SCons Documentation Toolchain
This text tries to give an overview of the current SCons documentation toolchain. The interested user should be able to better understand where and how the text he writes is processed. It is also a reference for core developers and the release team.
The diagram above roughly shows the steps that we currently need for creating all the MAN pages, User manuals and reference documents. You may think: "Geeez, that looks so complicated. Why can't they simply convert XML files to PDF with Docbook, or use reST?" Please be patient, and continue reading. Things will get a little clearer soon.
Our toolchain doesn't only produce HTML and PDF files that are nice to look at, it also performs a lot of processing under the covers. We try to have our documentation as consistent as possible to the current behaviour of the source code, but this requires some extra steps.
So let's start right at the top...
The toolchain is set up, such that the User has a very restricted view on this whole "document processing thingy". All he should be concerned about is to edit existing text or write new sections and paragraphs. Sometimes even a completely new chapter has to be added, in general he can fire up his XML editor of choice and type away.
If he is a really nice user, he cares about validating his XML files against our special "SCons Docbook DTD/XSD". Either during typing, supported by his XML editor, or by executing a special script
from the top source folder afterwards. Preferably both.
Everything's looking okay, all validation passed? Good, then he simply commits his new work, and creates a pull request on Bitbucket. That's it!
Additionally, he can create the single documents on his side if he wants to get a feel for how the final result looks (and who doesn't?). Each of the document folders (design, developer, man, python10, reference, and user) contains an SConstruct file along with the actual XML files. You can call
from within the directory, and have the MAN pages or HTML created...even PDF, if you have a renderer installed (fop, xep or jw).
Just a few more words about the validation step. We are using our own DTD/XSD as a kind of hook, which only exists to link our own SCons documentation tags into the normal Docbook XSD. For the output, we always have an intermediary step (see diagram above), where we rewrite tags like cvar into a block of Docbook formatting elements representing it.
The toolchain, and all the Python scripts supporting it, are based on the prerequisite that all documents are valid against the SCons Docbook XSD. This step guarantees that we can accept the pull request of a user/writer with all his changes, and can create the documentation for a new release of SCons without any problems at a later time.
We are using entities for special keywords like SCons that should appear with the same formatting throughout the text. These are kept in a single file doc/scons.mod which gets included by the documents.
Additionally, for each Tool, Builder, Cvar and Function, a bunch of linkends in the form of entities get defined. They can be used in the MAN page and the User manual.
When you add an XML file in the src/engine/Tools folder, e.g. for a tool named foobar, you can use the two entities
- which prints the name of the Tool, and
- which is a link to the description of the Tool in the Appendix
of the User guide's text.
By calling the script
you can recreate the lists of entities (*.mod) in the generated folder, if required. At the same time, this will generate the *.gen files, which list the full description of all the Builders, Tools, Functions and CVars for the MAN page and the guide's appendix.
For more information about how to properly describe these elements, refer to the start of the Python script bin/SConsDoc.py. It explains the available tags and the exact syntax in detail.
In the User Guide, we support automatically created examples. This means that the output of the specified source files and SConstructs, is generated by running them with the current SCons version. We do this to ensure that the output displayed in the manual, is identical to what you get when you run the example on the command-line.
A short description about how these examples have to be defined, can be found at the start of the file bin/SConsExamples.py. Call
from the top level source folder, to run all examples through SCons.
Before this script starts to generate any output, it checks whether the names of all defined examples are unique. Another important prerequisite is, that for every example all the single scons_output blocks need to have a suffix attribute defined. These suffixes also have to be unique, within each example.
All example output files (*.xml) get written to the folder doc/generated/examples, together with all files defined via the scons_example_file tag. They are put under version control, too. Like this, it is easier to compare whether the output got broken for a new version of SCons.
Note, that these output files are not actually needed for editing the documents. When loading the User manual into an XML editor, you will always see the example's definition. Only when you create some output, the files from doc/generated/examples get XIncluded and all special scons* tags are transformed into Docbook elements.
Documents are in the folders design, developer, man, python10, reference, and user.
- Prepared configuration sets for the validating WYSIWYG XML editors XmlMind and Serna. You'll probably want to try the latter, because the XXE config requires you to have a full version (costing a few hundred bucks) and is therefore untested. For installing the Serna config, simply copy the scons folder into the plugins directory of your installation. Likewise, the XXE files from the xmlmind folder have to be copied into ~/.xxe4/ under Linux.
- Entity lists and outputs of the UserGuide examples. They get generated by the update scripts bin/docs-update-generated.py and bin/docs-create-example-outputs.py.
- Images for the overview.rst document.
- The SCons Docbook schema (XSD), based on the Docbook v4.5 DTD/XSD.
- XSLT transformation scripts for converting the special SCons tags like scons_output to valid Docbook during document processing.