% pdfTeX manual, rewritten in LaTeX from the ConTeXt original, % Karl Berry, 2023. Content not significantly changed. % Released under the GFDL v1.2+, same as the ConTeXt version. % % Not quite tagging-ready, numerous warnings like: % Package tagpdf Warning: The structure Sect can not be closed. % and ends with: % ./pdftex.tex:4336: Package tagpdf Error: The number of automatic begin % (1351) and end (1352) ... % %\DocumentMetadata{testphase={phase-III,table}} \documentclass{pdftexmanual} \svnscan $Id: pdftex.tex 924 2024-02-23 02:20:18Z karl $ % \title{The pdf\TeX\ user manual} \author{\THANH\ and others} \date{\PDFTEX\ \currentpdftex\\ \hspace*{1em}\\ \today\\ \hspace*{1em}\\ \url{https://pdftex.org} } \begin{document} \hypersetup{pageanchor=false} % avoid hyperref duplicate page.1 warning \maketitle \hypersetup{pageanchor=true} % https://tex.stackexchange.com/questions/18924 \setcounter{page}{2} \tableofcontents \chapter{Introduction} The main purpose of the \PDFTEX\ project is to create and maintain an extension of \TEX\ that can produce \PDF\ output directly from \TEX\ source files and improve\slash enhance the result of \TEX\ typesetting with the help of \PDF\ output. When \PDF\ output is not selected, \PDFTEX\ produces standard \DVI\ output. An important aspect of the project is to investigate alternative justification algorithms; notably, the margin kerning and font expansion algorithms following the \HZ\ micro\hyph typography algorithm by Prof.~Hermann Zapf. \PDFTEX\ is maintained by \THANH, the original author, and others. The \PDFTEX\ home page is \useurl{ptexorg}. Please send bug reports, suggestions, etc., to the mailing list (\useurl{ptexbugs}). \PDFTEX\ is based on the original \TEX\ sources and \WEBC, and has been successfully compiled on \UNIX, \WIN\ and many other systems. It is actively maintained, with great care taken to keep new \PDFTEX\ versions backward\hyph compatible with earlier ones. A conservative successor to \TEX, named \ETEX, was developed in the 1990s. Since \PDFTEX\ version 1.40, the \ETEX\ extensions are compiled as part of the \PDFTEX\ engine and thus always available. For documentation on the \ETEX\ extensions, see \useurl{etexctan}. Furthermore, \PDFTEX\ itself has acquired plenty of extensions over the years which are not related specifically to \PDF\ output, generally new primitives for various features that are inconvenient or impossible to implement at the \TEX\ level. Many of these extensions have been adopted across all engines (sometimes with different primitive names or variant functionality), and some are required by \LATEX. Therefore, in most distributions \type{etex} is a link to \type{pdftex}; the difference being whether \DVI\ or \PDF\ output is generated by default. Other extensions are \MLTEX\ and \ENCTEX; these are also included in the \PDFTEX\ implementation, although they are rarely used for new documents. \PDFTEX\ does not natively support \UTFEIGHT\ input text, Unicode-encoded fonts, or anything else related to Unicode, as it was written long before Unicode became widespread. Making those changes to the engine now would necessarily create unacceptable incompatibilities, so there are no plans to do so. Thus, when using \PDFTEX, \LaTeX\ and other formats handle \UTFEIGHT\ (and other) input at the \TEX\ macro level, which works well enough in practice for most documents. It is also possible to use TrueType and OpenType fonts with \PDFTEX, if you choose an 8-bit subset to be encoded. If you need a \TeX\ engine with native support for Unicode input, TrueType fonts, OpenType fonts, please look into \LUATEX\ (\useurl{luatex}) or \XETEX\ (\useurl{xetex}). \section{About this manual} This manual revision (\rcsrevision) covers \PDFTEX\ development up to version \currentpdftex. The primary repository for both the manual and the \PDFTEX\ sources is \useurl{ptexdevel}. The typeset manual in \PDF\ format can be found on \CTAN\ in \useurl{ptexctan}. Thanks to the many people who have contributed to the manual. Improvements are always possible, and bugs not unlikely. Please send questions or suggestions via email at \useurl{ptexbugs}. \section{Legal notice} Copyright \copyright\ 1996--2024 \THANH. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled ``GNU Free Documentation License''. \section{About \PDF} Let's start with a very brief introduction to \PDF\ format. For our example, the bit of \TeX\ source below (figure~\ref{fig.titlesource}) generates the nearly-minimal \PDF\ file shown on the next page (figure~\ref{fig.titleout}). Since compression is disabled, such a \PDF\ file is rather verbose and readable. The first line (\texttt{\%PDF-1.4}) specifies the \PDF\ version used. \PDF\ viewers are supposed to silently skip over all elements they cannot handle. \begin{figure}[b] \bigskip \hrule \medskip \begin{verbatim} \pdfoutput=1 \pdfcompresslevel=0 \pdfobjcompresslevel=0 \pdfmapline{ptmr8r Times-Roman 2 <8r.enc} \font\tenrm=ptmr8r \tenrm Welcome to pdf\TeX! \end \end{verbatim} \vspace{-\baselineskip} \caption{This plain \TeX\ source generates \PDF\ output shown in figure~\ref{fig.titleout}.} \label{fig.titlesource} \end{figure} \begin{figure}[p] % Why isn't \vsize enough? \includegraphics[height=1.27\vsize,alt={small PDF file}] {incl/pdfmin-crop.pdf} \caption{The \PDF\ output for the \TeX\ source in figure~\ref{fig.titlesource}.} \label{fig.titleout} \end{figure} A \PDF\ file consists of objects. These objects can be recognized by their number and keywords. For example, over in the second column, we can see (modulo editorial line breaks): \begin{verbatim} 9 0 obj << /Type /Catalog /Pages 5 0 R >> endobj \end{verbatim} Here \verb|9 0 obj ... endobj| is the object capsule. The first number is the object number (no.~9). The sequence \verb|5 0 R| is an object reference, that is, a pointer to another object (no.~5). The second number (here a zero) is currently not used in \PDFTEX; it is the version number of the object. It is, for instance, used by \PDF\ editors, when they replace objects by new ones. When a viewer opens a \PDF\ file, it goes to the end of the file, looking for the keyword \type{startxref}. The number after \type{startxref} gives the absolute position (byte offset from the file start) of the so\hyph called ``object cross-reference table'' that begins with the keyword \type{xref}. This table in turn tells the byte offsets of all objects that make up the \PDF\ file, providing fast random access to the individual objects (here the \type{xref} table shows 11~objects, numbered from~0 to~10; object no.~0 is always unused). The actual starting point of the file's object structure is defined after the \type{trailer}: the \type{/Root} entry points to the \type{/Catalog} object (no.~9). In this object the viewer can find the pointer \type{/Pages} to the page list object (no.~5). In our example we have only one page. The trailer also usually holds an \type{/Info} entry, which points to an object (no.~10) with a bit more about the document. We can follow the thread: \begin{quote} \type{/Root} $\longrightarrow$ object~9 $\longrightarrow$ \type{/Pages} $\longrightarrow$ object~5 $\longrightarrow$ \type{/Kids} $\longrightarrow$ object~2 $\longrightarrow$ \type{/Contents} $\longrightarrow$ object~3 \end{quote} As soon as we add annotations, a fancy word for hyperlinks and the like, some more entries will be present in the catalog. We invite users to take a look at the \PDF\ code of this file to get an impression of that. The page content is a stream of drawing operations. Such a stream can be compressed, where the level of compression can be set with \cs{pdfcompresslevel} (compression is switched off for the title page). Let's take a closer look at this stream in object~3. Often (but not in our example) there is a transformation matrix, six numbers followed by \type{cm}. As in \POSTSCRIPT, the operator comes after the operands. Between \type{BT} and \type{ET} comes the text. A font is selected by a \type{Tf} operator, which is given a font resource name \type{/F..} and the font size. The actual text goes into parentheses \type{()} so that it creates a \POSTSCRIPT\ string. The numbers in bracket pairs provide horizontal movements like spaces and fine glyph positioning (kerning). When one analyzes a file produced by a less sophisticated typesetting engine, whole sequences of words can be recognized. In \PDF\ files generated by \PDFTEX\ however, many words come out rather fragmented, mainly because a lot of kerning takes place; in our example the \type{80} moves the text \type{(elcome)} left towards the letter \type{(W)} by 80/1000 of the font size. \PDF\ viewers in search mode simply ignore the kerning information in these text streams. When a document is searched, the search engine reconstructs the text from these (string) snippets. Every \type{/Page} object points also to a \type{/Resources} object (no.~1) that gives all ingredients needed to assemble the page. In our example only a \type{/Font} object (no.~4) is referenced, which in turn tells that the text is typeset in \type{/Font} \type{/Times-Roman}. The \type{/Font} object points also to a \type{/Widths} array (object no.~7) that tells for each character by how much the viewer must move forward horizontally after typesetting a glyph. More details about the font can be found in the \type{/FontDescriptor} object (no.