It's well-known that a consistent user interface is easier to use. Consistency helps users focus on the task rather than the user interface. Likewise, a consistent documentation style helps users focus on the information, rather than the formatting.
A related goal is to design the documentation so that it is easy to maintain, so that it tends to remain internally consistent with the framework itself.
Do it now. Do it once. Do it well.
Overall, there are three goals for the documentation system.
- Say it all
- Say it once
- Say it well
First, we want the documentation to be both complete and concise. This is job one! The documentation should also be a quick but practical introduction to the framework, so newcomers can get started as easily as possible. To keep people coming back, the documentation should also be a repository of the tips and tricks we use in our own applications, so that people can find it here instead of asking over and over again on the list.
Second, the documentation should be easy to maintain. Ideally, we should cover the detail of each topic once, and draw as much detail from the source code and examples as possible (using the snippet macro).
Third, the documentation should be text-book quality; if not in the first draft, then in the next. Don't hesitate to hack in a new page. Better that we have the page than we don't. (See Job One!) But, as time allows, we should try to make each page the best that it can be. A great many people access the documentation, and it's worth the effort to make the "documentation experience" productive and enjoyable.
Capitalization of common terms
- Java
- Javadoc
- HTML
- XML
General Punctuation and Grammar
Good online resources for punctuation, grammar, and text style include
In print, two excellent (and inexpensive!) resources are
Also excellent, but more expensive:
Quick Tips
- Use as few words as possible. Instead of "but there are some quirks about it" try "but there are quirks".
- If a list of items includes both a term and an explanation, consider using a table instead of bullets.
- Avoid using "This" by itself. Instead of "This lets us" try "This strategy lets us".
- Ask yourself: "This what?"
- References to other wiki pages can be unqualified. For example: "See Documentation Style Guide."
Don't be smurfy!
A lot of API members use the term "action". We have
- action extensions on pages,
- action attributes in forms,
- action elements in configuration files, and
- Action Java classes, some of which may implement the
- Action interface.
Here are some terms that can be used to help clarify which action is which.
- Use "the framework" or "Struts 2" to refer to the codebase as a whole, including any frameworks we use internally, like XWork and OGNL.
- Use "Action class" or "action handler" to refer to the Java class incorporated by the action element.
- Use "action mapping" to refer to the object created by the action element.
Page Save Comment
Try to include a brief description of a change when saving a page. The comments are included in the page's history. The comments are also included on the daily change report. In a group environment, it's important to help each other follow along.
Parent Pages
Use the Parent Page feature to create a hierarchy of pages. The parent pages are reflected in the "bread crumb" menu. If properly used, parent pages can help browsers "visualize" the documentation as an outline.
The root of the documentation is the "Home" page, which is also the "Welcome" page. The documentation is ordered into three main areas: Tutorials, FAQs, and Guides. Each area has a contents page, whose parent is Home. Other pages within each section can also serve as parents, to help organize the documentation into a coherent outline.
Labels
Pages can be cross-indexed with the Label feature. Labels are not be used much yet, except for internal authoring.
FIXME |
A page that mentions a problem in the distribution that we intend to fix. Review these pages before tagging a distribution to see if the issue has been resolved. |
---|---|
TODO |
A page that is incomplete. Try to complete these pages before tagging a distribution |
Shortcuts Links
The Shortcut Link feature should be used for any external reference that may be used elsewhere.
Shortcuts being used include
Shortcut |
Purpose |
Usage |
Result |
---|---|---|---|
primer |
A bookmark in our Key Technologies Primer |
[javabeans@primer] |
|
jira |
A ticket in our issue tracker |
[WW-2111@jira] |
|
s2plugins |
S2 Plugin Repository |
[tiles-plugin.html@s2plugins] |
|
s2site |
The Struts 2 website |
[docs/home.html@s2site] |
About Headings
This section refers to: Notation Guide >> Headings.
About h1
Don't use h1.
at the top of each page. The page title serves as the "top level header". This is not as obvious online, but it is very apparent when the documentation is exported to HTML or PDF.
Try to start each page with some introductory text, to separate the page title from the rest of content.
Likewise, try to have some content between all page headings. Avoid placing headings one after the other.
Document sections
Headings can help you divide your document in sections, subsections, sub-subsections and so forth.
Advantages
Your document becomes more organized.
Disadvantages
Too many headings can fragment the text.
