Overview of rendering engine.
The module contains the RakuDoc::Processor class.
The class method 'render' is the principal one for rendering RakuDoc v2. The Raku compiler expects to call render with an AST and to return a Str. It can be called with the option :pre-finalised to get the processed data, see the example in Debugging.
The aim is for a RakuDoc::Processor object (an RPO) to be as generic as possible, allowing for other classes to instantiate it, and to attach templates that will generate more specific formats, such as HTML, MarkDown, or EPub pages.
An RPO relies on Templates, PromiseStrings, and ScopedData.
It is also necessary to have a good understanding of RakuDoc v2.
Some environment variables provide for some output control.
The following describes choices that are left by RakuDoc v2 to the renderer.
An RPO contains a default set of Text templates.
The aim of the text templates is that they are the lowest level templates needed to render any RakuDoc source.
When a renderer is needed to output a new format, eg., HTML, the renderer instantiates an RPO and attaches new templates to it using the add-template( Pair $p, :source ) method or the add-templates( Hash $h, :source ) method. Setting the source option to some value will help determine which module introduced the templates.
The design of the Templates object means that new additions to the object push the previous definition onto a linked list, and so they will always be available by default.
The set of template keys needed to create a renderer is tabulated here.
All aspects of the output can be defined using the templates.
The generic render method transforms the AST of the input file using a handle method for each type of Block and Markup code. A template exists for each block and markup code.
After all the blocks have been processed, the Table of Contents, Index, and Footnotes are rendered, with templates for each item and for the list of items.
Any forward references and numerations are then resolved.
The intermediate structure of the processed file can then be accessed before it is finally rendered into a string.
The final rendering is done by the 'final' template.
Once a string has been obtained, it can be post-processed using the post-process method. For example, text output can be transformed by wrapping lines.
In order to offer the maximum flexibility to a RakuDoc author, if there is an error condition after the RakuDoc source has been compiled, a set of warnings are generated.
By default these are rendered using the 'warnings' template, and the rendered version is appended to the output file.
RakuDoc allows for customisability. This renderer handles customisability via the template mechanism.
When custom block is detected according to the RakuDoc v2 spelling rules, an RPO will check whether a template exists with that name.
If a template has been attached to an RPO, then the rendered content of the block (including rendered embedded RakuDoc instructions) will be provided to the template as contents and the verbatim contents with no renderering will be provided as raw.
So a new block can be created by attaching a template with the blockname to the RPO.
RakuDoc v2 allows for custom markup codes in two ways:
the M< DISPLAY | FUNCTION ; list of comma delimited strings >
a single Unicode (not ASCII ) character with the upper property.
When a M< DISPLAY | FUNCTION ; LIST > is encountered, the renderer will look for a template named in the FUNCTION position. If no such template exists, the text will be rendered verbatim and a warning logged.
When some custom markup is encountered, the renderer will
check that the character, eg Ɵ (Greek Capital Theta), has the unicode property Upper;
look for a template called (eg) markup-Ɵ. For clarity, the Unicode character chosen for the markup code is prefixed by markup- in order to name the template.
Thus in order to create a new markup code, create an appropriately named template and add it to the RPO.
More information is available in Templates.
All templates can attach items to the Table of Contents, Index, Footnotes, and Warnings structures of the rendered source using the helper methods.
The following helper methods are available:
escape.($s), where $s is a string. Note the . after escape
In addition, within a template, it is possible to attach data to globals, and to retrieve the data in another template.
For an example of this, see test xt/030-customisation-data.rakutest. (test number may be changed, but otherwise the filename will be the same).
When developing new templates or renderers based on an RPO, several debug options can be attached to the RPO, eg.
my RakuDoc::Processor $rdp .=new( :test );
$rdp.debug( AstBlock, Templates);
$ast = Q:to/QAST/.AST;
=begin rakudoc
=head This is a header
Some text
=end rakudoc
QAST
$rv = $rdp.render( $ast, :pre-finalised );
'myOutput'.IO.spurt: $rdp.finalise
Commentary on the code
new( :test ) this attaches the test templates to the RPO
:pre-finalised returns the RakuDoc::Processed object
if called with :pre-finalised, calling finalise returns a Str that can then be stored as a file.
The following options are available:
BlockType information about which RakuAST::Doc::Block is being processed
It is also possible to get the result of one template (so as to reduce the amount of output information). This is done eg for the 'table' template: $rdp.verbose( 'table' ); $rv = $rdp.render( $ast, :pre-finalised );
The Test and Pretty options described in the Templates documentation can be set on an RPO, eg. $rdp.test( True ); $rdp.pretty( True );
Bear in mind that the pretty flag overrides the test flag, and both override the debug and verbose flags.
By setting the environment variable POSTPROCESSING=1 the text output will be naively wrapped.
For example,
POSTPROCESSING=1 bin/RenderTextify --pretty rakudociem-ipsum
If the environment variable WIDTH is also set, the text output will be wrapped to the value.
WIDTH by default is set at 80 chars. To set at 70, use:
POSTPROCESSING=1 WIDTH=70 bin/RenderTextify rakudociem-ipsum
Since sometimes it is necessary to target a specific paragraph, some paragraphs get an automated id based on the last Seven hex chars of the SHA1 encoding of its contents.
Seven chars should be adequate so long as the number of paragraphs in a single document is below 16,000.
Based on the expectation that we would see collision in a repository with 2^(2N) objects when using object names shortened to first N bits Stackoverflow question
Long texts or books will probably need more to avoid a conflict. This can be done by setting paragraph-id-length in the %structure-data to the required number of hex digits.
Typically, it should only be necessary to change the templates of an existing renderer to alter the way output is formated. But for a new type of format, a new renderer will be needed.
It is best to look at the code for the HTML renderer to see how to write a renderer. The following are only very brief notes.
A new renderer for a new output format will need:
A processor class, which is sub-classed from RakuDoc::Processor. The escape method will need to be over-ridden.
A Renderer class which must be called RakuDoc::To::NEWFORMAT. This is required so that the code
raku --rakudoc=NEWFORMAT input-file.rakudoc > output-filename.newformat
will work.
The Renderer class must contain the method render( $ast ), which is needed by the raku compiler, as noted above.