It might appear that it is a fairly simple matter to produce documents in Adobe\*(rg\~"Portable\~Document\~Format", commonly known as PDF, using groff ) GNU\~Troff\~( as the document formatter. Indeed, groff 's default output format is the native Adobe\*(rg\~PostScript\*(rg format, which PDF producers such as Adobe\*(rg Acrobat\*(rg Distiller\*(rg, or GhostScript, expect as their input format. Thus, the PDF production process would seem to entail simply formatting the document source with groff , to produce a PostScript\*(rg version of the document, which can subsequently be processed by Acrobat\*(rg Distiller\*(rg or GhostScript, to generate the final PDF document.

For many PDF production requirements, the production cycle described above may be sufficient. However, this is a limited PDF production method, in which the resultant PDF document represents no more than an on screen image of the printed form of the document, if groff 's PostScript\*(rg output were printed directly.

The Portable Document Format provides a number of features, which significantly enhance the experience of reading a document on screen, but which are of little or no value to a document which is merely printed. It .EM is possible to exploit these PDF features, which are described in the Adobe\*(rg
This is an example of a resource reference specified by URI ...
We may need to refer often to the Adobe pdfmark Reference Manual,
so we create the internet link definition using a macro, to make
it reusable.

Note also, that we protect the description of the reference by
preceding it with "--", to avoid "invalid character in name" type
error messages from groff (caused by the use of "\~").

.pdfhref W -D http://partners.adobe.com/asn/acrobat/docs/pdfmark.pdf \ -P " -A "\\$1 -- pdfmark\~Reference\~Manual .. .pdfmark-manual , with some refinement of the simple PDF production method, provided appropriate "feature implementing" instructions can be embedded into groff 's PostScript\*(rg rendering of the document. This, of course, implies that the original document source, which groff will process to generate the PostScript\*(rg description of the document, must include appropriate markup to exploit the desired PDF features. It is this preparation of the groff document source to exploit a number of these features, which provides the principal focus of this document.

The markup techniques to be described have been utilised in the production of the PDF version of this document itself. This has been formatted using groff 's ms macro package; thus, usage examples may be found in the document source file, \n(.F , to which comments have been added, to help identify appropriate markup examples for implementing PDF features, such as:\(en .QS

All of the techniques described have been tested on .EM both GNU/Linux, and on Microsoft\*(rg Windows\(tm2000 operating platforms, using groff 1.19.1 ,\c .pdfhref L -D footnote1 -- \** .FS .pdfhref M footnote1 Later versions should, and some earlier versions may, be equally suitable. See .pdfhref W \*[GROFF-WEBSITE] for information and availability of the latest version. .FE in association with AFPL GhostScript 8.14 .\c .pdfhref L -D footnote2 -- \** .FS .pdfhref M footnote2 Again, other versions may be suitable. See .pdfhref W http://ghostscript.com for information and availability. .FE Other tools employed, which should be readily available on .EM any UNIX\(tm or GNU/Linux system, are sed , awk and make , together with an appropriate text editor, for creating and marking up the groff input files. These additional utilities are not provided, as standard, on the Microsoft\*(rg Windows\(tm platform, but several third party implementations are available. Some worth considering include the MKS\*(rg\~Toolkit,\** .FS A commercial offering; see .pdfhref W http://mkssoftware.com/products/tk/default.asp for information. .FE Cygwin,\** .FS A .EM free but comprehensive POSIX emulation environment and UNIX\(tm toolkit for 32\(hybit Microsoft\*(rg Windows\(tm platforms; see .pdfhref W http://cygwin.com for information and download. .FE or MSYS.\** .FS Another free, but minimal suite of common UNIX\(tm tools for 32\(hybit Microsoft\*(rg Windows\(tm, available for download from .pdfhref W -A ; http://www.mingw.org it .EM does include those tools listed above, and is the package which was actually used when performing the Windows\(tm2000 platform tests referred to in the text. .FE This list is by no means exhaustive, and should in no way be construed as an endorsement of any of these packages, nor to imply that other similar packages, which may be available, are in any way inferior to them. .bp We may wish a section heading to represent a named destination,
so that we can create a linked reference to it, from some other
part of the PDF document, (or even from another PDF document).

Here we use the "-N" option of the "XN" macro, to create a named
PDF link destination, at the location of the heading. Notice that
we also use the "--" marker to separate the heading text from the
preceding option specification; it is not strictly necessary in
this case, but it does help to set off the heading text from the
option specification.

.XN -N pdf-features -- Exploiting PDF Document Features

To establish a consistent framework for adding PDF features, a groff macro package, named pdfmark.tmac , has been provided. Thus, to incorporate PDF features in a document, the appropriate macro calls, as described below, may be placed in the groff document source, which should then be processed with a groff command of the form .QP .fam C groff -Tps [-m name "] -m" pdfmark options [- "file ..." "...] "

It may be noted that the pdfmark macros have no dependencies on, and no known conflicts with, any other groff macro package; thus, users are free to use any other macro package, of their choice, to format their documents, while also using the pdfmark macros to add PDF features. .XN -N pdfmark-operator -- The \F[C]pdfmark\F[] Operator

All PDF features are implemented by embedding instances of the \F[C]pdfmark\F[] operator, as described in the Adobe\*(rg .pdfmark-manual , into groff 's PostScript\*(rg output stream. To facilitate the use of this operator, the pdfmark macro package defines the primitive pdfmark macro; it simply emits its argument list, as arguments to a pdfmark operator, in the PostScript\*(rg output stream.