~8); if a font file is embedded, this object points to the font program stream. But as the Times-Roman font used for our example is one of the 14 so\hyph called standard fonts that should always be present in any \PDF\ viewer and therefore need not be embedded in the \PDF\ file, it is left out here for brevity. However, when we use for instance a Computer Modern Roman font, we have to make sure that this font is later available to the \PDF\ viewer, and the best way to do this is to embed the font. It's highly recommended nowadays to embed even the standard fonts; you can't know how it looks exactly at the viewer side unless you embed every font. In this simple file we don't specify in what way the file should be opened, for instance full screen or clipped. A closer look at the page object no.~2 (\type{/Type/Page}) shows that a mediabox (\type{/MediaBox}) is part of the page description. A mediabox acts like the (high-resolution) bounding box in a \POSTSCRIPT\ file. \PDFTEX\ users can add dictionary entries to page objects with the \cs{pdfpageattr} primitive. Although in most cases macro packages will shield users from these internals, \PDFTEX\ provides access to many of the entries described here, either automatically by translating the \TEX\ data structures into \PDF\ ones, or manually by pushing entries to the catalog, page, info or self-created objects. One can for instance create an object by using \cs{pdfobj}, after which \cs{pdflastobj} returns its number. So \begin{verbatim} \pdfobj { << /Type/ExtGState /LW 2 >> } \end{verbatim} \noindent inserts an object into the \PDF\ file (it creates a ``graphics state'' object setting the line width to 2~units), and \cs{pdflastobj} now returns the number \PDFTEX\ assigned to this object. Unless objects are referenced by others, they will just end up as isolated entities, not doing any real harm but bloating the \PDF\ file. In general this rather direct way of pushing objects in the \PDF\ files by primitives like \cs{pdfobj} is not very useful, and only makes sense when implementing, say, fill\hyph in field support or annotation content reuse. We will come to that later. Of course, this is just the barest introduction to \PDF\ format. For those who want to learn more about technical \PDF\ details, the best bet is to read the \PDFReference\ (\useurl{pdfreference}). We turn now to the specifics of \PDFTEX. \chapter{Invoking \PDFTEX} \PDFTEX\ has many command line options and can use many environment variables and configuration file settings. Except for the simple and rarely-used \type{-draftmode} and \type{-output-format} options, they are all inherited from the common framework for \TeX\ engines as implemented in \WEBC\ and Kpathsea. Their documentation is available at \useurl{webc} and \useurl{kpathsea}. Two additional environment variables need more description here: first, \type{SOURCE_DATE_EPOCH} (introduced in version 1.40.17, 2016). If this is set, it must be a positive integer (with one trivial exception: if it is set but empty, that is equivalent to 0). Non-integer values cause a fatal error. The value is used as the current time in seconds since the usual Unix ``epoch'': the beginning of 1970-01-01, UTC. Thus, a value of \type{32} would result in a \type{/CreationDate} and \type{/ModDate} values of \type{19700101000032Z}. This is useful for reproducible builds of documents. (See also \cs{pdfinfoomitdate}, \cs{pdfsuppressptexinfo}, et al.) The second, related, environment variable is \type{FORCE_SOURCE_DATE}. If this is set to~\type{1}, \TEX's time-related primitives are also initialized from the value of \type{SOURCE_DATE_EPOCH}. These primitives are \cs{year}, \cs{month}, \cs{day}, and \cs{time}. If \type{SOURCE_DATE_EPOCH} is not set, setting \type{FORCE_SOURCE_DATE} has no effect. If \type{FORCE_SOURCE_DATE} is unset, set to the empty string, or set to~\type{0}, the primitives reflect the current time as usual. Any other value elicits a warning, and the current time is used. This is useful if one wants to make reproducible \PDF{}s for a set of documents without changing them in any way, \eg\ an operating system distribution with manuals that use \cs{today}. Except in such unusual circumstances, it is better not to set this, and let the \TEX\ primitives retain the meaning they have always had. In addition, if both \type{SOURCE_DATE_EPOCH} and \type{FORCE_SOURCE_DATE} are set, \cs{pdffilemoddate} returns a value in UTC, ending in \type{Z}. (The values of the environment variables are irrelevant in this case.) Finally, just to have the list of options and basic invocation at hand, here is a verbatim listing of the \type{--help} and \type{--version} output. All options can be specified with one or two dashes and unambiguously abbreviated. \myverbatiminput{incl/pdftex-help.txt} \section{Macro packages supporting \PDFTEX} Currently all mainstream macro packages offer \PDFTEX\ support, with automatic detection of \PDFTEX\ as engine. So there is normally no need to turn on \PDFTEX\ support explicitly. \begin{itemize} \item For \LATEX\ users, the \type {hyperref} package (originally written by Sebastian Rahtz and Heiko Oberdiek; now maintained by the \LaTeX\ team, has substantial support for \PDFTEX\ and provides access to most of its features. In the simplest and most common case, the user merely needs to load \type{hyperref}, and all cross\hyph references will be converted to \PDF\ hypertext links. \PDF\ output is automatically selected, compression is turned on, and the \PDF\ page size is set up correctly. Bookmarks are created to match the table of contents. \item The standard \LATEX\ packages \type{graphics}, \type{graphicx}, and \type{color} also have built-in \PDFTEX\ support, which allows use of color, text rotation, and graphics inclusion commands. \item The \CONTEXT\ MkII system by Hans Hagen has full support for \PDFTEX\ in its generalized hypertext features. The latest \CONTEXT\ supports advanced \PDF\ output, but uses a engine (\LMTX). \item \PDF\ from \GNU\ \TEXINFO\ documents can be created by running \PDFTEX\ on the \TEXINFO\ file, instead of \TEX. Alternatively, run the shell command \type{texi2pdf} instead of \type{texi2dvi}. \item For plain \TEX\ users, the \filename{miniltx.tex} file from the \type{graphics-pln} package allows loading \type{graphics}, \type{graphicx}, and \type{color}. \EPLAIN\ provides additional support for hyperlinks. \item A modification of \filename {webmac.tex}, named \filename {pdfwebmac.tex}, allows production of hyperlinked \PDF\ versions of the literate source code written in \WEB, such as \PDFTEX. \end{itemize} For \TeX\ developers: as \PDFTEX\ generates the final \PDF\ output without help of a postprocessor, macro packages that take care of these \PDF\ features have to be set up properly. Tasks include handling color, graphics, hyperlink support, threading, fonts, page imposition and manipulation. All of these \PDF\hyph specific tasks can be controlled by \PDFTEX's own primitives (a few also by a \PDFTEX\hyph specific \verb|\special{pdf: ...}|). Any other \cs{special} commands, like the ones defined for various \DVI\ postprocessors, are simply ignored by \PDFTEX\ when in \PDF\ output mode; a warning is given for non\hyph empty \cs{special} commands. When a macro package written for classical \TEX\ with \DVI\ output is to be modified for use with \PDFTEX, it is helpful to get some insight to what extent \PDFTEX\hyph specific support is needed. This info can be gathered, for instance, by outputting the various \cs{special} commands via \cs{message}. as in: \begin{verbatim} \pdfoutput=1 \let\special\message \end{verbatim} or, if this leads to confusion, \begin{verbatim} \pdfoutput=1 \def\special#1{\write16{special: #1}} \end{verbatim} and see what happens. As soon as one \type{special:} message turns up, one knows for sure that some kind of \PDFTEX\hyph specific support is needed, and often the message itself gives a indication of what is needed. \chapter{Setting up fonts} \PDFTEX\ can work with Type~1 and TrueType fonts (and to a small extent also with OpenType fonts). Font files should be available and embedded for all fonts used in the generated \PDF. It is possible to use \METAFONT\hyph generated fonts in \PDFTEX---but it is strongly recommended not to use these fonts if an equivalent is available in Type~1 or TrueType format, if only because bitmap Type~3 fonts render poorly under enlargement. \section{Map files} \label{sec.mapfile} Font map files provide the connection between \TEX\ \TFM\ font files and outline font file names. They contain also information about re\hyph encoding arrays, partial font embedding (``subsetting''), and character transformation parameters (like SlantFont and ExtendFont). Those map files were first created for \DVI\ postprocessors. But, as \PDFTEX\ in \PDF\ output mode includes all \PDF\ processing steps, it also needs to know about font mapping, and therefore reads in one or more map files. Map files are not read in when \PDFTEX\ is in \DVI\ mode. Bitmap fonts can (and normally should) be used without being listed in the map file. By default, \PDFTEX\ reads the map file \filename{pdftex.map}. In \WEBC, map files are searched for using the \type{TEXFONTMAPS} config file value and environment variable. By default, the current directory and various system directories are searched. Within the map file, each font is listed on a single line. The syntax of each line is upward\hyph compatible with \type{dvips} map files and can contain the following fields (some are optional; explanations follow): \begin{quote} \varname{tfmname psname fontflags special encodingfile fontfile} \end{quote} Here are two (real-world) examples. All the values are explained in detail in the following. {\small \begin{verbatim} cmr10 CMR10 >test.map \end{verbatim} There are a limitations with TrueType fonts in comparison with Type~1 fonts: \begin{itemize} \item The special effects SlantFont\slash ExtendFont did not work before version 1.40.0. \item To subset a TrueType font, the font must be specified as re\hyph encoded, therefore an encoding vector must be given. \item TrueType fonts used in embedded \PDF\ files are kept untouched; they are not replaced or merged with the same font used in the document, as happens with Type~1. \end{itemize} For much more about \PDFTEX\ and TrueType fonts, including many details on handling glyph names, see ``A closer look at TrueType fonts and \PDFTEX'', \titleref{TUGboat} 30:1 (2009), pp.