Here we go again!
This segment is an example of overusing headings. This whole "Headings" section has so few paragraphs that it really should have been written in just one section. The "advantages" and "disadvantages" would be just as easy to render as a table.
Headings capitalization
Try to use initial capitals for h1
and h2
headers.
For h3
and smaller headings, try to capitalize only the first word, and any proper nouns.
By using different capitalization styles, we emphasize the importance of bigger headings.
Avoid skipping headers
The headers form an outline for the page. When writing term papers, it is not a good practice to skip outline levels. When writing hypertext, it is not a good practice to skip heading levels either. Try not to skip from a h2
to a h4
.
Too many headings?
If you find yourself writing too many h2
headings in a single page, consider breaking the page into child pages.
More on Text Effects
This section refers to: Notation Guide >> Text Effects.
Text effects like strong, emphasis, and inserted can be used in the usual way to denote important parts of a sentence.
Monospaced
should be used to files, tags, and methods, like struts.xml
, <xmltag />
, and execute
. Class and Interface names may be left in normal face, like Action and Interceptor.
A panel should be preferred to a block quote.
The color fonts should be avoided or used only with great care. Some people have difficulty seeing some colors, and the colors may not be apparent if the page is printed.
Text Breaks
This section refers to: Notation Guide >> Text Breaks.
Text breaks should not be used to format blocks on the screen. If there is an issue with the way paragraphs or headings are being rendered, we should customize the stylesheet.
Lists
This section refers to: Notation Guide >> Lists.
Unordered lists should be created only with the *
(star) notation.
Ordered list should be used when numbering the items is important. Otherwise, prefer unordered lists.
- This is an unordered list in star notation;
- Items can have sub-items
- That can have sub-items
- That can have sub-items ...
- What is the limit?
- That can have sub-items ...
- That can have sub-items
- Mixing ordered and unordered lists is possible:
- One;
- Two;
- Three.
Images
This section refers to: Notation Guide >> Images and Notation Guide >> Misc.
Avoid using external images for bullets or icons. Prefer the equivalents provided with Confluence.
Images can be included by URL or annexing the binary to the page. Prefer annexing when possible, since URLs are subject to change.
Always observe copyright issues. Do not annex images unless it an original or public domain work, or the author has donated the image to the foundation.
Example:
Icons
Use , , , and to bullet important one-liners. Use to highlight cross references.
Used carefully, icons can make the content easier to read and understand.
However, if icons are overused, they lose impact (and can make a page look like a ransom note).
Casual icons like and should be used with care or avoided.
Tables
This section refers to: Notation Guide >> Tables.
Prefer lists for single-value entries. Prefer tables for lists with multiple columns.
Tables are very useful when lists just don't do it. Meaning: don't write a table when a list suffices. Tables are more organized, because you can align the text in columns. Since the markup text for tables in Confluence is not easy to read, complex and big tables can be hard to maintain.
File |
Optional |
Location (relative to webapp) |
Purpose |
---|---|---|---|
no |
/WEB-INF/ |
Web deployment descriptor to include all necessary WebWork components |
|
no |
/WEB-INF/classes/ |
Main configuration, contains result/view types, action mappings, interceptors, and so forth |
Advanced Formatting
This section refers to: Notation Guide >> Advanced Formatting.
Panels should be used as needed. Try to select the right panel for the content.
Try to give all panels and {code} blocks meaningful titles. People scan the pages looking for likely tips and examples.
Avoid generic titles like "Warning" or "Example". Style the headings like they were h3. or smaller.
When a panel contains a file or a class, the panel title should refer to the filename or classname.
Try to specify the language for {code} blocks.
/** Hello World class. */ public class HelloWorld { /** Main method. */ public static void main(String[] args) { System.out.println("Hello, World!"); } }
Try to use snippets for code blocks whenever possible!
Avoid tabs in code blocks, use two spaces instead. Long lines should be formatted to fit in a 800x600 resolution screen, without resorting to horizontal scrolling.
A typical example of noformat
would be the command line statements to compile and run the code above.
Either the code or noformat block can be used to represent command line windows. The terminal notation ({$}} should be used to represent a system prompt.
$ javac HelloWorld.java $ java HelloWorld Hello, World!
Change Happens
Anyone who has worked with databases knows the value of normalizing the schema. Ideally, we want to store each fact exactly once, and then use query system to retrieve that fact whereever it is needed. If we store a fact once, we only need to update it once, and we avoid inconsistencies in our data set.
