6.3 Structures And Processing

A document is represented as a part, as described in Parts, Flows, Blocks, and Paragraphs. This representation is intended to be independent of its eventual rendering, and it is intended to be immutable; rendering extensions and specific data in a document can collude arbitrarily, however.

A document is processed in four passes:

• The traverse pass traverses the document content in document order so that information from one part of a document can be communicated to other parts of the same document. The information is transmitted through a symbol-keyed mapping that can be inspected and extended by traverse-elements and traverse-blocks in the document. The traverse pass iterates the traversal until it obtains a fixed point (i.e., the mapping from one iteration is unchanged from the previous iteration).

• The collect pass globally collects information in the document that can span documents that are built at separate times, such as targets for hyperlinking.

• The resolve pass matches hyperlink references with targets and expands delayed elements (where the expansion should not contribute new hyperlink targets).

• The render pass generates the result document.

None of the passes mutate the document representation. Instead, the traverse pass, collect pass, and resolve pass accumulate information in a side hash table, collect-info table, and resolve-info table. The collect pass and resolve pass are effectively specialized version of traverse pass that work across separately built documents.

6.3.1 Parts, Flows, Blocks, and Paragraphs

This diagram shows the large-scale structure of the type hierarchy for Scribble documents. A box represents a struct or a built-in Racket type; for example part is a struct. The bottom portion of a box shows the fields; for example part has three fields, title, blocks, and subparts. The substruct relationship is shown vertically with navy blue lines connected by a triangle; for example, a compound-paragraph is a block. The types of values on fields are shown via dark red lines in the diagram. Doubled lines represent lists and tripled lines represent lists of lists; for example, the blocks field of compound-paragraph is a list of blocks. Dotted lists represent functions that compute elements of a given field; for example, the block field of a traverse-block struct is a function that computes a block.

The diagram is not completely accurate: a table may have 'cont in place of a block in its cells field, and the types of fields are only shown if they are other structs in the diagram. A prose description with more detail follows the diagram.

A part is an instance of part; among other things, it has a title content, an initial flow, and a list of subsection parts. There is no difference between a part and a full document; a particular source module just as easily defines a subsection (incorporated via include-section) as a document.

A flow is a list of blocks.

A block is either a table, an itemization, a nested flow, a paragraph, a compound paragraph, a traverse block, or a delayed block.

• A table is an instance of table; it has a list of list of blocks corresponding to table cells.

• A itemization is an instance of itemization; it has a list of flows.

• A nested flow is an instance of nested-flow; it has a flow that is typeset as sub-flow.

• A paragraph is an instance of paragraph; it has a content:

  • An content can be a string, one of a few symbols, an instance of element (possibly link-element, etc.), a multiarg-element, a a traverse element, part-relative element, a delayed element, or a list of content.

    • A string is included in the result document verbatim, except for space, and unless the content’s enclosing style is 'hspace. In a style other than 'hspace, consecutive spaces in the output may be collapsed togther or replaced with a line break. In the style 'hspace, all text is converted to uncollapsable spaces that cannot be broken across lines.

    • A symbol content is either 'mdash, 'ndash, 'ldquo, 'lsquo, 'rdquo, 'rsquo, 'larr, 'rarr, or 'prime; it is rendered as the corresponding HTML entity (even for Latex output).

    • An instance of element has a content plus a style. The style’s interpretation depends on the renderer, but it can be one of a few special symbols (such as 'bold) that are recognized by all renderers.

    • An instance of link-element has a tag for the target of the link.

    • An instance of target-element has a tag to be referenced by link-elements. An instance of the subtype toc-target-element is treated like a kind of section label, to be shown in the “on this page” table for HTML output.

    • An instance of index-element has a tag (as a target), a list of strings for the keywords (for sorting and search), and a list of contents to appear in the end-of-document index.

    • An instance of image-element incorporates an image from a file into the rendered document.

    • An instance of multiarg-element combines a style with a list of content, where the style corresponds to a rendered command that takes multiple arguments.

    • An instance of collect-element has a procedure that is called in the collect pass of document processing to record information used by later passes.

    • A traverse element is an instance of traverse-element, which ultimately produces content, but can accumulate and inspect information in the traverse pass.