~32\hyph 34, \useurl{thanhtruetypetub}. \chapter{\PDFTEX\ primitives} \label{sec.primitives} Here follows a description of the primitives added by \PDFTEX\ to the original \TEX\ engine (other extensions by \ETEX, \MLTEX\ and \ENCTEX\ are not described). Many of these primitives are described further in the \filename {samplepdf.tex} file in the \PDFTEX\ distribution (q.v.). If the output is \DVI, then the \PDFTEX-specific dimension parameters are not used at all. However, some \PDFTEX\ integer parameters can affect \DVI\ as well as \PDF\ output (specifically, \cs{pdfoutput} and \cs{pdfadjustspacing}). A warning to macro writers: if you define macros whose names start with \cs{pdf}, you risk name clashes with new primitives that may be introduced in future versions of \PDFTEX. General warning: many of these new primitives, for example \cs{pdfdest} and \cs{pdfoutline}, write their arguments directly to the \PDF\ output file (when producing \PDF), as \PDF\ string constants. This means that \emph{you} (or, more likely, the macros you write) must escape characters as necessary (namely \cs{}, \type{(}, and \type{)}. Otherwise, an invalid \PDF\ file may result. The \type{hyperref} and \TEXINFO\ packages have code which may serve as a starting point for implementing this, although it will certainly need to be adapted to any particular situation. \section{Document setup} \subsection{\cs{pdfoutput}} \pdftexprimitive{\Syntax{\cs{pdfoutput} \Whatever{integer}}} This parameter specifies whether the output format should be \DVI\ or \PDF. A positive value means \PDF\ output, otherwise (default 0) one gets \DVI\ output. This primitive is the only one that must be set to produce \PDF\ output (the command-line option \type{-output-format=pdf} may alternatively be used); all other primitives are optional. This parameter cannot be specified \emph{after} shipping out the first page. In other words, to get \PDF\ output, we have to set \cs{pdfoutput} before \PDFTEX\ the first page. A simple way of making macros aware of \PDFTEX\ in both \PDF\ or \DVI\ mode is: \begin{verbatim} \ifx\pdfoutput\undefined \csname newcount\endcsname\pdfoutput \fi \ifcase\pdfoutput DVI CODE \else PDF CODE \fi \end{verbatim} Using the \type{ifpdf.sty} file, which works with both \LATEX\ and plain \TeX, is a cleaner way of doing this. Originally, the simple test \verb|\ifx\pdfoutput\undefined| sufficed; but for many years, the \PDFTEX\ engine is used in distributions also for non-\PDF\ formats (\eg\ \LATEX), so \cs{pdfoutput} may be defined even when the output format is \DVI. \subsection{\cs{pdfmajorversion}, \cs{pdfminorversion}} \pdftexprimitive{\Syntax{\cs{pdfmajorversion} \Whatever{integer}}} \pdftexprimitive{\Syntax{\cs{pdfminorversion} \Whatever{integer}}} Together, these two primitives specify the \PDF\ version for generated \PDF\ output. The defaults compiled into the \PDFTEX\ program are \cs{pdfmajorversion=1} and \cs{pdfminorversion=4}, thus \PDF~1.4 is generated by default. However, distributions typically alter the engine's compiled default minor version of~4 when building formats. For example, as of 2010, \TEXLIVE\ sets \cs{pdfminorversion=5} when formats are built. This is so object compression can be enabled (see \cs{pdfobjcompress} below). This value also defines the highest \PDF\ version for included \PDF{}s to be allowed without error, by default (see \cs{pdfinclusionerrorlevel}). The values for both must be $\ge1$ but are not checked further. Furthermore, they are set independently; setting only \cs{pdfmajorversion=2} would result in \PDF~2.4 output; it's necessary to additionally set \cs{pdfminorversion}. If specified, these primitives must appear before any data is written to the generated \PDF\ file. The \cs{pdfmajorversion} primitive was introduced in \PDFTEX\ 1.40.21. \cs{pdfminorversion} was originally a synonym of the \cs{pdfoptionpdfminorversion} command, which is now obsolete. \introduced{1.30.0} \subsection{\cs{pdfcompresslevel}} \pdftexprimitive{\Syntax{\cs{pdfcompresslevel} \Whatever{integer}}} This integer parameter specifies the level of stream compression. Zero means no compression, 1~means fastest, 9~means best, $2..8$ means something in between. A value outside this range will be adjusted to the nearest meaningful value. This parameter is read each time \PDFTEX\ starts a stream. This compression applies to text, inline graphics, and embedded \PNG\ images (but only if they are un- and re-compressed during the embedding process). It is implemented by the \type{zlib} library). Setting \cs{pdfcompresslevel=0} is useful for \PDF\ stream debugging. \subsection{\cs{pdfobjcompresslevel}} \pdftexprimitive{\Syntax{\cs{pdfobjcompresslevel} \Whatever{integer}}} This integer parameter controls the compression of non-stream objects. If specified, the parameter must appear before any data is written to the \PDF\ output. \introduced{1.40.0} In the \PDF-1.4 specification, non-stream objects had to be written in the \PDF\ file as clear text, uncompressed. The \PDF-1.5 specification allows collecting non-stream objects as ``compressed objects'' into ``object stream'' objects (\type{/Type/ObjStm}, see the \PDFReference, 5th~ed., \S3.4.6). At the end of the \PDF\ file, an \type{/XRef} cross-reference stream is then written out instead of the object table. This can result in a considerably smaller \PDF\ file, particularly if lots of annotations and links are used. The writing of compressed objects is enabled by setting \cs{pdfobjcompresslevel} to a value between~1 and~3; it's disabled if~0 (default). Object compression also requires $\cs{pdfminorversion}\ge5$ (or $\cs{pdfmajorversion}\ge2$), else a warning is given and the compression is disabled. The value of \cs{pdfobjcompresslevel} is clipped to the range $0..3$, but using values outside this range is not recommended (for future extension). Values for \cs{pdfobjcompresslevel} have the following effects: \begin{itemize} \item When set to~0, no object streams are generated at all. \item When set to~1, all non-stream objects are compressed with the exception of any objects coming with embedded \PDF\ files (``paranoid'' mode, to avoid yet unknown problems), and also the \type{/Info} dictionary is not compressed for clear-text legibility. \item When set to~2, also all non-stream objects coming with embedded \PDF\ files are compressed, but the \type{/Info} dictionary is still not compressed. \item Finally, when set to~3, all non-stream objects are compressed, including the \type{/Info} dictionary (this means that the \type{/Info} can't be read as clear text any more). If object streams are to be used, currently \cs{pdfobjcompresslevel=2} is recommended, and is so specified in some distributions, including \TEXLIVE~2010 and later. \end{itemize} \emph{Compatibility caveats:} \PDF\ files generated with object streams enabled can't be read with (sufficiently old) \PDF\ viewers that don't understand \PDF-1.5 files. For widest distribution and unknown audience, don't activate object stream writing. The \PDF-1.5 standard describes also a hybrid object compression mode that gives some backward compatibility, but this is currently not implemented, as \PDF-1.5 was rather quickly adopted by modern \PDF\ viewers. Also not implemented is the optional \type{/Extends} key. \subsection{\cs{pdfdecimaldigits}} \pdftexprimitive{\Syntax{\cs{pdfdecimaldigits} \Whatever{integer}}} This integer parameter specifies the numeric accuracy of real coordinates as written to the \PDF\ file. It gives the maximal number of decimal digits after the decimal point. Valid values are in the range $0..4$. A higher value means more precise output, but also results in a larger file size and more time to display or print. In most cases the optimal value is~2. This parameter does not influence the precision of numbers used in raw \PDF\ code, like that used in \cs{pdfliteral} and annotation action specifications; also multiplication items (\eg\ scaling factors) are not affected and are always output with best precision. This parameter is read when \PDFTEX\ writes a real number to the \PDF\ output. When including huge \METAPOST\ images using \filename {supp-pdf.tex}, one can limit the accuracy to two digits