To the extent possible, we want to "normalize" our technical documentation. Like a database, all technical documentation is subject to change. When change happens, we want the documentation to be as easy to update as possible. One way to do that is to try and minimize redundancy (without sacrificing ease of use).
Single sourcing with snippets
The "holy grail" of technical documentation is single sourcing. One way we try to single-source documentation is to pull content directly from the Javadocs and source code into the documentation.
Using a snippet macro, we are able to tag portions of any file for reuse. The macro fetches those snippets from a repository and merges the content into the documentation.
Use the Source!
Before writing any new content, ask yourself if we could place the content in the repository in either one of the example applications or the Javadocs. Rather than contrive an example, can you pull a snippet from one of the applications? Rather than reiterate Javadoc, could we update the Javadoc and make it a snippet? It is preferable to use snippets from the Struts example apps over Javadoc snippets for anything except plain text content as this ensures that the content's syntax has been validated.
Example snippet usage
{snippet:id=example|lang=xml|javadoc=true|url=struts2/core/src/main/java/org/apache/struts2/components/If.java}
|
Snippet Attributes |
---|---|
id |
The name of the snippet (optional - defaults to "all", meaning the entire file). |
url |
The URL where the snippet can be found (required). |
linenumbers |
If true line numbers are displayed. Numbering always starts at 1. |
lang |
lang=java would surround the snippet with {code:java}snippet{code}. If this snippet is simply text, don't include this parameter and the content will be printed outside of a code block. |
javadoc |
If true, the content is within a Javadoc block. If this is set to true, then the preceeding "* " (asterisk-space) characters will be stripped before merging the content. Also, the content is assumed to be already HTML escaped and won't be escaped again. |
All snippets are marked off by the pattern START SNIPPET: XXX
and END SNIPPET: XXX
where XXX
is the name
of the snippet that is assigned in the id
attribute of the macro. The URL is typically a location that points to the project's source control contents. |
About URLs
A URL must start with a valid prefix. There are two types of prefixes:
- com.opensymphony.xwork2. Notice the period. This syntax is better when you want to include content from a class because they allow you to use the fully qualified classname as the URL.
- struts2/ Notice the trailing slash. This syntax better when you want to include content content from non-class files such as xml or properties files. They may also be needed if a class based prefix for a sub-project has not be setup.
To include a snippet from http://svn.apache.org/repos/asf/struts/struts2/trunk/apps/showcase/src/main/java/org/apache/struts2/showcase/DateAction.java the two possible methods are:
{snippet:lang=java|url=struts2/apps/showcase/src/main/java/org/apache/struts2/showcase/DateAction.java} {snippet:lang=java|url=struts2/apps.showcase.src.main.java.org.apache.struts2.showcase.DateAction}
To include a snippet from https://svn.apache.org/repos/asf/struts/struts2/trunk/xwork-core/src/main/java/com/opensymphony/xwork2/validator/validators/StringLengthFieldValidator.java the two possible methods are:
{snippet:id=javadoc|javadoc=true|url=com.opensymphony.xwork2.validator.validators.StringLengthFieldValidator} {snippet:id=javadoc|javadoc=true|url=com.opensymphony.xwork2.validator/validators/StringLengthFieldValidator.java}
The list of available prefixes:
About snippet markers
When possible, all snippet markers should be in comment blocks. How they are commented depends on where the snippet is being embedded.
<!-- START SNIPPET: xxx --> ... <!-- END SNIPPET: xxx -->
if (true != false) { // START SNIPPET: xxx System.out.println("This is some silly code!"); // END SNIPPET: xxx }
If the snippet is embedded within Javadoc comments use HTML comments to declare the snippet as they won't render in the Javadocs.
When using the <pre>
tag within Javadoc comments embed the snippet markers inside the <pre>
tag.
* <pre> * <!-- START SNIPPET: example --> * <!-- records only the action's execution time --> * <action name="someAction" class="com.examples.SomeAction"> * <interceptor-ref name="completeStack"/> * <interceptor-ref name="timer"/> * <result name="success">good_result.ftl</result> * </action> * <!-- END SNIPPET: example --> * </pre>
A <pre>
tag within a Javadoc comment would be escaped and rendered as part of the snippet. See TimerInterceptor.java for an complete example.