.pdfhref M -N pdfmark-example To illustrate the use of the pdfmark macro, the following is a much simplified example of how a bookmark may be added to a PDF document outline .QP ".pdfmark \e"

.fam C
/Count 2 \e
/Title (An Example of a Bookmark with Two Children) \e
/View [/FitH \en[PDFPAGE.Y]] \e
/OUT

In general, users should rarely need to use the pdfmark macro directly. In particular, the above example is too simple for general use; it .EM will create a bookmark, but it does .EM not address the issues of setting the proper value for the /Count key, nor of computing the PDFPAGE.Y value used in the /View key. The pdfmark macro package includes a more robust mechanism for creating bookmarks,
Here is an example of how a local reference may be planted,
using the automatic formatting feature of the "pdfhref" macro.

This is a forward reference to the named destination "add-outline",
which is defined below, using the "XN" wrapper macro, from the
"spdf.tmac" macro package. The automatically formatted reference
will be enclosed in parentheses, as specified by the use of
"-P" and "-A" options.

.pdfhref L -P ( -A ), -D add-outline
which addresses these issues automatically. Nevertheless, the pdfmark macro may be useful to users wishing to implement more advanced PDF features, than those currently supported directly by the pdfmark macro package. .XN -N docview -- Selecting an Initial Document View

By default, when a PDF document is opened, the first page will be displayed, at the default magnification set for the reader, and outline and thumbnail views will be hidden. When using a PDF reader, such as Acrobat\*(rg\~Reader, which supports the /DOCVIEW class of the pdfmark operator, these default initial view settings may be overridden, using the pdfview macro. For example .QP ".pdfview /PageMode /UseOutlines"

will cause Acrobat\*(rg\~Reader to open the document outline view, to the left of the normal page view, while .QP ".pdfview /PageMode /UseThumbs"

will open the thumbnail view instead.

Note that the two /PageMode examples, above, are mutually exclusive \(em it is not possible to have .EM both outline and thumbnail views open simultaneously. However, it .EM is permitted to add /Page and /View keys, to force the document to open at a page other than the first, or to change the magnification at which the document is initially displayed; see the .pdfmark-manual for more information.

It should be noted that the view controlling meta\(hydata, defined by the pdfview macro, is not written immediately to the PostScript\*(rg output stream, but is stored in an internal meta\(hydata "cache", (simply implemented as a groff diversion). This "cached" meta\(hydata must be written out later, by invoking the pdfsync macro,
Here is another example of how we may introduce a forward reference.
This time we are using the shorter notation afforded by the "XR" macro
provided by "spdf.tmac"; this example is equivalent to the native
"pdfmark.tmac" form
.pdfhref L -D pdfsync -P ( -A ).

.XR pdfsync ). (
.XN -N docinfo -- Adding Document Identification Meta-Data

In addition to the /DOCVIEW class of meta\(hydata described above, .XR docview ), ( we may also wish to include document identification meta\(hydata, which belongs to the PDF /DOCINFO class.

To do this, we use the pdfinfo macro. As an example of how it is used, the identification meta\(hydata attached to this document was specified using a macro sequence similar to:\(en I .pdfinfo /Title PDF Document Publishing with GNU Troff .pdfinfo /Author Keith Marshall .pdfinfo /Subject How to Exploit PDF Features with GNU Troff .pdfinfo /Keywords groff troff PDF pdfmark Notice that the pdfinfo macro is repeated, once for each /DOCINFO record to be placed in the document. In each case, the first argument is the name of the applicable /DOCINFO key, which .EM must be named with an initial solidus character; all additional arguments are collected together, to define the value to be associated with the specified key.

As is the case with the pdfview macro, .XR docview ), ( the /DOCINFO records specified with the pdfinfo macro are not immediately written to the PostScript\*(rg output stream; they are stored in the same meta\(hydata cache as /DOCVIEW specifications, until this cache is explicitly flushed, by invoking the pdfsync macro, .XR pdfsync ). ( .XN -N add-outline -- Creating a Document Outline

A PDF document outline comprises a table of references, to "bookmarked" locations within the document. When the document is viewed in an "outline\~aware" PDF document reader, such as Adobe\*(rg Acrobat\*(rg Reader, this table of "bookmarks" may be displayed in a document outline pane, or "Bookmarks" pane, to the left of the main document view. Individual references in the outline view may then be selected, by clicking with the mouse, to jump directly to the associated marked location in the document view.

The document outline may be considered as a collection of "hypertext" references to "bookmarked" locations within the document. The pdfmark macro package provides a single generalised macro, pdfhref , for creating and linking to "hypertext" reference marks. This macro will be described more comprehensively in a later section, .XR pdfhref ); ( the description here is restricted to its use for defining document outline entries. .XN -N basic-outline -- A Basic Document Outline

In its most basic form, the document outline comprises a structured list of headings, each associated with a marked location, or "bookmark", in the document text, and a specification for how that marked location should be displayed, when this bookmark is selected.

To create a PDF bookmark, the pdfhref macro is used, at the point in the document where the bookmark is to be placed, in the form .QP .fam C ".pdfhref O" level > < "descriptive text ..."

in which the reference class B O " " stipulates that this is an outline reference.

Alternatively, for those users who may prefer to think of a document outline simply as a collection of bookmarks, the pdfbookmark macro is also provided \(em indeed, pdfhref invokes it, when processing the B O " " reference class operator. It may be invoked directly, in the form .QP .fam C .pdfbookmark level > < "descriptive text ..."