with \cs{twodigitMPoutput}. \subsection{\cs{pdfhorigin}} \pdftexprimitive{\Syntax{\cs{pdfhorigin} \Whatever{dimen}}} This parameter can be used to set the horizontal offset of the output box from the top left corner of the page. A value of 1~inch corresponds to the normal \TEX\ offset. This parameter is read when \PDFTEX\ starts shipping out a page to the \PDF\ output. For all normal purposes, this parameter should always be kept at 1~true inch. If you want to shift text on the page, use \TEX's own \cs{hoffset} primitive. To avoid surprises, after global magnification has been changed by the \cs{mag} primitive, the \cs{pdfhorigin} parameter should still be 1~true inch, \eg\ by setting \cs{pdfhorigin=1 true in} after the \cs{mag} setting. Or, you can preadjust the \cs{pdfhorigin} value before typing \cs{mag}, so that its value after the \cs{mag} command ends up at 1~true inch again. \subsection{\cs{pdfvorigin}} \pdftexprimitive{\Syntax{\cs{pdfvorigin} \Whatever{dimen}}} This parameter is the vertical companion of \cs{pdfhorigin}, and the notes above regarding \cs{mag} and true dimensions apply. Also keep in mind that the \TEX\ coordinate system starts in the top left corner (downward), while \PDF\ coordinates start at the bottom left corner (upward). \subsection{\cs{pdfpagewidth}} \pdftexprimitive{\Syntax{\cs{pdfpagewidth} \Whatever{dimen}}} This dimension parameter specifies the page width of the \PDF\ output (the screen, the paper, etc.). \PDFTEX\ reads this parameter when it starts shipping out a page. If magnification has been changed by the \cs{mag} primitive, check that this parameter reflects the desired true page width. When part of the page falls off the paper or screen, it's quite possible that this parameter is set wrong. If the value is set to zero, the page width is calculated as \[ \hbox{\textit{width}}_{\hbox{\scriptsize box being shipped out}} + 2 \times (\cs{horigin} + \hbox{\cs{hoffset}}). \] It is not wise to rely on this default calculation, since box widths may vary unexpectedly. \subsection{\cs{pdfpageheight}} \pdftexprimitive{\Syntax{\cs{pdfpageheight} \Whatever{dimen}}} Similar to the previous item, this dimension parameter specifies the page height of the \PDF\ output; the notes above apply. If set to zero, the page height will be calculated analogously to the above. \section{Document info and catalog} \label{sec.docinfocatalog} \subsection{\cs{pdfomitinfodict}} \pdftexprimitive{\Syntax{\cs{pdfomitinfodict} \Whatever{integer}}} If nonzero, omit the \type{/Info} dictionary completely, as required by the \PDF\ A-4 standard. \introduced{1.40.25} \subsection{\cs{pdfinfo}} \pdftexprimitive{\Syntax{\cs{pdfinfo} \Something{general text}}} This primitive allows the user to specify information for the document info dictionary. Provided information can be extracted from the output \PDF\ by, for instance, the \type{pdfinfo} program. The \Something{general text} is a collection of key\hyph value\hyph pairs. The key names are preceded by a \type{/}, and the values, being strings, are given between parentheses. All keys, and the primitive itself, are optional. Possible keys are:\newline \type{/Title},\newline \type{/Author},\newline \type{/Subject},\newline \type{/Keywords},\newline \type{/Producer} (defaults to \hbox{\tt pdfTeX-\currentpdftex}),\newline \type{/Creator} (defaults to \type{TeX}),\newline \type{/CreationDate} (defaults to current date and time, with time zone),\newline \type{/ModDate} (same default),\newline \type{/Trapped} (defaults to \type{/False},\newline \type{/PTEX.Fullbanner} (defaults to the \cs{pdftexbanner} string, q.v.). \type{/CreationDate} and \type{/ModDate} are expressed in the form \type{D:YYYYMMDDhhmmssTZ}, where `\type{D:}' is a constant string prefix, \type{YYYY} is the year, \type{MM} is the month, \type{DD} is the day, \type{hh} is the hour, \type{mm} is the minutes, \type{ss} is the seconds, and \type{TZ} is an optional string denoting the time zone, \type{Z} for universal time. For example, this is the Unix epoch, the beginning of 1970-01-01 UTC, in this format: \type{D:19700101000000Z}. If the time zone is not UTC, it is given as \type{+HH'mm'} or \type{-HH'mm'}, indicating an offset of the given hours and minutes from UTC, with the given either later (\type{+}) or earlier (\type{-}) than UTC. For more details, see the \PDFReference, \S7.9.4. Multiple appearances of \cs{pdfinfo} are concatenated. Usually if a key is given more than once, the first appearance will be used, but this is viewer\hyph dependent. Except for standard \TeX\ macro expansion, \PDFTEX\ does not perform any further operations in the \Something{general text} provided by the user. Here is an example of using \cs{pdfinfo} to include the information not supplied by \PDFTEX: \begin{verbatim} \pdfinfo { /Title (example.pdf) /Author (Tom and Jerry) /Subject (Example) /Keywords (mouse, cat) } \end{verbatim} \subsection{\cs{pdfinfoomitdate}} \pdftexprimitive{\Syntax{\cs{pdfinfoomitdate} \Whatever{integer}}} If nonzero, omit the \type{/CreationDate} and \type{/ModDate} keys from the document info dictionary described above. This can be useful in making reproducible \PDF{}s. \introduced{1.40.17} \subsection{\cs{pdfsuppressptexinfo}} \pdftexprimitive{\Syntax{\cs{pdfsuppressptexinfo} \Whatever{integer}}} This value is treated as a bit mask, specifying which \type{PTEX.*} keys to omit from the output: \begin{smalltable} \begin{tabular}{cl} \bf value & \bf suppresses \cr 1 & \type{PTEX.Fullbanner} \cr 2 & \type{PTEX.FileName} \cr 4 & \type{PTEX.PageNumber} \cr 8 & \type{PTEX.InfoDict} \cr \end{tabular} \end{smalltable} \noindent Thus, the value \type{-1}, setting all the bits, suppresses everything. \type{PTEX.Fullbanner} is included by default in the general document info dictionary, as mentioned under \cs{pdfinfo} above. The other \type{PTEX.*} keys are included when a \PDF\ is included in the document (and not otherwise), as described in section~\ref{sec.addpdfkeys}. This conditional suppression can be useful in making reproducible \PDF{}s. \introduced{1.40.17} \subsection{\cs{pdfcatalog}} \pdftexprimitive{\Syntax{\cs{pdfcatalog} \Something{general text} \Optional{\Literal{openaction} \Something{action spec}}}} Similar to the document info section is the document catalog, where possible keys are \type{/URI}, which specifies the base \URL\ of the document, and \type {/PageMode}, which determines how the \PDF\ viewer displays the document on startup. The possibilities for the latter are given in this table: \begin{smalltable} \begin{tabular}{ll} \bf value & \bf meaning \cr \tt /UseNone & neither outline nor thumbnails visible \cr \tt /UseOutlines & outline visible \cr \tt /UseThumbs & thumbnails visible \cr \tt /FullScreen & full\hyph screen mode \cr \end{tabular} \end{smalltable} The default \type{/PageMode} setting is \type{/UseNone}. In full\hyph screen mode, there is no menu bar, window controls, nor any other window present. After the \Something{general text} of the catalog, a construct \Literal{openaction} \Something{action spec} can be given, where \Literal{openaction} is a \PDFTEX\ keyword, and \Something{action spec} specifies the action to be taken when opening the document. This \Something{action spec} is the same as for internal links; see section~\ref{sec.linking}. Several settings can be made in one \cs{pdfcatalog} call, for example: \begin{verbatim} \pdfcatalog { /PageMode /FullScreen } openaction goto page 2 {/Fit} \end{verbatim} \subsection{\cs{pdfcreationdate}} \pdftexprimitive{\Syntax{\cs{pdfcreationdate} \Whatever{expandable}}} Expands to the date string \PDFTEX\ uses in the info dictionary of the document, \eg\ for this file: {\tt\pdfcreationdate}. \introduced{1.30.0} \subsection{\cs{pdfnames}} \pdftexprimitive{\Syntax{\cs{pdfnames} \Something{general text}}} This primitive inserts the \Something{general text} to the \type{/Names} array. The text must conform to the specifications as laid down in the \PDFReference, or the document may be invalid. \subsection{\cs{pdftrailer}} \pdftexprimitive{\Syntax{\cs{pdftrailer} \Something{general text}}} This command puts its argument text verbatim into the file trailer dictionary. Example: \verb|\pdftrailer{/mytrlrkey /mytrlrval}|. \introduced{1.11a} \subsection{\cs{pdftrailerid}} \pdftexprimitive{\Syntax{\cs{pdftrailerid} \Something{general text}}} Use the \Something{general text} to seed the \type{/ID} value in the trailer, instead of the default combination of the input file name and starting time. If the argument is empty, the \type{/ID} is omitted entirely. Example: \verb|\pdftrailerid{}|. This can be useful in making reproducible \PDF{}s. \introduced{1.40.17} \section{Fonts} \subsection{\cs{pdfadjustspacing}} \pdftexprimitive{\Syntax{\cs{pdfadjustspacing} \Whatever{integer}}} This primitive controls whether font expansion happens (this operation is described in detail at \cs{pdffontexpand}). By default, \cs{pdfadjustspacing} is set to~0; then font expansion is disabled, so that the \PDFTEX\ output is identical to that from the original \TEX\ engine. Font expansion can be activated in two modes. When \cs{pdfadjustspacing} is set to~1, font expansion is applied \emph{after} \TEX's normal paragraph breaking routines have broken the paragraph into lines. In this case, line breaks are identical to standard \TEX\ behavior. When set