murex

GoDoc Go Report Card CodeBuild CircleCI codecov

About murex

murex is a shell, like bash / zsh / fish / etc. It follows a similar syntax to POSIX shells like Bash however supports more advanced features than you'd typically expect from a $SHELL.

It aims to be similar enough to traditional shells that you can retain most of your muscle memory, while not being afraid to make breaking changes where "bash-isms" lead to unreadable, hard to maintain, or unsafe code.

murex is designed for DevOps productivity so it isn't suited is high performance workloads beyond what you'd typically run in Bash (eg pipelines forked as concurrent processes).

A non-exhaustive list features would include:

The language employs a relatively simple syntax modelled loosely on functional and stack-based programming paradigms (albeit without the LISP-style nested parentheses that scare a lot of developers) while maintaining compatibility with POSIX-like shells (eg Bash) wherever sensible. For example, a program structure could look like the following:

command | command | [ index ] | if { command }

However where murex differs is that, while the pipe token, | is supported for compatibility with existing shells, murex idioms prefer -> used for readability as it clearly demonstrates the direction of the data flow. eg

command -> command -> [ index ] -> if { command }

The language supports multiple data types natively; such as JSON, YAML, CSV, S-Expressions and even loosely tabulated terminal output (eg ps, ls -l). This makes passing data through the pipeline and parsing output easier when dealing with more complex arrangements of data than a simple byte stream in traditional shells like Bash.

Interactive shell

Aside from murex being carefully designed with scripting in mind, the interactive shell itself is also built around productivity. To achieve this, we wrote our own readline library. Below is an example of that library in use:

asciicast

The above demo includes the following features of murex's bespoke readline library:

Concise yet predictable

Despite the amount of features added to shell, we have tried to keep the amount of "magic" to a minimum and follow a pretty standard structure so the language is predictable. However there are times when a little magic goes a long way. For example murex's support for complex data objects of differing formats is managed in the pipeline so you don't need to think about the data format when querying data from them.

open: file.csv  -> [ column_name ] # returns specific columns (or rows) in CSV file
open: file.json -> [ index ]       # returns specific items from JSON

The index function ([) alters its matching algorithm depending on the piped data type and open sets the data type depending on the file extension or MIME type.

Sometimes you will want less guesswork or just the robustness of a forced behavior. On those occasions you can remove one layer of magic by casting the data type:

open: file.txt -> cast csv  -> [ column_name ]
open: file.txt -> cast json -> [ index ]

This awareness of data structures is also utilised in foreach (which will cycle through each index in an array) and formap (key/value iteration against complex objects). See GUIDE.control-structures for more details on these and other control structures.

More robust scripts / shell one liners

murex employs a few methods to make shell scripting more robust:

Bash, for all it's power, is littered with hidden traps. The aim of murex is to address as many of them as we can without taking the flexibility or power away from the interactive command line. This is achieved through a couple of key concepts:

Everything is a function

The biggest breaking change from regular shells is how globbing isn't expanded by the shell by default. This is instead done by inlining functions as arrays:

# Bash
ls -l *.go

# Murex
ls -l @{g *.go}

The advantage of murex's method is that we can now offer other ways of matching file system objects that follows the same idiomatic pattern:

# Match files by regexp pattern
ls -l @{rx \.go$}

# Match only directories
ls -l @{f +d}

(more information on g, rx and f are available in GUIDE.quick-start).

However there will be occasions when you just want an inlined expansion (eg when using an interactive shell) and that can be achieved via the @g command prefix:

@g ls -l *.go

Powerful autocompletion

murex takes a slightly different approach to command line autocompletion, both from a usability perspective as well as defining completion rules.

Inspired by IDEs, murex queries man pages directly for flags as well as "tooltip" descriptions. Custom completions are defined via JSON meaning simple commands are much easier to define and complex commands can still fallback to using dynamic shell code just like they are in other shells.

