Frequently Asked Questions

  1. Basic Questions
  2. Using Augeas
  3. Describing file formats

Basic Questions

What license is Augeas ditributed under ?
Augeas is distributed under the GNU Lesser General Public License (LGPL) A copy of the license is included in the file COPYING in the source distribution.
Can I use Augeas from a proprietary application ?
Yes. The LGPL allows you to use Augeas from a proprietary application. It would be very helpful to both you and the community if you send bug fixes and improvements as patches for possible incorporation in the main development tree. It will decrease your maintenance costs anyway if you do so.
How do I get Augeas ?
The download page explains where you can find released RPM's and tarballs and how you can check the latest development version out from Mercurial
What's with that name ?
To quote Wikipedia: "In Greek mythology, Augeas ... was King of Elis ... He is best known ... for his stables, which housed the single greatest number of cattle in the country and had never been cleaned."
How do I contribute to Augeas ?
The main means of communication is the Augeas mailing list For more information on contributing, see the contribute page
I don't know any C — can I still help ?
Of course. Besides writing code, there are many more areas where Augeas needs help: evaluating existing tree representations of configuration files, writing descriptions for new file formats, and documenting Augeas' tree structure are all areas where help is sorely needed.
Can I use Augeas for other text files ?
Yes, there is actually nothing specific to configuration files in Augeas. If you have some other unstructured text file that you'd like to transform into a tree, all you need to do is write a file format description for it.
Can I change a file with Augeas and still edit it by other means ?
Yes. One of the main design goals of Augeas is to not interfere with any other ways in which you might change your configuration files. Augeas works very hard to preserve comments and formatting details such as the number of spaces used to separate fields etc. It also does not depend on any additional information embedded in comments — you can always alternate between changing a configuration file with your favorite editor or custom script, and changing it with Augeas.

Using Augeas

Can I use Augeas in a shell script ?
Yes, the augtool command can be used to manipulate the tree and the underlying files with shell scripts. Run augtool help for a list of commands augtool supports.
What language bindings are available ?
Bindings for Ruby, Python, OCaml, Perl, Haskell, PHP, Lua, and Java are available for download. Patches/additions for other languages are eagerly anticipated.
Do I need to know how file formats are described ?
No, that is only necessary if you want to change how files are mapped into the tree or map new files. For using Augeas, it is enough to understand the functionality provided by augtool, or the equivalent calls in C or the language bindings.
How is the Augeas tree structured ?
Each node in the Augeas tree consists of three pieces of data: a label and a value, both strings, and a list of child nodes. The label and value are both strings. The label is used in the path to the node and child nodes, much as file names are used in the paths to files in a file system. The label must not contain the characters / or * or end with [N] where N is a positive number.
Can a node have multiple children with the same label ?
Yes, that is the biggest difference between the Augeas tree and files in a filesystem (and makes it similar to DOM's for XML documents). A node can have any number of children called foo; in paths, a specific child can be picked out with foo[N] where N is a positive number – the first child with label foo is picked with foo[1], and the last can be picked with foo[last()]. Just foo in a path picks all the children with that label. This convention follows XPath
What kind of data can be stored in the tree ?
Augeas only stores strings in the tree. So far, there hasn't been a compelling reason to support other data types.
Where do a find the contents of a file ?
The contents of a file /some/path are mapped into the tree as /files/some/path, which is usually a fine-grained tree view of that file; additional (meta-)information about each file is available in /augeas/files/some/path.

Describing file formats

What are lenses ?
Lenses are the building blocks of the file <-> tree transformation; they combine parsing a file and building the tree, the get transformation, with turning the tree back into an (updated) file, the put transformation.
Why are they called lenses ?
The name comes from the Harmony project. In Harmony's lingo, Augeas maps back and forth between a concrete view (the configuration file) and an abstract view (the tree representation). And calling something that goes back and forth between two views a 'lens' seems only natural, doesn't it ?
What is a bidirectional language ?
A bidirectional language is one where the program expresses a transformation from input to output, and from (possibly modified) output back to the corresponding input. They are called bidirectional rather than bijective because there are generally many inputs for the same output and vice versa. In Augeas case, many input files (e.g., /etc/hosts with varying amounts of whitespace between fields) map to the same tree, and there are many ways to map the same tree back to a file. Other examples of bidirectional languages are Harmony (mapping trees to trees) and Boomerang (mapping strings to strings), both from the Harmony project.
How is a file parsed ?
The lenses match regular expressions, different lenses do different things with the strings that match (create a new tree node with a certain label, store a value in a tree node, combine tree nodes into a larger tree); writing the regular expressions is the main task when describing a file format.
What is the regular expression syntax ?
Augeas uses extended POSIX regexps. There are a few more esoteric features of extended POSIX regexps that are not supported, the most notable of them being POSIX named character classes, collating elements and back references. Named character classes should be supported soon (patches welcome).
What is the language in which file format descriptions are written ?
The file format descriptions are written in a small subset of ML (a popular functional language.) If you don't know ML, don't despair: there are only two or three constructs that you need to know to write file format descriptions. The Augeas language is statically typed; besides the 'usual' checks of a type checker, checks also make sure that the transformations described by lenses are safe.
How do I test my file format description ?
The Augeas language has a built-in test construct for running unit tests; additional round-trip tests can be found in the tests/ directory in the source distribution. They are run by the augtest command in the same directory.
Where do I find existing file format descriptions ?
File format descriptions are contained in files with the extension .aug. In the source distribution, they are in the lenses directory, which, by default, is installed into /usr/share/augeas/lenses.
Can I send you my file format description ?
Yes, please. One of Augeas' goals is to collect file format descriptions, and build a library of them, so that eventually there will be a 'canonical' tree representation of configuration data.

libfa — Finita Automata

What is libfa ?
Libfa is a library for manipulating finite automata. A finite automaton is a representation of all strings matching a regular expression. Augeas uses it for typechecking lenses and ensuring they don't contain ambiguities.
Can I use libfa outside of Augeas ?
Yes. Libfa is also licensed under the LGPL, and has no dependencies on Augeas.
What operations does libfa support ?
A list of operations can be found here.
What regular expression syntax does libfa support ?
It supports extended POSIX regular expressions, exactly the same as Augeas.
Why is there no operation to match a regular expression against a string ?
The GNU C library (and GNU Regex) contains a much better matcher than I would ever hope to write for libfa — being able to use that matcher is the main reason for using extended POSIX syntax.