to~2, the width changes that are the result of stretching and shrinking are taken into account \emph{while} the paragraph is broken into lines. In this case, line breaks are likely to be different from those of standard \TEX. Paragraphs may well become longer or shorter. Both alternatives require a collection of \TFM\ files that are related to the \Something{stretch} and \Something{shrink} settings for the \cs{pdffontexpand} primitive, unless this is given with the \type{autoexpand} option. \subsection{\cs{pdffontexpand}} \pdftexprimitive{\Syntax{\cs{pdffontexpand} \Something{font} \Something{stretch} \Something{shrink} \Something{step} \Optional{\Literal{autoexpand}}}} This extension to \TEX's font definitions controls a major \PDFTEX\ feature called \emph{font expansion}. To enable font expansion, \cs{pdfadjustspacing} must be set to a value greater than zero. We describe the basic process with an example: \begin{verbatim} \font\somefont=sometfm at 10pt \pdffontexpand\somefont 30 20 10 autoexpand \pdfadjustspacing=2 \end{verbatim} The \verb|30 20 10| means this: ``hey \TEX, when line breaking is going badly, you may stretch the glyphs from this font as much as 3\% or shrink them as much as 2\%.'' For practical reasons \PDFTEX\ uses discrete expansion steps, in this example, 1\%. Roughly speaking, the idea is as follows: When \TEX\ cannot break a line in the appropriate way, the unbreakable parts of the last word may stick into the margin. When \PDFTEX\ sees this, it will try to scale (shrink) the glyphs in that line using fixed steps, until the line fits. When lines are too spacey, the opposite happens: \PDFTEX\ starts scaling (stretching) the glyphs until the white space gaps are acceptable. This glyph stretching and shrinking is called \emph{font expansion}. There are two different modes for font expansion, depending on whether \type{autoexpand} is specified: \begin{enumerate} \item If the \type{autoexpand} keyword is given---this is recommended mode---only a single map entry is needed for all expanded font versions, using the name of the unexpanded \TFM\ file (\varname{tfmname} in section~\ref{sec.map-tfmname}). No expanded \varname{tfmname} versions need be mentioned (indeed, they are ignored), as \PDFTEX\ generates expanded instances of the unexpanded \TFM\ data structures and keeps them in its memory. As of \PDFTEX~1.40.0, the \type{autoexpand} work is done within the page stream by modification of the text matrix (\PDF\ operator ``\type{Tm}''), and not at the font file level, giving the advantage that it now works not only with Type~1 but also with TrueType and OpenType fonts (and even without embedding a font file; but that's not recommended). In this mode \PDFTEX\ requires only unexpanded font files. \item Second, if the \type{autoexpand} keyword is not given, setting up font expansion requires considerably more work, as there must be map entries for \TFM\ files in all required expansion values. The expanded \varname{tfmname} variants are constructed by adding the font expansion value to the \varname{tfmname} of the base font, \eg\ there must be a map entry with \varname{tfmname} \type{sometfm+10} for 1\% stretch or \type{sometfm-15} for 1.5\% shrink. This also means that for each expanded font variant a \TFM\ file with properly expanded metrics must exist. In addition to the \TFM\ file, it is necessary to provide, for each expansion value, an individually crafted font file with the expanded glyphs. Thus, this allows the absolute best possible output, controlling the glyphs for every expanded variant of the font. It is a rare document indeed for this to be worth the trouble. One technical drawback of non-\type{autoexpand} mode is that all needed individual font files need to be embedded in the \PDF\ output for each expanded font, leading to significantly larger \PDF\ files than in \type{autoexpand} mode. Another caveat for non-\type{autoexpand} mode: when \cs{pdffontexpand} is executed, \PDFTEX\ immediately loads two fonts, at the maximum stretch and shrink; in our example, \type{sometfm+30} and \type{sometfm-20}. (If they aren't available, \type{mktextfm} may be uselessly called, and then an error message issued.) This happens even if those fonts never end up being used, which is arguably undesirable, but hard to change. It is not a problem when using \type{autoexpand}. \end{enumerate} The font expansion mechanism is inspired by an optimization first introduced by Prof.~Hermann Zapf, which in itself goes back to optimizations used in the early days of typesetting: use different glyphs to optimize the grayness of a page. So, there are many, slightly different~a's, e's, etc. For practical reasons \PDFTEX\ does not use such huge glyph collections; it uses horizontal scaling instead. This is sub\hyph optimal, and possibly offensive to the design, though the expansions are tiny. It is up to the user to decide whether such slightly remastered fonts are acceptable. As an example, this document is typeset with font expansion and margin kerning activated (via the \texttt{microtype} \LaTeX\ package). \subsection{\cs{efcode}} \pdftexprimitive{\Syntax{\cs{efcode} \Something{font} \Something{8-bit number} \Whatever{integer}}} We haven't yet told the whole story. One can imagine that some glyphs are visually more sensitive to stretching or shrinking than others. Then the \cs{efcode} primitive can be used to influence the expandability of individual glyphs within a given font, as a factor applied to the expansion setting from the \cs{pdffontexpand} primitive. The syntax is similar to \cs{sfcode} (but with the \Something{font} required), and it defaults to~1000, meaning 100\% expandability. The given integer value is clipped to the range $0..1000$, corresponding to a usable expandability range of $0..100$\%. Examples: \begin{verbatim} \efcode\somefont`A=800 \efcode\somefont`O=0 \end{verbatim} Here the A in \cs{somefont} may only stretch or shrink by up to 80\% of the current expansion value for that font, and expansion for the~O is disabled. The actual expansion is still bound to the steps as defined by \cs{pdffontexpand} primitive. Changes to this table are global, \ie\ ignore \TeX's usual grouping, and apply only to the given \Something{font}. \subsection{\cs{pdfprotrudechars}} \pdftexprimitive{\Syntax{\cs{pdfprotrudechars} \Whatever{integer}}} Yet another way of optimizing paragraph breaking is to let certain characters move into the margin (`character protrusion'). Character protrusion is disabled when \cs{pdfprotrudechars=0} or negative. When \cs{pdfprotrudechars=1}, the glyphs qualified as such will make this move when applicable, without changing the line-breaking. When \cs{pdfprotrudechars=2} (or greater), character protrusion will be taken into account while considering breakpoints, so line-breaking might be changed. This qualification and the amount of shift are set by the primitives \cs{rpcode} and \cs{lpcode}. If you want to protrude an item other than a character (\eg\ an \cs{hbox}), you can do so by padding the item with an invisible zero\hyph width character for which protrusion is activated. \subsection{\cs{rpcode}, \cs{lpcode}} \pdftexprimitive{\Syntax{\cs{rpcode} \Something{font} \Something{8-bit number} \Whatever{integer}}} \pdftexprimitive{\Syntax{\cs{lpcode} \Something{font} \Something{8-bit number} \Whatever{integer}}} The amount that a character from a given font may shift into the right margin (`character protrusion') is set by the primitive \cs{rpcode}. The protrusion distance is the integer value given to \cs{rpcode}, multiplied by 0.001em from the current font. The given integer value is clipped to the range $-1000..1000$, corresponding to a usable protrusion range of $-$1em..1em. \cs{lpcode} is exactly analogous to \cs{rpcode}, but affects the amount by which characters may protrude into the left margin. Examples: \begin{verbatim} \rpcode\somefont`,=200 \rpcode\somefont`-=150 \end{verbatim} Here the comma may shift by 0.2em into the margin and the hyphen by 0.15em. All these small bits and pieces will help \PDFTEX\ to give you better paragraphs (use \cs{rpcode} judiciously; don't overdo it). Remark: old versions of \PDFTEX\ use the character width as measure. This was changed to a proportion of the em-width after \THANH\ finished his master's thesis. Changes to these tables are global, \ie\ ignore \TeX's usual grouping, and apply only to the given \Something{font}. \subsection{\cs{leftmarginkern}, \cs{rightmarginkern}} \pdftexprimitive{\Syntax{\cs{leftmarginkern} \Something{box number} \Whatever{expandable}}} \pdftexprimitive{\Syntax{\cs{rightmarginkern} \Something{box number} \Whatever{expandable}}} The \cs{leftmarginkern} \Something{box number} primitive expands to the width of the margin kern at the left side of the horizontal list stored in \cs{box} \Something{box number}. The expansion string includes the unit \type{pt}. \Eg\ when the left margin kern of \cs{box0} amounts to $-$10pt, \cs{leftmarginkern0} will expand to \type{-10pt}. The primitive \cs{rightmarginkern} works analogously for the right margin. \introducedplural{1.30.0} These are auxiliary primitives to make character protrusion more versatile. When using the \TEX\ primitives \cs{unhbox} or \cs{unhcopy}, the margin kerns at either end of the unpackaged hbox will be removed (\eg\ to avoid weird effects if several