    • A part-relative element is an instance of part-relative-element, which has a procedure that is called in the collect pass of document processing to obtain content. When the part-relative element’s procedure is called, collected information is not yet available, but information about the enclosing parts is available.

    • A delayed element is an instance of delayed-element, which has a procedure that is called in the resolve pass of document processing to obtain content.

    • An instance of render-element has a procedure that is called in the render pass of document processing.

• A compound paragraph is an instance of compound-paragraph; like blockquote, it has list of blocks, but the blocks are typeset as a single paragraph (e.g., no indentation after the first block) instead of inset.

• A traverse block is an instance of traverse-block, which ultimately produces another block, but can accumulate and inspect information during the traverse pass.

• A delayed block is an instance of delayed-block, which has a procedure that is called in the resolve pass of document processing to obtain a block.

6.3.2 Tags

A tag is a list containing a symbol and either a string, a generated-tag instance, or an arbitrary list. The symbol effectively identifies the type of the tag, such as 'part for a tag that links to a part, or 'def for a Racket function definition. The symbol also effectively determines the interpretation of the second half of the tag.

A part can have a tag prefix, which is effectively added onto the second item within each tag whose first item is 'part, 'tech, or 'cite, or whose second item is a list that starts with 'prefixable:

• The prefix is added to a string second item by creating a list containing the prefix and string.

• The prefix is added to a list second item after 'part, 'tech, or 'cite using cons.

• The prefix is added to a second item that starts 'prefixable by adding it to the list after 'prefixable.

• A prefix is not added to a generated-tag item.

The prefix is used for reference outside the part, including the use of tags in the part’s tags field. Typically, a document’s main part has a tag prefix that applies to the whole document; references to sections and defined terms within the document from other documents must include the prefix, while references within the same document omit the prefix. Part prefixes can be used within a document as well, to help disambiguate references within the document.

Some procedures accept a “tag” that is just the string part of the full tag, where the symbol part is supplied automatically. For example, section and secref both accept a string “tag”, where 'part is implicit.

The scribble/tag library provides functions for constructing tags.

6.3.3 Styles

A style combines a style name with a list of style properties in a style structure. A style name is either a string, symbol, or #f. A style property can be anything, including a symbol or a structure such as color-property.

A style has a single style name, because the name typically corresponds to a configurable instruction to a renderer. For example, with Latex output, a string style name corresponds to a Latex command or environment. For more information on how string style names interact with configuration of a renderer, see Extending and Configuring Scribble Output. Symbolic style names, meanwhile, provide a simple layer of abstraction between the renderer and documents for widely supported style; for example, the 'italic style name is supported by all renderers.

Style properties within a style compose with style names and other properties. Again, symbols are often used for properties that are directly supported by renderers. For example, 'unnumbered style property for a part renders the part without a section number. Many properties are renderer-specific, such as a hover-property structure that associates text with an element to be shown in an HTML display when the mouse hovers over the text.

6.3.4 Collected and Resolved Information

The collect pass, resolve pass, and render pass processing steps all produce information that is specific to a rendering mode. Concretely, the operations are all represented as methods on a render<%> object.

The result of the collect method is a collect-info instance. This result is provided back as an argument to the resolve method, which produces a resolve-info value that encapsulates the results from both iterations. The resolve-info value is provided back to the resolve method for final rendering.

Optionally, before the resolve method is called, serialized information from other documents can be folded into the collect-info instance via the deserialize-info method. Other methods provide serialized information out of the collected and resolved records.

During the collect pass, the procedure associated with a collect-element instance can register information with collect-put!.

During the resolve pass, collected information for a part can be extracted with part-collected-info, which includes a part’s number and its parent part (or #f). More generally, the resolve-get method looks up information previously collected. This resolve-time information is normally obtained by the procedure associated with a delayed block or delayed element.

The resolve-get information accepts both a part and a resolve-info argument. The part argument enables searching for information in each enclosing part before sibling parts.

6.3.5 Structure Reference

The tags indicates a list of tags that each link to the section. Normally, tags should be a non-empty list, so that hyperlinks can target the section.

The title-content field holds the part’s title, if any.

For the style field, the currently recognized symbolic style names are as follows:

• 'index — The part represents an index.

The recognized style properties are as follows:

• 'unnumbered — A section number is not computed or rendered for the section.

• 'hidden-number — A section number is computed for the section, but it is not rendered as part of the section name.

