arara version 6 released

After approximately one year, we are proud to announce the release of arara version 6. Just ready for TeX Live 2021, we have finished many exciting new features and are already prepared to work on a technically even more intriguing version 7.

Important changes

Please consult our changelog for a detailed view on all changes and references to our issue tracker. This post which is targeted at our users will only cover the most relevant changes.

More configurable and default preambles

Preambles are the user's way of not writing a rule yet reusing certain directives. Up to now, preambles have been prepended to the file regardless of it containing other directives. That means, if you have the following in your configuration:

preambles:
  twopdflatex: |
    % arara: pdflatex
    % arara: pdflatex

and your file.tex contains

% arara: lualatex

you might actually want to compile with lualatex. In version 5 and before, a call to arara -p twopdflatex file.tex would result in two compilations using pdflatex and one using lualatex.

If you do not want arara to perform these preamble compilations if there are directives present in your file, you may now set the following option in your arararc file:

prependPreambleIfDirectivesGiven: false

The default is still true to preserve backward compatibility.

Another major improvement for preambles is the ability to specify a defaultPreamble, i.e. you may now even omit -p preamble if you specified that key in your configuration file. This is especially useful if all files in your project shall be run with the same configuration. You may then set the following in your arararc:

preambles:
  twopdflatex: |
    % arara: pdflatex
    % arara: pdflatex
defaultPreamble: twopdflatex

and can call arara file.tex file2.tex file3.tex where arara will automatically run the default preamble on each of these files.

Combining those features is even more intriguing. If you insert a snippet like

preambles:
  twopdflatex: |
    % arara: pdflatex
    % arara: pdflatex
defaultPreamble: twopdflatex
prependPreambleIfDirectivesGiven: false

into your global arara configuration file, you can run arara file.tex on any file around there and it will either

That way, arara can be set as the default compiler in your editor and you can even compile before having inserted any arara directive.

Important note: Of course, this destroys the portability arara directives use to create. Please do not get lazy and specify directives in documents you want to share.

Passing command-line parameters in arara's evaluation

In previous versions, the only way to communicate with arara in rules and (with some trickery) directives were environment variables (as of version 5) and file I/O (since the beginning). This made automated builds only configurable with writing own rules which some users tend to avoid for simple directive patterns.

In version 6, we introduce a new command line parameter -P to allow specifying command-line options to be passed to arara's session map. Calling arara like

$ arara -P key=value file.tex

will allow you to write the following directives in file.tex:

% arara: pdflatex
% arara: pdflatex if (getSession().get('arg:key') == 'value')

Of course this is a poor example. But you get the gist of what is now possible using this command-line parameter. Imagine writing preambles in your arararc.yaml file that perform tests like getSession().get('arg:run') == 'fast') to leave certain compilations out for fast runs.

💡 A useful tip
Remember you can also use getSession().get('environment:PATH') and similar to get the PATH environment variable.

Expansion within options

The new command-line parameters are one of the few commands that have been made available within the options part of directives. This is especially useful to set output names or job names like

% arara: pdflatex: { options: ["-jobname=@{getSession().get("arg:key")}"] }

Apart from getSession you may use getBasename and getOriginalReference with their documented meanings in the options expansion.

A new safe mode

For quite some time, arara has been a very powerful tool. With some of the new changes, e.g. letting the user specify a full path anywhere on the file system to place the log file at, the wishes for a safe mode seemed very sensible.

The new safe mode may be enabled using the command-line flags -S or --safe-run. Currently, the following features are restricted:

Improving safety of flags and defaults in rules

In rules, you specify flags and defaults for parameters. We changed the handling of both. Let's start with default. Up to now, default could compute expensive operations and in many cases duplicated lots of code from the corresponding flag. Because of that, arara changed behavior and now treats the default like user input. To make that more clear, let's consider the example directive

% arara: pdflatex: {branch: stable}

where the stable branch is actually the default value. In previous versions, you would write (in your rule)

arguments:
- identifier: branch
  flag: >
    @{
        return [ 'stable' : 'pdflatex', 'developer' : 'pdflatex-dev' ]
                 .get(parameters.branch);
    }
  default: 'pdflatex'

which expresses that you want to run the stable branch by default but only because the rule author knows that if the user uses stable the flag will return pdflatex.

To express the same thing in v6, i.e. to use the stable branch of this rule by default, you would now write

arguments:
- identifier: branch
  flag: >
    @{
        return [ 'stable' : 'pdflatex', 'developer' : 'pdflatex-dev' ]
                 .get(parameters.branch);
    }
  default: 'stable'

which is very clear, even to the user who only knows the documented options of the rule. Furthermore, it now expresses the intention. As with user input, the flag will be called on the default value and determine the real value.