hboxes are unpackaged behind each other into the same horizontal list). These \cs{unhbox} or \cs{unhcopy} commands are often used together with \cs{vsplit} for dis- and re\hyph assembling of paragraphs, \eg\ to add line numbers. Paragraphs treated like this do not show character protrusion by default, as the margin kerns have been removed during the unpacking process. The \cs{leftmarginkern} and \cs{rightmarginkern} primitives allow access to the margin kerns and store them away before unpackaging the hbox. \Eg\ the following code snippet restores margin kerning of a horizontal list stored in \cs{box}\cs{testline}, resulting in a hbox with the original margin kerning, now inserted by ordinary kerns. \begin{verbatim} \dimen0=\leftmarginkern\testline \dimen1=\rightmarginkern\testline \hbox to\hsize{\kern\dimen0\unhcopy\testline\kern\dimen1} \end{verbatim} \subsection{\cs{letterspacefont}} \pdftexprimitive{\Syntax{\cs{letterspacefont} \Something{control sequence} \Something{font} \Something{integer}}} This primitive creates an instance of \Something{font} with the widths of all glyphs increased by \Something{integer} thousandths of an em (as defined by \Something{font}). The effect is letter spacing, but the glyphs are actually wider, as the sidebearings are increased, so a single glyph will take more space. For instance, the following creates a font \cs{lsfont} whose glyphs are all 1.2pt larger than those of \cs{normalfont}: \begin{verbatim} \font\normalfont=myfont at 12pt \letterspacefont\lsfont\normalfont 100 \end{verbatim} Negative values are allowed for \Something{integer}. Letter spacing works natively in \PDF\ mode only, unless special fonts are devised (in our example, a \type{myfont+100ls} font), as with font expansion. \subsection{\cs{pdfcopyfont}} \pdftexprimitive{\Syntax{\cs{pdfcopyfont} \Something{control sequence} \Something{font}}} This primitive defines \Something{control sequence} as a synonym for \Something{font}. \subsection{\cs{pdffakespace}} \pdftexprimitive{\Syntax{\cs{pdffakespace}}} Always insert a fake interword space in the output (PDF only; this primitive is an error in DVI mode), regardless of the values of \cs{pdfinterwordspaceon} and \cs{pdfinterwordspaceoff} (q.v.). Example: \begin{verbatim} Text with a fake interword \pdffakespace space. \end{verbatim} \introduced{1.40.15} \subsection{\cs{pdffontattr}} \pdftexprimitive{\Syntax{\cs{pdffontattr} \Something{font} \Something{general text}}} This primitive inserts the \Something{general text} into the \type{/Font} dictionary. The text must conform to the specifications as laid down in the \PDFReference, or the document may be invalid. For examples, see the \type{cmap} and \type{CJK} packages. \subsection{\cs{pdffontname}} \pdftexprimitive{\Syntax{\cs{pdffontname} \Something{font} \Whatever{expandable}}} In \PDF\ files produced by \PDFTEX\ one can recognize a font resource by the prefix~\type{/F} followed by a number, for instance \type{/F12} or \type{/F54}. For a given \TEX\ \Something{font}, this primitive expands to the number from the corresponding font resource name. \Eg\ if \type{/F12} corresponds to some \TEX\ font \cs{foo}, then \cs{pdffontname}\cs{foo} expands to the number~\type{12}. In the current implementation, when \cs{pdfuniqueresname} (see below) is set to a positive value, the \cs{pdffontname} still returns only the number from the font resource name, without the appended random string. \subsection{\cs{pdffontobjnum}} \pdftexprimitive{\Syntax{\cs{pdffontobjnum} \Something{font} \Whatever{expandable}}} This command is similar to \cs{pdffontname}, but it returns the \PDF\ object number of the font dictionary instead of the number from the font resource name. \Eg\ if the font dictionary (\type{/Type/Font}) in \PDF\ object~3 corresponds to some \TEX\ font \cs{foo}, then \cs{pdffontobjnum}\cs{foo} gives the number~\type{3}. Use of \cs{pdffontname} and \cs{pdffontobjnum} allows users full access to all the font resources used in a document. \subsection{\cs{pdffontsize}} \pdftexprimitive{\Syntax{\cs{pdffontsize} \Something{font} \Whatever{expandable}}} This primitive expands to the font size of the given font, with unit \type{pt}. \Eg\ when using the plain \TeX\ macro package, the call \cs{pdffontsize}\cs{tenrm} expands to \type{10.0pt}. \subsection{\cs{pdfgentounicode}} \pdftexprimitive{\Syntax{\cs{pdfgentounicode} \Whatever{integer}}} By default, \PDFTEX\ does not include a \type{/ToUnicode} resource when including fonts in the output. Such a resource (also called a CMap resource) maps glyph names to Unicode characters in a \PDF\ file. Lacking such a resource, it is the \PDF\ reader which determines how and whether searching in the \PDF\ file works. In practice, searching for basic \ASCII\ characters generally works, but searching for anything beyond those, including ligatures such as `fi', is likely to fail. If \cs{pdfgentounicode} is set to \type{1} when the job ends, the \type{/ToUnicode} resource will be included in the output, with mappings for Type~1 fonts used---unless \cs{pdfnobuiltintounicode} (q.v.) is set for a given font. The mapping is created as follows: for each glyph in the font, look for its \type{ToUnicode} value in a global hash table. By default that global hash table is empty, in which case \PDFTEX\ merely emits a warning. Entries are added to the table using the command \cs{pdfglyphtounicode}, described next. \subsection{\cs{pdfglyphtounicode}} \pdftexprimitive{\Syntax{\cs{pdfglyphtounicode} \Something{general text} \Something{general text}}} The first argument is the name of a glyph, the second is a string of Unicode numeric values denoting characters, separated by spaces. For instance: \begin{verbatim} \pdfgentounicode=1 \pdfglyphtounicode{ff}{0066 0066} \end{verbatim} \noindent maps the \type{ff} ligature to a pair of \type{f} characters (whose code is \type{U+0066}, that is, \ASCII\ \type{0x66}). Once a single \cs{pdfglyphtounicode} definition is made, whether it is used or not, another feature comes into play: all glyph names of the form \type{uniXXXX} or \type{uXXXX} are mapped to the natural \type{U+XXXX}. Many fonts use this style of naming. In addition, the \type{glyphtounicode.tex} file (distributed with \PDFTEX\ and other software) contains thousands of such definitions, covering most common glyph names. So, for practical purposes, one would probably want: \begin{verbatim} \input glyphtounicode \pdfgentounicode=1 \end{verbatim} \noindent (\LATEX\ users could load the \type{cmap} package to achieve the same effect.) By default, these glyph name-to-unicode mappings are global. Thus, \begin{verbatim} \pdfglyphtounicode{abc}{1234} \end{verbatim} \noindent would map the glyph named \type{abc} to \type{U+1234} for every font. However, it's possible to make a mapping for a single font using a \type{tfm:} prefix: \begin{verbatim} \pdfglyphtounicode{tfm:foo/abc}{5678} \end{verbatim} \noindent means that for the font \type{foo.tfm}, only, the glyph \type{abc} is mapped to \type{U+5678}. Glyph names sometimes contain a dot, as in \type{somechar.sc}. \PDFTEX\ simply strips the dot and everything after it before looking up the name, so in this case it would look for \type{somechar} (even if \type{somechar.sc} exists in the mappings, it will not be used). This behavior could be made smarter if there is a demand for it. \subsection{\cs{pdfincludechars}} \pdftexprimitive{\Syntax{\cs{pdfincludechars} \Something{font} \Something{general text} \Whatever{expandable}}} This command causes \PDFTEX\ to treat the characters in \Something{general text} as if they were used with \Something{font}, which means that the corresponding glyphs will be embedded into the font resources in the \PDF\ output. Nothing is appended to the list being built. \subsection{\cs{pdfinterwordspaceon}, \cs{pdfinterwordspaceoff}, \cs{pdfspacefont}} \pdftexprimitive{\Syntax{\cs{pdfinterwordspaceon}}} \pdftexprimitive{\Syntax{\cs{pdfinterwordspaceoff}}} \pdftexprimitive{\Syntax{\cs{pdfspacefont} \Something{general text}}} The first two commands insert corresponding whatsit nodes which turn on/off generation of faked interword spaces in the \PDF\ output (they cause errors in \DVI\ output). This allows for better reflowing of text on the fly by \PDF\ readers, better extraction of textual content, and is required by \PDF/A. It does not affect the normal \TeX\ justification with glue of the typeset output. This works roughly as follows: with \cs{pdfinterwordspaceon}, \PDFTEX\ will guess when an interword space should be inserted, based on movement within some limits in horizontal direction. When found, \PDFTEX\ inserts a true space character into the \PDF\ page description for the viewers, and adjusts the next movement so that the next character will be in the expected position, according to normal \TEX\ behavior. Where does that ``true space character'' come from? There are two possibilities. \begin{itemize} \item If the current font has a real space character, it is used. \PDFTEX\ considers a font to have such a space character if 1)~the font has an encoding vector (\type{.enc} file) specified in its map entry, and 2)~the encoding has a glyph named \type{space} (that is, the PostScript name \type{/space}) at slot 32. For example, the font \type{texnansi-lmr10} uses the encoding file \type{lm-texnansi.enc}, which has such an entry. \item If the current font does not have such a space character (this is the case for most traditional \TEX\ fonts, such as \type{cmr10} and \type{ec-lmr10}), \PDFTEX\ will use the space character from a special fallback font named (by default) \type{pdftexspace}[\type{.tfm}]. \PDFTEX\ automatically defines a map entry for this font which looks like this: \begin{verbatim} \pdfmapline{=pdftexspace PdfTeX-Space 0$, \type{/ProcSet} is always omitted; if it is $<0$, \type{/ProcSet} is always included. For information about what \type{/ProcSet} is, see the \PDFReference\ or other documentation. \cs{ProcSet} was considered obsolete as of \PDF~1.4, but conforming writers should continue to output it. It was formally deprecated in \PDF~2.0.