Irrespective of which of the above macro forms is employed, the I level > < argument is required. It is a numeric argument, defining the nesting level of the "bookmark" in the outline hierarchy, with one being the topmost level. Its function may be considered analagous to the .EM "heading level" of the document's section headings, for example, as specified with the NH macro, if using the ms macros to format the document.

All further arguments, following the I level > < argument, are collected together, to specify the heading text which will appear in the document's outline view. Thus, the outline entry for this section of this document, which has a level three heading, might be specified as .QP .pdfhref O 3 \*(SN A Basic Document Outline

or, in the alternative form using the pdfbookmark macro, as .QP .pdfbookmark 3 \*(SN A Basic Document Outline .XN Hierarchical Structure in a Document Outline

When a document outline is created, using the pdfhref macro as described in
Here is an example of how we can temporarily modify the format of
a reference link, in this case to indicate only the section number
of the link target, in the form "section #", (or, if we define
"SECREF.BEGIN" before the call, its content followed by the
section number).

We first define a macro, which will get the reference data from
pdfhref, as arguments, and will return the formatted output, as we
require it, the string "PDFHREF.TEXT".

.while \\n(.$ \{\ . ie '\\$1'section' \{\ . if !dSECREF.BEGIN .ds SECREF.BEGIN \\$1 . ds PDFHREF.TEXT \\*[SECREF.BEGIN]\~\\$2 . rm SECREF.BEGIN . shift \\n(.$ . \} . el .shift . \} .. We now tell "pdfhref" to use our formatting macro, in place of
its builtin default formatter, before we specify the reference.

.pdfhref F SECREF .pdfhref L -A , -D basic-outline
At this point, we would normally revert the "pdfhref" formatter
to use its default, built in macro. However, in this particular
case, we want to use our custom format one more time, before we
revert it, so we will omit the reversion step this time.

and any entry is added at a nesting level greater than one, then a hierarchical structure is automatically defined for the outline. However, as was noted in the simplified .pdfhref L -D pdfmark-example -- example in .pdfhref L -A , -D pdfmark-operator
And now, we revert to default "pdfhref" formatting behaviour,
by completing the call we delayed above.

.pdfhref F
the data required by the pdfmark operator to create the outline entry may not be fully defined, when the outline reference is defined in the groff document source. Specifically, when the outline entry is created, its /Count key must be assigned a value equal to the number of its subordinate entries, at the next inner level of the outline hierarchy; typically however, these subordinate entries will be defined .EM later in the document source, and the appropriate /Count value will be unknown, when defining the parent entry.

To resolve this paradox, the pdfhref macro creates the outline entry in two distinct phases \(em a destination marker is placed in the PostScript\*(rg output stream immediately, when the outline reference is defined, but the actual outline entry is stored in an internal "outline cache", until its subordinate hierarchy has been fully defined; it can then be inserted in the output stream, with its /Count value correctly assigned. Effectively, to ensure integrity of the document outline structure, this means that each top level outline entry, and .EM all of its subordinates, are retained in the cache, until the .EM next top level entry is defined.

One potential problem, which arises from the use of the "outline cache", is that, at the end of any document formatting run, the last top level outline entry, and any subordinates defined after it, will remain in the cache, and will .EM not be automatically written to the output stream. To avoid this problem, the user should follow the guidelines given in
Here is a more conventional example of how to temporarily change
to the format used to display reference links. We will again use
the "SECREF" format, which we defined above, but on this occasion
we will immediately revert to the default format, after the link
has been placed.

.pdfhref F SECREF .pdfhref L -D pdfsync -A , .pdfhref F
to synchronise the output state with the cache state, .XR pdfsync ), ( at the end of the groff formatting run. .XN -N outline-view -- Associating a Document View with an Outline Reference

Each "bookmark" entry, in a PDF document outline, is associated with a specific document view. When the reader selects any outline entry, the document view changes to display the document context associated with that entry.

The document view specification, to be associated with any document outline entry, is established at the time when the outline entry is created. However, rather than requiring that each individual use of the pdhref macro, to create an outline entry, should include its own view specification, the actual specification assigned to each entry is derived from a generalised specification defined in the string PDFBOOKMARK.VIEW , together with the setting of the numeric register PDFHREF.VIEW.LEADING , which determine the effective view specification as follows:\(en .QS

If desired, the view specification may be changed, by redefining the string PDFBOOKMARK.VIEW , and possibly also the numeric register PDFHREF.VIEW.LEADING . Any alternative definition for PDFBOOKMARK.VIEW .EM must be specified in terms of valid view specification parameters, as described in the Adobe\*(rg .pdfmark-manual .

Note the use of the register PDFPAGE.Y , in the default definition of PDFBOOKMARK.VIEW above. This register is computed by pdfhref , when creating an outline entry; it specifies the vertical position of the "bookmark", in basic groff units, relative to the .EM bottom edge of the document page on which it is defined, and is followed, in the PDFBOOKMARK.VIEW definition, by the grops u " " operator, to convert it to PostScript\*(rg units on output. It may be used in any redefined specification for PDFBOOKMARK.VIEW , (or in the analogous definition of PDFHREF.VIEW , described in .XR-NO-PREFIX pdfhref-view ), but .EM not in any other context, since its value is undefined outside the scope of the pdfhref macro.

Since PDFPAGE.Y is computed relative to the .EM bottom of the PDF output page, it is important to ensure that the page length specified to troff correctly matches the size of the logical PDF page. This is most effectively ensured, by providing .EM identical page size specifications to groff , grops and to the PostScript\*(rg to PDF converter employed, and avoiding any page length changes within the document source.