• 'toc-hidden — The part title is not shown in tables of contents, including in “on this page” boxes. For Latex rendering, the part title is omitted only if it is unnumbered or has a hidden number.

• 'hidden — The part title is not shown; for Latex output, the part title is not shown only if its is empty, and in that case, it is also excluded from tables of contents. The 'toc-hidden style property usually should be included with 'hidden (for consistency in non-Latex output).

• 'grouper — The part is numbered with a Roman numeral, and its subsections continue numbering as if they appeared in the preceeding part. In other words, the part acts like a “part” in a book where chapter numbering is continuous across parts.

• 'toc — Sub-parts of the part are rendered on separate pages for multi-page HTML mode.

• 'non-toc — Initial sub-parts of the part are not rendered on separate pages for multi-page HTML mode; this style property applies only to the main part.

• 'reveal — Shows sub-parts when this part is displayed in a table-of-contents panel in HTML output (which normally shows only the top-level sections).

• 'quiet — In HTML output and most other output modes, hides entries for sub-parts of this part in a table-of-contents or local-table-of-contents listing except when those sub-parts are top-level entries in the listing.

• 'no-toc — As a style property for the main part of a rendered page, causes the HTML output to not include a margin box for the main table of contents; the “on this page” box that contains toc-element and toc-target-element links (and that only includes an “on this page” label for multi-page documents) takes on the location and color of the main table of contents, instead.

• 'no-sidebar — As a style property for the main part of a document, causes the HTML output to not include an “on this page” margin box.

• document-version structure — A version number for this part and its sub-parts (except as overridden). When it is not "" may be used when rendering a document; at a minimum, a non-"" version is rendered when it is attached to a part representing the whole document. The default version for a document is (version). In rendered form, the version is normally prefixed with the word “Version,” but this formatting can be controlled by overriding .version:before and/or .versionNoNav:before in CSS for HTML rendering or by redefining the \SVersionBefore macro for Latex rendering (see Extending and Configuring Scribble Output).

• document-date structure — A date for the part, normally used on a document’s main part for for Latex output. The default date for a document is #f, which avoids explicitly specifying a date at the Latex level, so that the current date is used as the document date. Set the date to "" to suppress a date in an output document.

• body-id structure — Generated HTML uses the given string id attribute of the <body> tag; this style property can be set separately for parts that start different HTML pages, otherwise it is effectively inherited by sub-parts; the default is "scribble-racket-lang.org", but setup-plt installs "doc-racket-lang.org" as the id for any document that it builds.

• attributes structure — Provides additional HTML attributes for the <html> tag when the part corresponds to its own HTML page.

• head-extra structure — Provides additional HTML content for the <head> tag when the part corresponds to its own HTML page.

• color-property structure — For HTML, applies a color to the part title.

• background-color-property structure — For HTML, applies a color to the background of the part title.

• hover-property structure — For HTML, adds a text label to the title to be shown when the mouse hovers over it.

• render-convertible-as structure — For HTML, controls how objects that subscribe to the file/convertible protocol are rendered.

The to-collect field contains content that is inspected during the collect pass, but ignored in later passes (i.e., it doesn’t directly contribute to the output).

The blocks field contains the part’s initial flow (before sub-parts).

The parts field contains sub-parts.

For the style field, a string style name corresponds to a CSS class for HTML output or a macro for Latex output (see Implementing Styles). The following symbolic style names are recognized:

• 'author — Typeset as the author of a document. Such paragraphs normally should appear only in the initial flow of a part for a document, where they are treated specially by the Latex renderer by moving the author information to the title.

• 'pretitle — Typeset before the title of the enclosing part.

• 'wraps — Like a #f style name, but not boxable in the sense of box-mode for Latex output.

When a paragraph’s style is #f, then it is boxable in the sense of box-mode for Latex output.

The currently recognized style properties are as follows:

• 'omitable — When a table cell contains a single paragraph with the 'omitable style property, then when rendering to HTML, no <p> tag wraps the cell content.

• 'div — Generates <div> HTML output instead of <p> (unless a alt-tag property is provided).

• alt-tag structure — Generates the indicated HTML tag instead of <p> or <div>.

• attributes structure — Provides additional HTML attributes for the <p>, <div>, or alternate tag.