\introduced{1.40.25} \subsection{\cs{pdfpageref}} \pdftexprimitive{\Syntax{\cs{pdfpageref} \Something{page number} \Whatever{expandable}}} This primitive expands to the number of the page object that contains the dictionary for page \Something{page number}. If the page \Something{page number} does not exist, a warning will be issued, a fresh unused \PDF\ object will be generated, and \cs{pdfpageref} will expand to that object number. \Eg\ if the dictionary for page~5 of the \TEX\ document is contained in \PDF\ object no.~18, \cs{pdfpageref5} expands to the number 18. \subsection{\cs{pdfpageresources}} \pdftexprimitive{\Syntax{\cs{pdfpageresources} \Whatever{tokens}}} These tokens are added to the resource dictionary for all pages, before the font, XObject, and ProcSet resources. For example: \begin{verbatim} \pdfpageresources{/MyPageResourceAttribute /MyValue} \end{verbatim} \section{Form XObjects} The next three primitives support a \PDF\ feature called ``object reuse'' in \PDFTEX. The idea is first to create a `form XObject'. The content of this object corresponds to the content of a \TEX\ box; it can contain pictures and references to other form XObjects as well. After creation, the form XObject can be used multiple times by simply referring to its object number. This feature can be useful for large documents with many similar elements, to reduce the duplication of identical objects. These commands behave similarly to \cs{pdfobj}, \cs{pdfrefobj} and \cs{pdflastobj} (described in the previous section), but instead of taking raw \PDF\ code, they handle text typeset by \TEX. \subsection{\cs{pdfxform}} \pdftexprimitive{\Syntax{\cs{pdfxform} \Optional{\Something{attr spec}} \Optional{\Something{resources spec}} \Something{box number} \Modelist{h, v, m}}} This command creates a form XObject corresponding to the contents of the box \Something{box number}. The box can contain other raw objects, form XObjects, or images as well. However, it cannot contain annotations because those are laid out on a separate layer, are positioned absolutely, and have dedicated housekeeping. \cs{pdfxform} makes box \Something{box number} void, as \cs{box} does. When \Something{attr spec} is given, the text will be written as additional attribute into the form XObject dictionary. The \Something{resources spec} is similar, but the text will be added to the resources dictionary of the form XObject. The text given by \Something{attr spec} or \Something{resources spec} is written before other entries of the form dictionary and\slash or the resources dictionary and takes priority over later ones. \subsection{\cs{pdfrefxform}} \pdftexprimitive{\Syntax{\cs{pdfrefxform} \Something{object number} \Modelist{h, v, m}}} The form XObject is kept in memory and will be written to the \PDF\ output only when its object number is referred to by \cs{pdfrefxform} or when \cs{pdfxform} is preceded by \cs{immediate}. Nothing is appended to the list being built. The number of the most recently created form XObject is accessible via \cs{pdflastxform}. When issued, \cs{pdfrefxform} appends a whatsit node to the list being built. When the whatsit node is searched at shipout time, \PDFTEX\ will write the form \Something{object number} to the \PDF\ output if it is not written yet. \subsection{\cs{pdflastxform}} \pdftexprimitive{\Syntax{\cs{pdflastxform} \Whatever{read-only integer}}} The object number of the most recently created form XObject is accessible via \cs{pdflastxform}. As said, this feature can be used for reusing information. This mechanism also plays a role in typesetting fill\hyph in forms. Such widgets sometimes have visuals that show up on user request, but are hidden otherwise. \subsection{\cs{pdfxformname}} \pdftexprimitive{\Syntax{\cs{pdfxformname} \Something{object number} \Whatever{expandable}}} In \PDF\ files produced by \PDFTEX\ one can recognize a form XObject by the prefix~\type{/Fm} followed by a number, for instance \type{/Fm2}. For a given form XObject number, this primitive expands to the number in the corresponding form XObject name. \Eg\ if \type{/Fm2} corresponds to some form XObject with object number 7, the \cs{pdfxformname7} expands to the number~\type{2}. \introduced{1.30.0} \section{Graphics inclusion} \PDF\ provides a mechanism for embedding graphic and textual objects: form XObjects. In \PDFTEX\ this mechanism is accessed by means of \cs{pdfxform}, \cs{pdflastxform} and \cs{pdfrefxform} (described in the previous section). A special kind of XObjects are bitmap graphics and for manipulating them similar commands are provided. \subsection{\cs{pdfximage}} \pdftexprimitive{\Syntax{\cs{pdfximage} \Optional{\Something{rule spec}} \Optional{\Something{attr spec}} \Optional{\Something{page spec}} \Optional{\Something{pdf box spec}} \Optional{\Something{colorspace spec}} \Next \Something{general text} \Modelist{h, v, m} }} This command creates an image object from the filename given in \Something{general text}, \ie\ enclosed in braces. The image is kept in memory when this command is executed. It is written to the \PDF\ output only when either 1)~the \cs{pdfximage} command is preceded by \cs{immediate}, or 2)~its number is referred to by \cs{pdfrefximage} (q.v.). The dimensions of the image can be controlled via \Something{rule spec}. The default values are zero for depth and ``running'' for height and width. If all of them are given, the image will be scaled to fit the specified values. If some (but not all) are given, the rest will be set to a value corresponding to the remaining ones so as to make the image size to yield the same proportion of $width : (height+depth)$ as the original image size, where depth is treated as zero. If none are given then the image will have its natural size. An image inserted at its natural size often has a resolution of \type{\pdfimageresolution} (see below) given in dots per inch in the output file, but some images may contain data specifying the image resolution, and in such a case the image will be scaled accordingly. The dimensions of an image can be accessed by putting the \cs{pdfrefximage} command in a box and checking the box's dimensions: \begin{verbatim} \setbox0=\hbox{\pdfximage{somefile.png}\pdfrefximage\pdflastximage} \end{verbatim} Now we can use \cs{wd0} and \cs{ht0} to find the natural size of the image as determined by \PDFTEX. The image type is specified by the extension of the given file name: \type{.png} stands for \PNG\ image, \type{.jpg} (or \type{.jpeg}) for \JPEG, \type{.jbig2} (preferred, but \type{.jb2} works also) for \JBIGTWO, and \type{.pdf} for \PDF\ file. But once \PDFTEX\ has opened the file, it checks the file type first by checking the magic number at the file start, which gets precedence over the file name extension. This gives a certain degree of fault tolerance, if the file name extension is stated wrongly. Similarly to \cs{pdfxform}, the optional text given by \Something{attr spec} will be written as additional attributes of the image before other keys of the image dictionary. When using this, be aware that slightly different types of \PDF\ objects are created while including \PNG, \JPEG, or \JBIG2\ bitmaps and \PDF\ images. While working with \PDF\ or \JBIG2\ images, \Something{page spec} allows to specify which page of the document is to be included; the \Something{page spec} is irrelevant for the other two image formats. Commonly, \Something{page spec} will be {\tt page \Something{page number}}, but for \PDF\ images, you may also reference a so-called ``named destination'' with {\tt named \{\Something{destination name}\}}. (Such named destinations are created by \type{hyperref}, for example.) Starting with \PDFTEX\ 1.11, in the \PDF\ image case, you may also choose which page box of the image is to be treated as a final bounding box. If \Something{pdf box spec} is present, it overrides the default behavior specified by the \cs{pdfpagebox} parameter, and is overridden by the (obsolete) \cs{pdfforcepagebox} parameter. This option is irrelevant for non\hyph \PDF\ inclusions. Starting with \PDFTEX\ 1.21, \cs{pdfximage} supports a \type{colorspace} keyword followed by an object number (user\hyph defined colorspace for the image being included). This feature works for \JPEG\ images only. \PNG\ images use \RGB\ palettes, \JBIG2 bi-tonal, and \PDF\ images have self\hyph contained color space information. \subsection{\cs{pdfrefximage}} \pdftexprimitive{\Syntax{\cs{pdfrefximage} \Something{object number}}} \cs{pdfrefximage} appends a whatsit node to the list being built. When the whatsit node is searched at shipout time, \PDFTEX\ will write the image