Also note that PDFPAGE.Y is the only automatically computed "bookmark" location parameter; if the user redefines PDFBOOKMARK.VIEW , and the modified view specification requires any other positional parameters, then the user .EM must ensure that these are computed .EM before invoking the pdfhref macro. .XN -N outline-folding -- Folding the Outline to Conceal Less Significant Headings

When a document incorporates many subheadings, at deeply nested levels, it may be desirable to "fold" the outline such that only the major heading levels are initially visible, yet making the inferior subheadings accessible, by allowing the reader to expand the view of any heading branch on demand.

The pdfmark macros support this capability, through the setting of the PDFOUTLINE.FOLDLEVEL register. This register should be set to the number of heading levels which it is desired to show in expanded form, in the .EM initial document outline display; all subheadings at deeper levels will still be added to the outline, but will not become visible until the outline branch containing them is expanded. 'ne 5 For example, the setting used in this document: .QS .fam C .\e" Initialise the outline view to show only three heading levels, .\e" with additional subordinate level headings folded. .\e" .nr PDFOUTLINE.FOLDLEVEL 3 .QE

results in only the first three levels of headings being displayed in the document outline, .EM until the reader chooses to expand the view, and so reveal the lower level headings in any outline branch.

The initial default setting of PDFOUTLINE.FOLDLEVEL , if the document author does not choose to change it, is 10,000. This is orders of magnitude greater than the maximum heading level which is likely to be used in any document; thus the default behaviour will be to show document outlines fully expanded, to display all headings defined, at all levels within each document.

The setting of PDFOUTLINE.FOLDLEVEL may be changed at any time; however, the effect of each such change may be difficult to predict, since it is applied not only to outline entries which are defined .EM after the setting is changed, but also to any entries which remain in the outline cache, .EM at this time. Therefore, it is recommended that PDFOUTLINE.FOLDLEVEL should be set .EM once , at the start of each document; if it .EM is deemed necessary to change it at any other time, the outline cache should be flushed, .XR pdfsync ), ( .EM immediately before the change, which should immediately preceed a level one heading. .XN -N multipart-outline -- Outlines for Multipart Documents

When a document outline is created, using the pdfhref macro, each reference mark is automatically assigned a name, composed of a fixed stem followed by a serially generated numeric qualifier. This ensures that, for each single part document, every outline reference has a uniquely named destination.

As the overall size of the PDF document increases, it may become convenient to divide it into smaller, individually formatted PostScript\*(rg components, which are then assembled, in the appropriate order, to create a composite PDF document. While this strategy may simplify the overall process of creating and editing larger documents, it does introduce a problem in creating an overall document outline, since each individual PostScript\*(rg component will be assigned duplicated sequences of "bookmark" names, with each name ultimately referring to multiple locations in the composite document. To avoid such reference naming conflicts, the pdfhref macro allows the user to specify a "tag", which is appended to the automatically generated "bookmark" name; this may be used as a discriminating mark, to distinguish otherwise similarly named destinations, in different sections of the composite document.

To create a "tagged" document outline, the syntax for invocation of the pdfhref macro is modified, by the inclusion of an optional "tag" specification, .EM before the nesting level argument, i.e. .QP .fam C ".pdfhref O" -T [ tag >] < level > < "descriptive text ..."

The optional I tag > < argument may be composed of any characters of the user's choice; however, its initial character .EM "must not" be any decimal digit, and ideally it should be kept short \(em one or two characters at most.

By employing a different tag in each section, the user can ensure that "bookmark" names remain unique, throughout all the sections of a composite document. For example, when using the spdf.tmac macro package, which adds pdfmark capabilities to the standard ms package, .XR using-spdf ), ( the table of contents is collected into a separate PostScript\*(rg section from the main body of the document. In the "body" section, the document outline is "untagged", but in the "Table\~of\~Contents" section, a modified version of the TC macro adds an outline entry for the start of the "Table\~of\~Contents", invoking the pdfhref macro as .QP ".pdfhref O -T T 1 \e\e*[TOC]"

to tag the associated outline destination name with the single character suffix, T ". " Alternatively, as in the case of the basic outline, .XR basic-outline ), ( this may equally well be specified as .QP ".pdfbookmark -T T 1 \e\e*[TOC]" .XN Delegation of the Outline Definition

Since the most common use of a document outline is to provide a quick method of navigating through a document, using active "hypertext" links to chapter and section headings, it may be convenient to delegate the responsibility of creating the outline to a higher level macro, which is itself used to define and format the section headings. This approach has been adopted in the spdf.tmac package, to be described later, .XR using-spdf ). (

When such an approach is adopted, the user will rarely, if ever, invoke the pdfhref macro directly, to create a document outline. For example, the structure and content of the outline for this document has been exclusively defined, using a combination of the NH macro, from the ms package, to establish the structure, and the XN macro from spdf.tmac , to define the content. In this case, the responsibility for invoking the pdfhref macro, to create the document outline, is delegated to the XN macro. .XN -N pdfhref -- Adding Reference Marks and Links

.pdfhref F SECREF .pdfhref L -D add-outline .pdfhref F has shown how the pdfhref macro may be used to create a PDF document outline. While this is undoubtedly a powerful capability, it is by no means the only trick in the repertoire of this versatile macro.