• body-id structure — For HTML, uses the given string as an id attribute of the <p>, <div>, or alternate tag.

• 'never-indents — For Latex and compound paragraphs; see compound-paragraph.

• box-mode structure — For Latex output, uses an alternate rendering form for boxing contexts (such as a table cell); see box-mode.

A table has, roughly, a list of list of blocks. A cell in the table can span multiple columns by using 'cont instead of a block in the following columns (i.e., for all but the first in a set of cells that contain a single block).

Within style, a string style name corresponds to a CSS class for HTML output or an environment for Latex output (see Implementing Styles). The following symbolic style names are also recognized:

• 'boxed — Renders as a definition. This style name is not intended for use on a table that is nested within a 'boxed table; nested uses may look right for some renders of the style but not others.

• 'centered — Centers HTML output horizontally.

• 'block — Prevents pages breaks in Latex output.

The following style properties are currently recognized:

• table-columns structure — Provides column-specific styles, but only column-attributes properties (if any) are used if a table-cells structure is included as a style property. See table-cells for information about how a column style is used for each cell.

• table-cells structure — Provides cell-specific styles. See table-cells for information about how the styles are used.

• attributes structure — Provides additional HTML attributes for the <table> tag.

• body-id structure — For HTML, uses the given string as an id attribute of the <table> tag.

• 'aux — For HTML, include the table in the table-of-contents display for the enclosing part.

• 'never-indents — For Latex and compound paragraphs; see compound-paragraph.

For Latex output, a paragraph as a cell value is not automatically line-wrapped, unless a vertical alignment is specified for the cell through a table-cells or table-columns style property. To get a line-wrapped paragraph, use a compound-paragraph or use an element with a string style and define a corresponding Latex macro in terms of \parbox. For Latex output of blocks in the flow that are nested-flows, itemizations, compound-paragraphs, or delayed-blocks, the block is wrapped with minipage using \linewidth divided by the column count as the width.

In style, a string style name corresponds to a CSS class for HTML output or a macro for Latex output (see Implementing Styles). In addition, the following symbolic style names are recognized:

• 'compact — Reduces space between items.

• 'ordered — Generates <ol> HTML output instead of <ul> or an Latex enumeration instead of an itemization.

The following style properties are currently recognized:

• attributes structure — Provides additional HTML attributes for the <ul> or <ol> tag.

• body-id structure — For HTML, uses the given string as an id attribute of the <ul> or <ol> tag.

• 'never-indents — For Latex and compound paragraphs; see compound-paragraph.

In style, the style name is normally a string that corresponds to a CSS class for HTML <blockquote> output or a Latex environment (see Implementing Styles). The following symbolic style names are recognized:

• 'inset — Insets the nested flow relative to surrounding text.

• 'code-inset — Insets the nested flow relative to surrounding text in a way suitable for code. If the nested flow has a single block, then it is boxable in the sense of box-mode for Latex output.

• 'vertical-inset — Insets the nested flow vertically relative to surrounding text, but not horizontally. If the nested flow has a single block, then it is boxable in the sense of box-mode for Latex output.

The following style properties are currently recognized:

• 'command — For Latex output, a string style name is used as a command name instead of an environment name.

• 'multicommand — For Latex output, a string style name is used as a command name with a separate argument for each block in blocks.

• attributes structure — Provides additional HTML attributes for the <blockquote> tag.

• body-id structure — For HTML, uses the given string as an id attribute of the <blockquote> tag.

• 'never-indents — For Latex and compound paragraphs; see compound-paragraph.

• box-mode structure — For Latex output, uses an alternate rendering form for boxing contexts (such as a table cell); see box-mode.

• 'decorative — The content of the nested flow is intended for decoration. Text output skips a decorative nested flow.

• alt-tag structure — Generates the indicated HTML tag instead of <blockquote>.

For HTML, a paragraph block in blocks is rendered without a <p> tag, unless the paragraph has a style with a non-#f style name. For Latex, each block in blocks is rendered with a preceding \noindent, unless the block has the 'never-indents property (checking recursively in a nested-flow or compound-paragraph if the nested-flow or compound-paragraph itself has no 'never-indents property).

The style field of a compound paragraph is normally a string that corresponds to a CSS class for HTML output or Latex environment for Latex output (see Implementing Styles). The following style properties are currently recognized:

• 'command — For Latex output, a string style name is used as a command name instead of an environment name.

