Interfaces for interacting with the shell

Exoskeleton is a Scala library to make it easy to write command-line applications, in particular those which need argument parsing, and tab completions, making it trivial to write interactive command-line applications in Scala. Tab completions are automatically provided for bash, zsh and fish.

• POSIX-style parameter parsing
• unified programmatic tab-completions for bash, zsh and fish
• easy to specify conditional tab-completions
• functional API for application entry points
• simple stream access to POSIX interrupts (signals)
• easy installation of completion scripts for each shell

Exoskeleton has a modular design, so you can use as much or as little of it as you like. The modules include:

• args for parsing command-line arguments
• core to provide an enhanced entry-point into an application (requires args)
• completions to provide tab-completions for an application (requires core)

In its most trivial usage, Exoskeleton provides an application wrapper to implement a main method:

import exoskeleton.*
import anticipation.*, turbulence.*, rudiments.*, gossamer.*

@main
def myapp(args: IArray[Text]): Unit = application(args):
  Out.println(t"Hello world!")
  ExitStatus.Ok

Packaging this as a fat JAR, with its Main-Class specified as myapp will produce an executable JAR which can be invoked with,

java -jar myapp.jar

The body of the application method makes certain contextual values available. For example, arguments provides access to the application's arguments. Note that the return value is ExitStatus.Ok, though different exit statuses can be specified with ExitStatus.Fail(1) or higher numbers; but the value must be specified.

We make the following distinction between arguments and parameters: the word "arguments" is used to describe a linear sequence of textual values, while we use "parameters" to describe the interpreted meaning of arguments. So, for example, the arguments of grep -rA4 pattern would be {-rA4, pattern} while its parameters would be some representation of {"search recursively", "4 lines of trailing content", "search for pattern"}.

Exoskeleton interprets arguments as parameters. Different interpreters may be used, though most users will want to use the POSIX parameters interpreter, parameterInterpretation.posix to get a PosixParameters. With the contextual parameterInterpretation.posix in scope, the parameters method can be used to get an instance of PosixParameters instance.

An instance of PosixParameters distinguishes between flags, which are arguments beginning with - or --, and operands—all other arguments, and consists of the following:

1. a list of positional operands appearing before the first flag, often considered as "subcommands"
2. a map from flags to a list of operands (being all the arguments up to the next flag)
3. a list of unparsed positional arguments which appear after a -- argument

While each argument is clearly a string, each is represented by the type Argument which encapsulates not just the argument's Text value, but its position on the command line. (The first argument after the command is numbered 0, whereas in scripting languages it's typically 1.) Retaining the argument number turns out to be useful later, when providing completions.

The ability to press the tab key with the cursor at some position in an incomplete command line is an extremely useful feature of modern command-line applications, and offers users a highly interactive experience, allowing them to discover new features, preview possible flags and operand values and avoid typographic errors by having values completed without typing.

Tab completions are widely available in Bash, ZSH and Fish, though the support available in each shell differs greatly. The support in bash is most simplistic, with no more information than the completion text itself provided to the user, whereas Fish an ZSH both allow a description to be associated with each completion value.

Each shell implements completions differently, so most command-line applications which provide completions do so through a different script for each shell, often written by different users and potentially having subtle differences from each other or from the command itself.

Completion scripts also differ in how dynamic their completions are. Most try to duplicate the behavior of the application by hardcoding subcommands, flags and some completion values, or try to approximate operand values by calling other shell commands. Many go further to ensure that flags remain self consistent; for example, if the -t flag and the -x flags cannot be used together, then -x would not be suggested as a completion of - if -t already appears on the command line.

This work is admirable, but it's difficult to maintain, and liable to mistakes or compromises.

Exoskeleton takes the view that the application itself should be the only source of truth for the completions of arguments to a command, and that to the greatest extent possible, the same code that interprets a command when it is invoked should be used to evaluate what completions should be provided.

This makes it possible for the complex logic of determining completions to be moved out of scripts written in Bash, and have it evaluated entirely in the application itself—as long as the application has access to all the information that would have been available to the completions script.

This greatly simplifies the functionality of completions scripts for Bash, ZSH and Fish. Exoskeleton provides a generic completion script for each, which is greatly simplified. It's sufficient for each script to capture the current (incomplete) arguments of the command line, the position of the cursor, and the type of shell, and to invoke the application in "completions mode", passing it this information so that it can produce output which can provide the completions for just the current argument.

Within the Scala code, a further innovation is employed to reuse as much logic as possible from interpreting the arguments in producing the completions, while making an important distinction between the part of the program which runs when completions are requested, and the part which runs when the user presses enter and the command is invoked.

An application with tab-completions must delimit the side-effecting part of the program in a special execute block which will be run for an invocation, but not for tab-completions. However, code which processes the parameters will be executed in both cases. This ensures that requesting completions will not cause unwanted side-effects, but offers an opportunity to have completions inferred from the "prelude" code which runs in both modes. This is described in detail in the next section.

In the prelude, the act of checking for a flag is sufficient to infer that flag as a possible completion for any "free" argument. Checking the value of a parameter may, of course, be performed only conditionally in a branch which is dependent on other parameters, and this would cause that parameter to be proposed only conditionally too, on precisely the same condition!

For example, imagine a command which has a -v flag to make it produce output, and a --verbosity flag which can be used only if -v is present. The prelude would check for -v, and if it's present, would check for --verbosity. The whole expression could return a value such as Maybe[Verbosity] which would determine the verbosity level for the rest of the program.

If a user pressed tab after a single - character at the end of the command line, it would invoke the prelude, checking first for -v, and (as a side-effect) adding it to the list of flag completions. If there is no -v already on the command-line, then it would be proposed for the completion of -. But if there is a -v already present, the branch which checks for --verbosity would also run, and (as a side-effect) it would be added to the set of known flags, and would be proposed as a completion of -. (Since it only makes sense for -v to appear once on the command-line, it would not be proposed as a completion in this instance.)