The macro name, pdfhref , which is a contraction of "PDF HyperText Reference", indicates that the general purpose of this macro is to define .EM any type of dynamic reference mark, within a PDF document. Its generalised usage syntax takes the form .QP .fam C .pdfhref class > < "-options ..." ] [ [--] "descriptive text ..." ] [

where <\f(CIclass> represents a required single character argument, which defines the specific reference operation to be performed, and may be selected from:\(en .QS

The behaviour of a number of the pdfhref macro operations can be modified, by including .EM "option specifiers" " " after the operation specifying argument, but .EM before any other arguments normally associated with the operation. In .EM all cases, an option is specified by an .EM "option flag" ", " comprising an initial hyphen, followed by one or two option identifying characters. Additionally, .EM some options require .EM "exactly one" option argument; for these options, the argument .EM must be specified, and it .EM must be separated from the preceding option flag by one or more .EM spaces , (tabs .EM "must not" be used). It may be noted that this paradigm for specifying options is reminiscent of most UNIX\(tm shells; however, in the case of the pdfhref macro, omission of the space separating an option flag from its argument is .EM never permitted.

A list of .EM all general purpose options supported by the pdfhref macro is given below. Note that not all options are supported for all pdfhref operations; the operations affected by each option are noted in the list. For .EM most operations, if an unsupported option is specified, it will be silently ignored; however, this behaviour should not be relied upon.

The general purpose options, supported by the pdfhref macro, are:\(en .QS

In addition to the pdfhref options listed above, a supplementary set of two character options are defined. These supplementary options, listed below, are intended for use with the B L " " operator, in conjunction with the B -F \~\c BI file > < option, to specify alternate file names, in formats compatible with the file naming conventions of alternate operating systems; they will be silently ignored, if used in any other context.

The supported alternate file name options, which are ignored if the B -F \~\c BI file > < option is not specified, are:\(en .QS

The pdfhref macro may be used to create active links to any Internet resource, specified by its URI , or to any "named destination", either within the same document, or in another PDF document. Although the PDF specification allows link destinations to be defined in terms of a page number, and an associated view specification, this style of reference is not currently supported by the pdfhref macro, because it is not possible to adequately bind the specification for the destination with the intended reference context.

References to Internet resources are interpreted in accordance with the W3C standard for defining a URI ; hence the only prerequisite, for creating a link to any Internet resource, is that the URI be properly specified, when declaring the reference; .XR add-weblink ). ( In the case of references to "named destinations" in PDF documents, however, it is necessary to provide a mechanism for creating such "named destinations". This may be accomplished, by invoking the pdfhref macro in the form .QP .fam C ".pdfhref M" -N [ name >] < -X ] [ -E ] [ "descriptive text ..." ] [

This creates a "named destination" reference mark, with its name specified by I name > < , or, if the B -N option is not specified, by the first word of I descriptive \~\c I text ; (note that this imposes the restriction that, if the B -N option is omitted, then .EM "at least" one word of I descriptive \~\c I text .EM must be specified). Additionally, a reference view will be automatically defined, and associated with the reference mark, .XR pdfhref-view ), ( and, if any
.CWI descriptive
.CWI text
is specified, or the
and, if the B -X option is specified, and no document cross reference map has been imported, .XR import-map ), ( then a cross reference mapping record, .XR export-map ), ( will be written to the stdout stream; this may be captured, and subsequently used to generate a cross reference map for the document, .XR create-map ). (

When a "named destination" reference mark is created, using the pdfhref macro's B M " " operator, there is normally no visible effect in the formatted document; any I descriptive \~\c I text which is specified will simply be stored in the cross reference map, for use when a link to the reference mark is created. This default behaviour may be changed, by specifying the B -E option, which causes any specified I descriptive \~\c I text to be "echoed" in the document text, at the point where the reference mark is placed, in addition to its inclusion in the cross reference map. .XN -N export-map -- Mapping a Destination for Cross Referencing

Effective cross referencing of .EM any document formatted by groff requires multiple pass formatting. Details of how this multiple pass formatting may be accomplished, when working with the pdfmark macros, will be discussed later, .XR do-xref ); ( at this stage, the discussion will be restricted to the initial preparation, which is required at the time when the cross reference destinations are defined.

The first stage, in the process of cross referencing a document, is the generation of a cross reference map. Again, the details of .EM how the cross reference map is generated will be discussed in .pdfhref F SECREF L -D do-xref -A ; .pdfhref F however, it is important to recognise that .EM what content is included in the cross reference map is established when the reference destination is defined \(em it is derived from the reference data exported on the stderr stream by the pdfhref macro, when it is invoked with the B M " " operator, and is controlled by whatever definition of the string PDFHREF.INFO is in effect, when the pdfhref macro is invoked.

The initial default setting of PDFHREF.INFO is .QP ".ds PDFHREF.INFO page \e\en% \e\e$*"

which ensures that the cross reference map will contain at least a page number reference, supplemented by any I descriptive \~\c I text which is specified for the reference mark, as defined by the pdfhref macro, with its B M " " operator; this may be redefined by the user, to export additional cross reference information, or to modify the default format for cross reference links, .XR set-format ). ( .XN -N pdfhref-view -- Associating a Document View with a Reference Mark

In the same manner as each document outline reference, defined by the pdfhref macro with the B O " " operator, .XR add-outline ), ( has a specific document view associated with it, each reference destination marked by pdfhref with the B M " " operator, requires an associated document view specification.

The mechanism whereby a document view is associated with a reference mark is entirely analogous to that employed for outline references, .XR outline-view ), ( except that the PDFHREF.VIEW string specification is used, in place of the PDFBOOKMARK.VIEW specification. Thus, the reference view is defined in terms of:\(en .QS

Just as the view associated with outline references may be changed, by redefining PDFBOOKMARK.VIEW , so the view associated with marked reference destinations may be changed, by redefining PDFHREF.VIEW , and, if desired, PDFHREF.VIEW.LEADING ; such changes will become effective for all reference destinations marked .EM after these definitions are changed. (Notice that, since the specification of PDFHREF.VIEW.LEADING is shared by both outline reference views and marked reference views, if it is changed, then the views for .EM both reference types are changed accordingly).