• alt-tag structure — Generates the given HTML tag instead of <p>.

• attributes structure — Provides additional HTML attributes for the <p> or alternate tag.

• body-id structure — For HTML, uses the given string as an id attribute of the <p> or alternate tag.

• 'never-indents — For Latex within another compound paragraph; see above.

The traverse procedure is called with get and set procedures to get and set symbol-keyed information; the traverse procedure should return either a block (which effectively takes the traverse-block’s place) or a procedure like traverse to be called in the next iteration of the traverse pass.

All traverse-element and traverse-blocks that have not been replaced are forced in document order relative to each other during an iteration of the traverse pass.

The get procedure passed to traverse takes a symbol and any value to act as a default; it returns information registered for the symbol or the given default if no value has been registered. The set procedure passed to traverse takes a symbol and a value to registered for the symbol.

See also cond-block in scriblib/render-cond. The symbol 'scribble:current-render-mode is automatically registered to a list of symbols that describe the target of document rendering. The list contains 'html when rendering to HTML, 'latex when rendering via Latex, and 'text when rendering to text. The registration of 'scribble:current-render-mode cannot be changed via set.

The style field can be a style structure, but it can also be just a style name.

In style, a string style name corresponds to a CSS class for HTML output and a macro name for Latex output (see Implementing Styles). The following symbolic style names are recognized:

• 'tt, 'italic, 'bold, 'roman, 'sf, 'url, 'subscript, 'superscript, 'smaller, 'larger — Basic styles recognized by all renders.

• 'hspace — Renders its content as monospace blanks.

• 'newline — Renders a line break independent of the content.

• 'no-break — Prevents line breaks when rendering content.

The following style properties are currently recognized:

• target-url structure — Generates a hyperlink.

• url-anchor structure — For HTML, inserts a hyperlink target before content.

• color-property structure — Applies a color to the text of content.

• background-color-property structure — Applies a color to the background of content.

• alt-tag structure — Generates the given HTML tag instead of the default one (<span>, b, etc.).

• attributes structure — Provides additional HTML attributes for a tag.

• hover-property structure — For HTML, adds a text label to the content to be shown when the mouse hovers over it.

• script-property structure — For HTML, supplies a script alternative to content.

• body-id structure — For HTML uses the given string as an id attribute of the <span> tag.

• 'aux — Intended for use in titles, where the auxiliary part of the title can be omitted in hyperlinks. See, for example, secref.

• 'tt-chars — For Latex output, when the style name is a string, render the element’s content with escapes suitable for Latex tt mode.

• 'exact-chars — For Latex output, when the style name is a string, render the elements content exactly (without escapes).

• command-extras structure — For Latex output, adds strings as arguments to the Latex command.

For each string in suffixes, if the rendered works with the corresponding suffix, the suffix is added to path and used if the resulting path refers to a file that exists. The order in suffixes determines the order in which suffixes are tried. The HTML renderer supports ".png", ".gif", and ".svg", while the Latex renderer supports ".png", ".pdf", and ".ps" (but rendering Latex output to PDF will not work with ".ps" files, while rendering to Latex DVI output works only with ".ps" files). If suffixes is empty or if none of the suffixes lead to files that exist, path is used as-is.

The scale field scales the image in its rendered form.

When tag is a part tag and the content of the element is null, then the hyperlink uses the target part’s number and/or title as the content. In that case, if the section number is preceded by a word, the word starts in uppercase if the element’s style includes a 'uppercase property.

The following symbol is recognized as a style property:

• 'indirect-link — For HTML output, treats the link as “external”. When rendering to HTML and the set-external-tag-path method is called to provide an external-link URL, then the resolution of the hyperlink can be deferred until the link is clicked (or, in some cases, patched by JavaScript when the documentation is viewed in a browser).

The entry-seq list must have the same length as plain-seq. It provides the form of each key to render in the final document.

The desc field provides additional information about the index entry as supplied by the entry creator. For example, a reference to a procedure binding can be recognized when desc is an instance of procedure-index-desc. See scribble/manual-struct for other typical types of desc values.

See also index.

The sizer field is a procedure that produces a substitute content for the delayed element for the purposes of determining the delayed element’s width (see element-width).