This makes it easier to write completion rules as well as making the code more readable. An example of gits autocompletion definition:

private git-branch {
    # returns git branches and removes the current one from the list
    git branch -> [ :0 ] -> !match *
}

autocomplete set git { [{
    # define the top level flags
    "Flags": [
        "clone", "init", "add", "mv", "reset", "rm", "bisect", "grep",
        "log", "show", "status", "branch", "checkout", "commit", "diff",
        "merge", "rebase", "tag", "fetch", "pull", "push", "stash"
    ],

    # specify what values those flags support
    "FlagValues": {
        "init": [{
            "Flags": [ "--bare" ]
        }],
        "add": [{
            "IncFiles": true,
            "AllowMultiple": true
        }],
        "mv": [{
            "IncFiles": true
        }],
        "rm": [{
            "IncFiles": true,
            "AllowMultiple": true
        }],
        "checkout": [{
            "Dynamic": ({ git-branch }),
            "Flags": [ "-b" ]
        }],
        "merge": [{
            "Dynamic": ({ git-branch })
        }]
    }
}] }

murex also supports several different styles of completion suggestion "popups" to cater for different scenarios (demo above) as well as built in support for jumping to files within nested directories quickly and easily:

cat [ctrl+f]app.g[return]
# same as typing: cat config/app.go

Error handling

Like traditional shells, murex is verbose with errors by default with options to mute them. However murex also supports cleaner decision structures for when you want you want errors captured in a useful way:

try {
    # do something
}
catch {
    err: "Could not perform action"
}

As well as a saner if syntax:

# compare two strings
if { = `foo`==`bar` } then {
    out: "`foo` matched `bar`"
}

# check if command ran successfully
!if { foobar } then {
    err: "`foobar` could not be run"
}

Test and debugging frameworks

Unlike traditional shells, murex is designed with a test and debugging modes baked into the shell language. This means you can write tests against your shell scripts as part of the shell scripts themselves.

For example:

function: hello-world {
    test: define example {
        "StdoutRegex": (^Hello World$)
    }

    out: <test_example> "Hello Earth"
}

test: run { hello-world }

...will output:

Hello Earth
 Status  Definition Function                                           Line Col. Message
[FAILED] example    out                                                5    9    stdout: regexp did not match 'Hello Earth'

If test mode isn't enabled then any test commands are skipped without being executed so you can liberally include test cases throughout your functions without worrying about any performance impact.

murex also supports unit tests

For example:

test: unit function aliases {
    "PreBlock": ({
        alias ALIAS_UNIT_TEST=example param1 param2 param3
    }),
    "StdoutRegex": "([- _0-9a-zA-Z]+ => .*?\n)+",
    "StdoutType": "str",
    "PostBlock": ({
        !alias ALIAS_UNIT_TEST
    })
}

function: aliases {
    # Output the aliases in human readable format
    runtime: --aliases -> formap: name alias {
        $name -> sprintf: "%10s => ${esccli @alias}\n"
    } -> cast: str
}

test: run aliases

...will output:

 Status  Definition Function                                           Line Col. Message
[PASSED] (unit)     aliases                                            13   1    All test conditions were met

Language guides

  1. GUIDE.syntax is recommended first as it gives an overview if the shell scripting languages syntax and data types.
  2. GUIDE.type-system describes murex's type system. Most of the time you will not need to worry about typing in murex as the shell is designed around productivity.
  3. GUIDE.builtin-functions lists some of the builtin functions available for this shell.

Or if you're already a seasoned Bash developer then you read the Quick Start Guide, GUIDE.quick-start, to jump straight into using murex.

Install instructions

There are various ways you can load murex on to your system. See INSTALL for details.

CI/CD

murex makes heavy use of testing and CI/CD to ensure the latest builds are safe for use.

  1. Git pre-commit and pre-push files exist to help developers catch any regression errors before they even hit the feature branches.
  2. Each and every git push is validated against hundreds of distinct tests and race detectors (run in Circle CI). These tests are run 10 times to shake out any possible timing related bugs.

Each push to develop and master also creates a new docker container, lmorg/murex:develop and lmorg/murex:latest (for the master branch).

  1. Weekly automated builds are then run against the latest container. These builds are run in AWS CodeBuild and they generate the murex.rocks website and build pre-compiled binaries for download.

Known bugs / TODO

Please see GitHub's issue tracker: https://github.com/lmorg/murex/issues

This site is rebuilt weekly, the content is automatically generated from murex's source code. Last built on Mon Oct 19 06:12:17 UTC 2020 against commit 81299cb81299cbf18c05a5d38bdde93a5f5a3c5ff005b27. Downloadable murex binaries are also built weekly. Current version is 1.0.0000 BETA which has been verified against tests.