It may again be noted, that the PDFPAGE.Y register is used in the definition of PDFHREF.VIEW , just as it is in the definition of PDFBOOKMARK.VIEW ; all comments in .pdfhref F SECREF L -D outline-view .pdfhref F relating to its use, and indeed to page position computations in general, apply equally to marked reference views and to outline reference views. .XN -N link-named -- Linking to a Marked Reference Destination

Any named destination, such as those marked by the pdfhref macro, using it's B M " " operator, may be referred to from any point in .EM any PDF document, using an .EM "active link" ; such active links are created by again using the pdfhref macro, but in this case, with the B L " " operator. This operator provides support for two distinct cases, depending on whether the reference destination is defined in the same document as the link, .XR link-intern ), ( or is defined as a named destination in a different PDF document, .XR link-extern ). ( .XN -N link-intern -- References within a Single PDF Document

The general syntactic form for invoking the pdfhref macro, when creating a link to a named destination within the same PDF document is .QP .fam C .pdfhref L -D [ dest-name >] < -P [ prefix-text >] < -A [ affixed-text >] < \e

\0\0\0 -X ] [ -- ] [ "descriptive text ..." ] [

where I dest-name > < specifies the name of the link destination, as specified using the pdfhref B M " " operation; (it may be defined either earlier in the document, to create a backward reference, or later, to create a forward reference).
Here's a example of how to add an iconic annotation.

.pdfnote -T "Internal Cross References" \
This description is rather terse, and could benefit from \
the inclusion of an example.

If any I descriptive \~\c I text arguments are specified, then they will be inserted into the groff output stream, to define the text appearing in the "hot\(hyspot" region of the link; this will be printed in the link colour specified by the string, PDFHREF.TEXT.COLOUR , which is described in .XR-NO-PREFIX set-colour . If the B -X option is also specified, then the I descriptive \~\c I text will be augmented, by prefacing it with page and section number indicators, in accordance with the reference formatting rules which are in effect, .XR set-format ); ( such indicators will be included within the active link region, and will also be printed in the link colour.

Note that .EM either the B -D \~\c BI dest\(hyname > < option, .EM or the I descriptive \~\c I text arguments, .EM "but not both" , may be omitted. If the B -D \~\c BI dest\(hyname > < option is omitted, then the first word of I descriptive \~\c I text , i.e.\~all text up to but not including the first space, will be interpreted as the BI dest\(hyname > < for the link; this text will also appear in the running text of the document, within the active region of the link. Alternatively, if the B -D \~\c BI dest\(hyname > < option .EM is specified, and I descriptive \~\c I text is not, then the running text which defines the reference, and its active region, will be derived from the reference description which is specified when the named destination is marked, .XR mark-dest ), ( and will be formatted according to the reference formatting rules which are in effect, when the reference is placed, .XR set-format ); ( in this case, it is not necessary to specify the B -X option to activate automatic formatting of the reference \(em it is implied, by the omission of all I descriptive \~\c I text arguments.

The B -P \~\c BI prefix\(hytext > < and B -A \~\c BI affixed\(hytext > < options may be used to specify additional text which will be placed before and after the linked text respectively, with no intervening space. Such prefixed and affixed text will be printed in the normal text colour, and will not be included within the active region of the link. This feature is mostly useful for creating parenthetical references, or for placing punctuation adjacent to, but not included within, the text which defines the active region of the link.

The operation of the pdfhref macro, when used with its B L " " operator to place a link to a named PDF destination, may best be illustrated by an example. However, since the appearance of the link will be influenced by factors established when the named destination is marked, .XR mark-dest ), ( and also by the formatting rules in effect when the link is placed, the presentation of a suitable exanple will be deferred, until the formatting mechanism has been explained, .XR set-format ). ( .XN -N link-extern -- References to Destinations in Other PDF Documents

The pdfhref macro's B L " " operator is not restricted to creating reference links within a single PDF document. When the link destination is defined in a different document, then the syntactic form for invoking pdfhref is modified, by the addition of options to specify the name and location of the PDF file in which the destination is defined. Thus, the extended pdfhref syntactic form becomes .QP .fam C .pdfhref L -F file > < -D [ dest-name >] < \e

\0\0\0 -DF [ dos-file >] < -MF [ mac-file >] < -UF [ unix-file >] < \e

\0\0\0 -WF [ win-file >] < -P [ prefix-text >] < -A [ affixed-text >] < \e

\0\0\0 -X ] [ -- ] [ "descriptive text ..." ] [

where the B -F \~\c BI file > < option serves .EM two purposes: it both indicates to the pdfhref macro that the specified reference destination is defined in an external PDF file, and it also specifies the normal path name, which is to be used to locate this file, when a user selects the reference.