The plain field is a procedure that produces a substitute content when needed before the collect pass, such as when element->string is used before the collect pass.

The resolve function can call collect-info-parents to obtain a list of parts that enclose the element, starting with the nearest enclosing section. Functions like part-collected-info and collected-info-number can extract information like the part number.

Unlike delayed-element or part-relative-element, the element remains intact (i.e., it is not replaced) by either the collect pass or resolve pass.

If a render-element instance is serialized (such as when saving collected info), it is reduced to a element instance.

The length of the number list indicates the section’s nesting depth. Numbers in number correspond to the section’s number, it’s parent’s number, etc. A non-empty string is used for a 'grouper section. For an unnumbered section, #f is used in place of all numbers and "" in place of all non-empty strings.

When rendering to HTML, a color-property is also recognized for a block, part (and used for the title in the latter case)or cell in a table.

If a cell style has a string name, it is used as an HTML class for the <td> tag or as a Latex command name.

The following are recognized as cell-style properties:

• 'left — Left-align the cell content.

• 'right — Right-align the cell content top baselines.

• 'center — Center the cell content horizontally.

• 'top — Top-align the cell content.

• 'baseline — Align the cell content top baselines.

• 'bottom — bottom-align the cell content.

• 'vcenter — Center the cell content vertically.

• color-property structure — For HTML, applies a color to the cell content.

• background-color-property structure — For HTML, applies a color to the background of the cell.

• attributes — Provides additional HTML attributes for the cell’s <td> tag.

Changed in version 1.1 of package scribble-lib: Added color-property and background-color-property support.

For HTML table rendering, for each column that has a column-attributes property in the corresponding element of styles, the attributes are put into an HTML col tag within the table.

A boxing context starts with a table cell in a multi-column table, and the content of a block in a boxing context is also in a boxing context. If the cell’s content is boxable, then the content determines the width of the cell, otherwise a width is imposed. A paragraph with a #f style name is boxable as a single line; the 'wraps style name makes the paragraph non-boxable so that its width is imposed and its content can use multiple lines. A table is boxable when that all of its cell content is boxable.

To generate output in box mode, the box-mode property supplies Latex macro names to apply to the nested flow or paragraph content. The top-name macro is used if the box’s top line is to be aligned with other boxes, center-name if the box’s center is to be aligned, and bottom-name if the box’s bottom line is to be aligned. The box-mode* function creates a box-mode structure with the same name for all three fields.

A box-mode style property overrides any automatic boxed rendering (e.g., for a paragraph with style name #f). If a block has both a box-mode style property and a 'multicommand style property, then the Latex macro top-name, center-name, or bottom-name is applied with a separate argument for each of its content.

If p and info arguments are not supplied, then a pre-“collect” substitute is obtained for delayed elements. Otherwise, the two arguments are used to force the delayed element (if it has not been forced already).

For a list that is an info tag, the interpretation of the second element of the list is effectively determined by the leading symbol, which classifies the key. However, a #f value as the second element has an extra meaning: collected information mapped by such info keys is not propagated out of the part where it is collected; that is, the information is available within the part and its sub-parts, but not in ancestor or sibling parts.

Note that every tag is an info key.

The result is #f if the no value for the given key is found. Furthermore, the search failure is recorded for potential consistency reporting, such as when racket setup is used to build documentation.

Added in version 1.1 of package scribble-lib.

(recursive-contract

((symbol? any/c . -> . any/c)

(symbol? any/c . -> . any)

. -> . (or/c block-traverse-procedure/c

block?)))

(recursive-contract

((symbol? any/c . -> . any/c)

(symbol? any/c . -> . any)

. -> . (or/c element-traverse-procedure/c

content?)))

6.3.6 HTML Style Properties

The path field can be a result of path->main-collects-relative.

The alternatives in the types field are tried in order and the first one that succeeds is used in the html output.

The file indicated by path is referenced in place when render<%> is instantiated with refer-to-existing-files as true. Otherwise, it is copied to the destination directory and potentially renamed to avoid conflicts.

This style property is useful only when render<%> is instantiated with refer-to-existing-files as #f, and only when path does not match then name of any other file that is copied by the renderer to the destination.

6.3.7 Latex Style Properties

The path field can be a result of path->main-collects-relative.