This change also implies that there is no orb tag expansion in the default field anymore.

The other notable change for flags is the transformation into a List<String>. That means for the return types of a flag:

Consider the argument example above: This branch argument actually returns a String, e.g. pdflatex. Hence, it is part of the second category. Internally, arara transforms this return value into ['pdflatex'], i.e. a list with one single element.

If you need interoperability of complex command or flag code with older versions, use isList(variable) ? variable[0] : variable to get the value of previously non-List values.

Rule updates

Some rules have been updated or added. New rules include ghostscript, copy, move, ltx2any, llmk, spix, pdfcrop and sage. If you encounter any issues, please open an issue at GitLab. We are also always open to support more tools in future versions.

Important note on rules: Those rules included in the core distribution have been renamed to have a unique prefix in the texmf tree. File names should not be relied upon (see deprecation notice in v5).

Being forgiving about language choices

Earlier last year, we started a poll about arara's localization and the users we have got feedback from stated they would be good with arara being focused on the English user interface. In version 5 we already removed the localized command-line interface/help, even without knowing about user feedback.

For this version, we decided to let arara continue running in English if an invalid language is specified. Previously, specifying a language like quack would cause arara to end the execution which does not seem appropriate anymore.

Another related change has been the switch to IETF BCP 47 codes for language selection. Please refer to the changelog for details.

Removal of the arara shorthand

As announced, the (as of version 5.0.0) deprecated <arara> shorthand notation has been removed. This is a breaking change. If your document previously used rules with

flag: <arara> return ['a', 'b'];

you will now have to write the more elaborate but much clearer

flag: >
  @{
    return ['a', 'b'];
  }

Fixed some minor issues

In arara version 5 and even before, the exit status handling did not conform to the documentation. We only differentiate the exit values 0, 1 and 2 as valid exit values for arara. Prior to this fix you have gotten the exit code of the failing command directly. For scripting purposes this could be a nice enhancement but we decided to aim for reestablishing documented behavior.

The correct behavior of currentFile and getOriginalReference has been restored as version 5 incorrectly used the same object for both.

The next steps

For version 7 of arara, i.e. TeX Live 2022 we have big plans. Not all of them may eventually end up in the released version but we still want to share our ideas with you.

Trying to serve more platforms

arara is written in Kotlin, a very versatile language. It has been designed with compilation to different platforms in mind. The usual base of operations for Kotlin is the JVM and the Java ecosystem. But one may also compile Kotlin to JavaScript or even native binaries.

In version 6, we already made a split between an API, a core implementation, the MVEL rules and the CLI application to separate concerns. This was the first step in the direction of splitting out components that are bound to one platform. Aditionally, we hope to enable other people to develop their own tools based on the arara libraries if they wish to. With multiple library components this should be easy.

For version 7, we want to take larger steps towards platform-independence. Hence, we will make use of Kotlin/Multiplatform (MPP) to use a single codebase for all target platforms. Some components will have to be rewritten, some need different interfaces for different platforms but in the end, we hope to provide an even more multitalented tool.

The real challenge is a major change of the API so that we are able to interpret rules that have been written for MVEL (which is JVM-only) to work on native platforms (Windows/Linux/Mac binaries) with respective libraries. We will have to put quite some effort into improving our API.

Last but not least, we want to drop some word about the background of these plans: arara has never known to be the fastest player among the TeX build tools but nevertheless performance is important to us. That, combined with some complaints (e.g. by a big player in the TeX world) about the reliance on the JVM, made us curious where we can take arara in that direction.

Exploring new grounds

Up to now, arara has been developed as a file-based build tool. The tool does not track dependencies nor does it manage projects. You throw a file at it and it makes the best out of it. That results in directives like the following snippet from file.tex

% arara: metapost: { files: [a.mp, b.mp] }
% arara: pdflatex

which in fact does not (only) specify how to compile file.tex but it specifies three things:

  1. how to compile a.mp,
  2. how to compile b.mp and
  3. how to compile file.tex.

In fact, we are doing project management by hand. Obviously, the author intended to compile the metapost files before running LaTeX. That is a clear indication of dependency management.

As we have been asked about make-like support of directive specifications out of the source file, we think the only sensible approach to tackle it is to actually introduce the concept of a project to kill two birds with one stone.

Actually, we already worked on project support in combination with a new Kotlin-based DSL (domain-specific language) for version 6. To avoid major problems, we decided not to ship it and reconsider our plans. However, we already had project specifications like

project("My awesome book") {
  file("a.mp") {
    directives = listOf("metapost")
  }
  file("b.mp") {
    directives = listOf("metapost")
  }
  file("file.tex") {
    dependsOn("a.mp", "b.mp")
    // directives omitted, take from file
  }
}

available, so stay tuned.