In addition to the B -F \~\c BI file > < option, which .EM must be specified when referring to a destination in an external PDF file, the B -DF \~\c BI dos\(hyfile > < , B -MF \~\c BI mac\(hyfile > < , B -UF \~\c BI unix\(hyfile > < and B -WF \~\c BI win\(hyfile > < options may be used to specify the location of the file containing the reference destination, in a variety of operating system dependent formats. These options assign their arguments to the /DosFile , /MacFile , /UnixFile and /WinFile keys of the generated pdfmark respectively; thus when any of these options are specified, .EM "in addition to" the B -F \~\c BI file > < option, and the document is read on the appropriate operating systems, then the path names specified by BI dos\(hyfile > < , BI mac\(hyfile > < , BI unix\(hyfile > < and BI win\(hyfile > < will be searched, .EM instead of the path name specified by BI file > < , for each of the MS\(hyDOS \*(rg, Apple \*(rg Macintosh \*(rg, UNIX \(tm and MS\(hyWindows \*(rg operating systems, respectively; see the .pdfmark-manual , for further details.

Other than the use of these additional options, which specify that the reference destination is in an external PDF file, the behaviour of the pdfhref B L " " operator, with the B -F \~\c BI file > < option, remains identical to its behaviour .EM without this option, .XR link-intern ), ( with respect to the interpretation of other options, the handling of the I descriptive \~\c I text arguments, and the formatting of the displayed reference.

Once again, since the appearance of the reference is determined by factors specified in the document reference map, and also by the formatting rules in effect when the reference is placed, the presentation of an example of the placing of a reference to an external destination will be deferred, until the formatting mechanism has been explained, .XR set-format ). ( .XN -N add-weblink -- Linking to Internet Resources

In addition to supporting the creation of cross references to named destinations in PDF documents, the pdfhref macro also has the capability to create active links to Internet resources, or indeed to .EM any resource which may be specified by a Uniform Resource Identifier, (which is usually abbreviated to the acronym "URI", and sometimes also referred to as a Uniform Resource Locator, or "URL").

Since the mechanism for creating a link to a URI differs somewhat from that for creating PDF references, the pdfhref macro is invoked with the B W " " (for "web\(hylink") operator, rather than the B L " " operator; nevertheless, the invocation syntax is similar, having the form .QP .fam C .pdfhref W -D [ URI >] < -P [ prefix-text >] < -A [ affixed-text >] < \e

\0\0\0 -- ] [ "descriptive text ..."

where the optional B -D BI URI > < modifier specifies the address for the target Internet resource, in any appropriate .EM "Uniform Resource Identifier" format, while the I descriptive I text argument specifies the text which is to appear in the "hot\(hyspot" region, and the B -P BI prefix\(hytext > < and B -A BI affixed\(hytext > < options have the same effect as in the case of local document links, .XR link-intern ). (

Notice that it is not mandatory to include the B -D BI URI > < in the link specification; if it .EM is specified, then it is not necessary for the URI to appear, in the running text of the document \(em the I descriptive I text argument exactly defines the text which will appear within the "hot\(hyspot" region, and this need not include the URI. However, if the B -D \~\c BI URI > < specification is omitted, then the I descriptive I text argument .EM must be an .EM exact representation of the URI, which .EM will , therefore, appear as the entire content of the "hot\(hyspot". For example, we could introduce a reference to .pdfhref W -D \*[GROFF-WEBSITE] -A , the groff web site in which the actual URI is concealed, by using mark up such as:\(en I For example, we could introduce a reference to .pdfhref W -D \*[GROFF-WEBSITE] -A , the groff web site in which the actual URI is concealed, Alternatively, to refer the reader to the groff web site, making it obvious that the appropriate URI is .pdfhref W -A , \*[GROFF-WEBSITE] the requisite mark up might be:\(en I to refer the reader to the groff web site, making it obvious that the appropriate URI is .pdfhref W -A , \*[GROFF-WEBSITE] the requisite mark up might be:\e(en .XN -N set-format -- Establishing a Format for References

There are two principal aspects to be addressed, when defining the format to be used when displaying references. Firstly, it is desirable to provide a visual cue, to indicate that the text describing the reference is imbued with special properties \(em it is dynamically linked to the reference destination \(em and secondly, the textual content should describe where the link leads, and ideally, it should also describe the content of the reference destination.

The visual cue, that a text region defines a dynamically linked reference, is most commonly provided by printing the text within the active region in a distinctive colour. This technique will be employed automatically by the pdfhref macro \(em .XR set-colour \(em unless the user specifically chooses to adopt, and implement, some alternative strategy. .XN -N set-colour -- Using Colour to Demarcate Link Regions

Typically, when a PDF document contains .EM active references to other locations, either within the same document, or even in other documents, or on the World Wide Web, it is usually desirable to make the regions where these active links are placed stand out from the surrounding text. .XN -N user-format -- Specifying Reference Text Explicitly .XN -N auto-format -- Using Automatically Formatted Reference Text .XN -N custom-format -- Customising Automatically Formatted Reference Text

It is incumbent on the user, if employing automatic formatting of the displayed reference, .XR set-format ), ( to ensure that an appropriate reference definition is created for the reference destination, and is included in the reference map for the document in which the reference will appear; thus, it may be easiest to .EM always use manual formatting for external references. .XN Problematic Links

Irrespective of whether a pdfhref reference is placed using the B L " " operator, or the B W " " operator, there may be occasions when the resulting link does function as expected. A number of scenarios, which are known to be troublesome, are described below. .XN -N page-trap -- Links with a Page Transition in the Active Region

When a link is placed near the bottom of a page, it is possible that its active region, or "hot\(hyspot", may extend on to the next page. In this situation, a page trap macro is required to intercept the page transition, and to restart the mapping of the "hot\(hyspot" boundary on the new page.

The pdfmark macro package includes a suitable page trap macro, to satisfy this requirement. However, to avoid pre\(hyempting any other requirement the user may have for a page transition trap, this is .EM not installed as an active page trap, unless explicitly requested by the user.