This simple example illustrates how the exact same logic could be used in both scenarios. But that logic can be as complex as necessary, and depend on any number of variables. But in all cases, completions will be guaranteed to mirror execution.

A standalone application or daemon will not process tab completions by default, but they can be "turned on" by importing the contextual value:

import executives.completions

Having this in scope when application of daemon is invoked will replace the default executive with the completions executive. The role of the executive is to determine the return type of that invocation's execution block, as well as the types of contextual values that are available within the block.

This distinction is important for keeping the simple æsthetics of the entry-point API, while imposing necessary requirements on the structure of the implementation.

Executives are powerful idea, but in practice, only one difference is crucial: with the standard executive, the return value must be an ExitStatus instance, while for the completions executive, the return value must be an instance of Execution, which can be constructed with a delimited execute block; and an execute block must return an ExitStatus.

Here is an illustration of the difference in code. Compare the standard executive,

@main
def myapp(): Unit = daemon:
  // side-effecting code goes here
  ExitStatus.Ok

with the completions executive:

import executives.completions

@main
def myapp(): Unit = daemon:
  // parameter processing code goes here
  execute:
    // side-effecting code goes here
    ExitStatus.Ok

Because the only way to construct an Execution is with an execute block, and an execute block can only be written by returning an ExitStatus value, the structure of a completions application (standalone or daemon) will always follow this structure. It's not easy to forget the execution block or an exit status because doing so will produce a complie error.

The completions executive also applies some restrictions in the "prelude" code which do not exist for the standard executive: standard output is not permitted, except inside an execute block. Under the completions executive, a Stdio instance, which is required by calls to Out.println and Err.println, is not made available in the context of daemon or application, since there is nowhere for output to be sent during evaluation of completions. The execute block provides an Stdio, though.

One further contextual value is provided in an execute block: an instance of Effectful. This type is defined in Exoskeleton, but is not used anywhere. But any user-defined method can require it as a using parameter, and it will ensure that that method can only be called from within an execute block, since the only way to obtain an Effectful instance is from within an execute block. This makes it very difficult to invoke an "effectful" method elsewhere, by accident.

Exoskeleton is classified as fledgling. For reference, Soundness projects are categorized into one of the following five stability levels:

• embryonic: for experimental or demonstrative purposes only, without any guarantees of longevity
• fledgling: of proven utility, seeking contributions, but liable to significant redesigns
• maturescent: major design decisions broady settled, seeking probatory adoption and refinement
• dependable: production-ready, subject to controlled ongoing maintenance and enhancement; tagged as version 1.0.0 or later
• adamantine: proven, reliable and production-ready, with no further breaking changes ever anticipated

Projects at any stability level, even embryonic projects, can still be used, as long as caution is taken to avoid a mismatch between the project's stability level and the required stability and maintainability of your own project.

Exoskeleton is designed to be small. Its entire source code currently consists of 777 lines of code.

Exoskeleton will ultimately be built by Fury, when it is published. In the meantime, two possibilities are offered, however they are acknowledged to be fragile, inadequately tested, and unsuitable for anything more than experimentation. They are provided only for the necessity of providing some answer to the question, "how can I try Exoskeleton?".

1. Copy the sources into your own project

   Read the fury file in the repository root to understand Exoskeleton's build structure, dependencies and source location; the file format should be short and quite intuitive. Copy the sources into a source directory in your own project, then repeat (recursively) for each of the dependencies.

   The sources are compiled against the latest nightly release of Scala 3. There should be no problem to compile the project together with all of its dependencies in a single compilation.

2. Build with Wrath

   Wrath is a bootstrapping script for building Exoskeleton and other projects in the absence of a fully-featured build tool. It is designed to read the fury file in the project directory, and produce a collection of JAR files which can be added to a classpath, by compiling the project and all of its dependencies, including the Scala compiler itself.

   Download the latest version of wrath, make it executable, and add it to your path, for example by copying it to /usr/local/bin/.

   Clone this repository inside an empty directory, so that the build can safely make clones of repositories it depends on as peers of exoskeleton. Run wrath -F in the repository root. This will download and compile the latest version of Scala, as well as all of Exoskeleton's dependencies.

   If the build was successful, the compiled JAR files can be found in the .wrath/dist directory.

Contributors to Exoskeleton are welcome and encouraged. New contributors may like to look for issues marked beginner.

We suggest that all contributors read the Contributing Guide to make the process of contributing to Exoskeleton easier.

Please do not contact project maintainers privately with questions unless there is a good reason to keep them private. While it can be tempting to repsond to such questions, private answers cannot be shared with a wider audience, and it can result in duplication of effort.

Exoskeleton was designed and developed by Jon Pretty, and commercial support and training on all aspects of Scala 3 is available from Propensive OÜ.

Exoskeleton is a library for interacting with shells, which are their exterior skeletons—or Exoskeletons.

In general, Soundness project names are always chosen with some rationale, however it is usually frivolous. Each name is chosen for more for its uniqueness and intrigue than its concision or catchiness, and there is no bias towards names with positive or "nice" meanings—since many of the libraries perform some quite unpleasant tasks.

Names should be English words, though many are obscure or archaic, and it should be noted how willingly English adopts foreign words. Names are generally of Greek or Latin origin, and have often arrived in English via a romance language.

The logo shows a simplistic and imaginary arthropod, with a pair of wings on each side; and an exoskeleton.

Exoskeleton is copyright © 2025 Jon Pretty & Propensive OÜ, and is made available under the Apache 2.0 License.