example-cli-plugin is a demo Typescript command plugin based on this framework.

js-example-cli-plugin is a demo Javascript command plugin based on this framework.

The plantuml-proxy-cli provides a simple real world single command CLI use case.

The Flowscripter cli provides a complex real world multiple command CLI use case. This implements custom commands for a REPL and a script executor.

Code Documentation

Typescript documentation

Development

NOTE: End-to-end testing is only performed on MacOS and Linux (not Windows).

Firstly:

npm install

then:

Build: npm run build (excuse the rollup warning, npm pacote library brings in a lot of cruft)

Watch: npm run watch

Test: npm test

Lint: npm run lint

Docs: npm run docs

Debug logging is implemented for the internals of the framework and can be enabled with the DEBUG environment variable e.g.:

DEBUG=* myHelloWorldApp say hello

The following diagram provides an overview of the main classes:

Further Details

Further details on project configuration files and Javascript version support can be found in the template for this project.

Alternatives

There are two popular alternatives available. Both are well documented and feature rich and you are encouraged to explore them:

• oclif
• Gluegun

The essence of a CLI framework implementation consists of:

1. A plugin mechanism: Neither of the above alternatives provided a dynamic, abstracted plugin import mechanism based on ES2015 modules:

   • oclif relies on available plugins being declared in package.json.
   • Gluegun supports dynamic loading of plugins based on a required folder structure.

2. Parsing input/printing output: to achieve this functionality, both of the above alternatives rely heavily on other CLI support packages under the hood e.g. yargs-parser or colors.js etc.

   • This is sensible, but both seem to suffer from feature creep as they move towards being a 'product' in their own right and both come with a growing list of dependencies. This project aims for a lightweight, abstracted service model so that almost all features are optional and customisable. It also aims to keep the core runtime clear of ANY behavioural logic e.g. reading from configs, logging, help etc. Where dependencies necessarily creep in, the resulting distributable size is minimised thanks to tree-shaking.

   • This project had some pretty specific requirements for argument parsing to accommodate the ideas of global modifier commands. yargs-parser was a bit over-powered and yet still didn't quite fit the bill, despite some serious trials. A lot of time was also spent with command-line-args but the need to hack in support for positional args caused endless pain.

Additional reasons for the existence of this project include:

• Despite the wonders of transpilers and bundlers, both alternatives mentioned above proved quite hard to get working in a native ES2015 application.
• There was no native ES2015 CLI framework - until now...

License

MIT © Flowscripter