To enable proper handling of page transitions, which occur within the active regions of reference links, the user should:\(en .QS .nr ITEM 0 1

It may be observed that this initialisation of the pdfhref page transition hook is, typically, required only once .EM before document formatting begins. Users of document formatting macro packages may reasonably expect that this initialisation should be performed by the macro package itself. Thus, writers of such macro packages which include pdfmark bindings, should provide appropriate initialisation, so relieving the end user of this responsibility. The following example, abstracted from the sample ms binding package, spdf.tmac , illustrates how this may be accomplished:\(en I .\e" groff "ms" provides the "pg@bottom" macro, which has already .\e" been installed as a page transition trap. To ensure proper .\e" mapping of "pdfhref" links which overflow the bottom of any .\e" page, we need to install the "pdfhref" page transition hook, .\e" as an addendum to this macro. . .pdfhref I -PT pg@bottom .XN -N add-note -- Annotating a PDF Document using Pop-Up Notes .XN -N pdfsync -- Synchronising Output and \F[C]pdfmark\F[] Contexts

It has been noted previously, that the pdfview macro, .XR docview ), ( the pdfinfo macro, .XR docinfo ), ( and the pdfhref macro, when used to create a document outline, .XR add-outline ), ( do not immediately write their pdfmark output to the PostScript\*(rg data stream; instead, they cache their output, in a groff diversion, in the case of the pdfview and pdfinfo macros, or in an ordered collection of strings and numeric registers, in the case of the document outline, until a more appropriate time for copying it out. In the case of pdfview and pdfinfo "meta\(hydata", this "more appropriate time" is explicitly chosen by the user; in the case of document outline data, .EM some cached data may be implicitly written out as the document outline is compiled, but there will .EM always be some remaining data, which must be explicitly flushed out, before the groff formatting process is allowed to complete.

To allow the user to choose when cached pdfmark data is to be flushed to the output stream, the pdfmark macro package provides the pdfsync macro, (to synchronise the cache and output states). In its simplest form, it is invoked without arguments, i.e. .QP .fam C .pdfsync

This form of invocation ensures that .EM both the "meta\(hydata cache", containing pdfview and pdfinfo data, .EM and the "outline cache", containing any previously uncommitted document outline data, are flushed; ideally, this should be included in a groff "end macro", to ensure that .EM both caches are flushed, before groff terminates.

Occasionally, it may be desirable to flush either the "meta\(hydata cache", without affecting the "outline cache", or vice\(hyversa, at a user specified time, prior to reaching the end of the document. This may be accomplished, by invoking the pdfsync macro with an argument, i.e. .QP .fam C ".pdfsync M"

to flush only the "meta\(hydata cache", or .QP .fam C ".pdfsync O"

to flush only the "outline cache".

The "meta\(hydata cache" can normally be safely flushed in this manner, at any time .EM after output of the first page has started; (it may cause formatting problems, most notably the appearance of unwanted white space, if flushed earlier, or indeed, if flushed immediately after a page transition, but before the output of the content on the new page has commenced). Caution is required, however, when explicitly flushing the "outline cache", since if the outline is to be subsequently extended, then the first outline entry after flushing .EM must be specified at level 1. Nevertheless, such explict flushing may occasionally be necessary; for example, the TC macro in the spdf.tmac package, .XR using-spdf ), ( invokes ".pdfsync O" " " to ensure that the outline for the "body" section of the document is terminated, .EM before it commences the formatting of the table of contents section. .bp .XN -N pdf-layout -- PDF Document Layout

The pdfmark macros described in the preceding section, .XR pdf-features ), ( provide no inherent document formatting capability of their own. However, they may be used in conjunction with any other groff macro package of the user's choice, to add such capability.

In preparing this document, the standard ms macro package, supplied as a component of the GNU Troff distribution, has been employed. To facilitate the use of the pdfmark macros with the ms macros, a binding macro package, spdf.tmac , has been created. The use of this binding macro package is described in the following section, .XR using-spdf ); ( it may also serve as an example to users of other standard groff macro packages, as to how the pdfmark macros may be employed with their chosen primary macro package. .XN -N using-spdf -- Using \F[C]pdfmark\F[] Macros with the \F[C]ms\F[] Macro Package

The use of the binding macro package, spdf.tmac , allows for the use of the pdfmark macros in conjunction with the ms macros, simply by issuing a groff command of the form .QP .fam C groff -Tps -mspdf "-options ..." ] [ file(s) ...

When using the spdf.tmac package, the groff input files may be marked up using any of the standard ms macros to specify document formatting, while PDF features may be added, using any of the pdfmark macros described previously, .XR pdf-features ). ( Additionally, spdf.tmac defines a number of convenient extensions to the ms macro set, to better accomodate the use of PDF features within the ms formatting framework, and to address a number of ms document layout issues, which require special handling when producing PDF documents. These additional macros, and the issues they are intended to address, are described below. .XN \F[C]ms\F[] Section Headings in PDF Documents

Traditionally, ms provides the NH and SH macros, to specify section headings. However, there is no standard mechanism for generating a table of contents entry based on the text of the section heading; neither is there any recognised standard method for establishing a cross reference link to the section.

To address this ms limitation, spdf.tmac defines the XN macro, .XR xn-macro ), ( to be used in conjunction with the NH macro. .XN -N xn-macro -- The \F[C]XN\F[] Macro .XN The PDF Publishing Process .XN -N do-xref -- Resolving Cross References .XN -N create-map -- Creating a Document Reference Map .XN -N import-map -- Deploying a Document Reference Map .TC