with number \Something{object number} to the \PDF\ output if it has not been written yet. \subsection{\cs{pdflastximage}} \pdftexprimitive{\Syntax{\cs{pdflastximage} \Whatever{read-only integer}}} \type{\pdflastximage} is the number of the most recently created XObject image. \subsection{\cs{pdfximagebbox}} \pdftexprimitive{\Syntax{\cs{pdfximagebbox} \Something{integer} \Something{integer} \Whatever{expandable}}} The dimensions of the bounding box of a \PDF\ image loaded with \cs{pdfximage} are stored in a table. This primitive returns those dimensions as follows: \begin{verbatim} \pdfximage{afile.pdf} \pdfximagebbox\pdflastximage 1 % Returns lower-left x \pdfximagebbox\pdflastximage 2 % Returns lower-left y \pdfximagebbox\pdflastximage 3 % Returns upper-right x \pdfximagebbox\pdflastximage 4 % Returns upper-right y \end{verbatim} \subsection{\cs{pdflastximagecolordepth}} \pdftexprimitive{\Syntax{\cs{pdflastximagecolordepth} \Whatever{read-only integer}}} The color depth (\type{1} for 1-bit images, \type{2} for 2-bit images, and so on) of the last image accessed with \cs{pdfximage}. \subsection{\cs{pdflastximagepages}} \pdftexprimitive{\Syntax{\cs{pdflastximagepages} \Whatever{read-only integer}}} This read-only register returns the highest page number from a file previously accessed via the \cs{pdfximage} command. This is useful only for \PDF\ files; it always returns \type{1} for \PNG, \JPEG, or \JBIGTWO\ files. \subsection{\cs{pdfimageresolution}} \pdftexprimitive{\Syntax{\cs{pdfimageresolution} \Whatever{integer}}} The integer \cs{pdfimageresolution} parameter (given in dots per inch, dpi) is a last resort value, used only for bitmap (\JPEG, \PNG, \JBIGTWO) images, but not for \PDF{}s. The priorities are as follows: Often one image dimension (\type{width} or \type{height}) is stated explicitly in the \TEX\ file. Then the image is properly scaled so that the aspect ratio is kept. If both image dimensions are given, the image will be stretched accordingly, whereby the aspect ratio might get distorted. If no image dimensions are given in the \TEX\ file, the image size will be calculated from its width and height in pixels, using the $x$ and $y$ resolution values normally contained in the image file. If one of these resolution values is missing or weird (either $<$~0 or $>$~65535), the \cs{pdfimageresolution} value will be used for both $x$ and $y$ resolution when calculating the image size. And if the \cs{pdfimageresolution} is zero, a final fallback resolution of 72dpi is used. The \cs{pdfimageresolution} is read when \PDFTEX\ creates an image via \cs{pdfximage}. The given value is clipped to the range $0..65535$ (dpi). Currently this parameter is used particularly for calculating the dimensions of \JPEG\ images with \EXIF\ information (unless at least one dimension is stated explicitly); the resolution values included in \EXIF\ data are ignored. \subsection{\cs{pdfpagebox}} \pdftexprimitive{\Syntax{\cs{pdfpagebox} \Whatever{integer}}} When \PDF\ files are included, the command \cs{pdfximage} allows the selection of which \PDF\ page box to use in the optional \Something{image attr spec}. If the option isn't present, the page box defaults to the value of \cs{pdfpagebox} as follows: (1)~media box, (2)~crop box, (3)~bleed box, (4)~trim box, and (5)~artbox. \subsection{\cs{pdfforcepagebox}} \pdftexprimitive{\Syntax{\cs{pdfforcepagebox} \Whatever{integer}}} %- It is now possible to specify the pdf page box to use when including pdfs. % The syntax has been extended: % \pdfximage [] (h, v, m) % --> [] [] [] [] % --> mediabox|cropbox|bleedbox|trimbox|artbox % The default is cropbox (which defaults to mediabox), to be compatible with % previous versions of pdfTeX. % The page box specified at \pdfximage can be globally overridden with the % config parameter always_use_pdfpagebox and the command % \pdfoptionalwaysusepdfpagebox , where 0 is the default (use % whatever is specified with \pdfximage), 1 is media, 2 is crop, 3 is % bleed, 4 is trim and 5 is artbox. This can only be set once per % object, i.e., the value used at \pdfximage is remembered. % See the PDF Reference for an explanation of the boxes. The integer primitive \cs{pdfforcepagebox} allows globally overriding the choice of the page box used with \cs{pdfximage}. It takes the same values as \cs{pdfpagebox}. The command is available starting with \PDFTEX\ 1.30.0, as a shortened synonym of obsolete \cs{pdfoptionalwaysusepdfpagebox} instruction, but is itself now considered obsolete---a combination of \cs{pdfpagebox} and \Something{image attr spec} is better. \subsection{\cs{pdfinclusionerrorlevel}} \pdftexprimitive{\Syntax{\cs{pdfinclusionerrorlevel} \Whatever{integer}}} This controls the behavior of \PDFTEX\ when a \PDF\ file is included which has a newer \PDF\ version than the one specified by \cs{pdfmajorversion} and \cs{pdfminorversion}. If \cs{pdfinclusionerrorlevel} is set to~0 (the default), \PDFTEX\ gives a warning; if 1, \PDFTEX\ raises an error; if negative, no diagnostic at all is given. It was originally a shortened synonym of \cs{pdfoptionpdfinclusionerrorlevel}, which is now obsolete. \introduced{1.30.0} \subsection{\cs{pdfimagehicolor}} \pdftexprimitive{\Syntax{\cs{pdfimagehicolor} \Whatever{integer}}} This parameter, when set to~1, enables embedding of \PNG\ images with 16~bit wide color channels at their full color resolution. This embedding mode is defined only from \PDF\ version~1.5 onwards, so the \cs{pdfimagehicolor} functionality is automatically disabled in \PDFTEX\ if $\cs{pdfminorversion}<5$ and $\cs{pdfmajorversion}=1$; in this case, each 16~bit color channel is reduced to a width of 8~bits by stripping the lower 8~bits before embedding. The same stripping happens when \cs{pdfimagehicolor} is set to~0. If $\cs{pdfmajorversion}=1$ and $\cs{pdfminorversion}\ge5$, or $\cs{pdfmajorversion}\ge2$, the default value of \cs{pdfimagehicolor} is~1. If specified, the parameter must appear before any data is written to the \PDF\ output. \introduced{1.30.0} \subsection{\cs{pdfimageapplygamma}} \pdftexprimitive{\Syntax{\cs{pdfimageapplygamma} \Whatever{integer}}} This primitive, when set to~1, enables gamma correction while embedding \PNG\ images, taking the value of the primitive \cs{pdfgamma} as well as the gamma value embedded in the \PNG\ image into account. When \cs{pdfimageapplygamma} is set to~0, no gamma correction is performed. If specified, the parameter must appear before any data is written to the \PDF\ output. \introduced{1.30.0} \subsection{\cs{pdfgamma}} \pdftexprimitive{\Syntax{\cs{pdfgamma} \Whatever{integer}}} This primitive defines the `device gamma' for \PDFTEX. Values are in promilles (same as \cs{mag}). The default value of this primitive is~1000, defining a device gamma value of~1.0. If \cs{pdfimageapplygamma} is set to~1, then whenever a \PNG\ image is included, \PDFTEX\ applies a gamma correction. This correction is based on the value of the \cs{pdfgamma} primitive and the `assumed device gamma' that is derived from the value embedded in the actual image. If no embedded value can be found in the \PNG\ image, then the value of \cs{pdfimagegamma} is used instead. If specified, the parameter must appear before any data is written to the \PDF\ output. \introduced{1.30.0} \subsection{\cs{pdfimagegamma}} \pdftexprimitive{\Syntax{\cs{pdfimagegamma} \Whatever{integer}}} This primitive gives a default `assumed gamma' value for \PNG\ images. Values are in promilles (same as for \cs{pdfamma}). The default value of this primitive is~2200, implying an assumed gamma value of~2.2. When \PDFTEX\ is applying gamma corrections, images that do not have an embedded `assumed gamma' value are assumed to have been created for a device with a gamma of 2.2. Experiments show that this default setting is correct for a large number of images; however, if your images come out too dark, you probably want to set \cs{pdfimagegamma} to a lower value, like~1000. If specified, the parameter must appear before any data is written to the \PDF\ output. \introduced{1.30.0} \subsection{\cs{pdfpxdimen}} \pdftexprimitive{\Syntax{\cs{pdfpxdimen} \Whatever{dimen}}} While working with bitmap graphics or typesetting electronic documents, it might be convenient to base dimensions on pixels rather than \TeX's standard units like \type{pt} or \type{em}. For this purpose, \PDFTEX\ provides an extra unit named \type{px} that takes the dimension given to the \cs{pdfpxdimen} primitive. For example, to make the unit \type{px} corresponding to 96dpi pixel density (then $\mbox{1px}=72/96$bp), one can do the following calculation: \begin{verbatim} \pdfpxdimen=1in % 1 dpi \divide\pdfpxdimen by 96 % 96 dpi \hsize=1200px \end{verbatim} Then \cs{hsize} amounts to 1200~pixels in 96dpi, which is exactly 903.375pt, which \TeX\ rounds to 903.36914pt. The default value of \cs{pdfpxdimen} is 1.00001bp (for historical reasons), corresponding to a pixel density of (a few sp off from) 72dpi. The \LUATEX\ default is 1bp (also a few sp off from 72dpi), so to get precisely the same behavior in \PDFTEX\ and \LUATEX, set \cs{pdfpxdimen}\texttt{=1bp}. This primitive is completely independent of the \cs{pdfimageresolution} and \cs{pdfpkresolution} parameters. \introduced{1.30.0} It used to be an integer register that gave the dimension 1px as number of scaled points, defaulting to 65536 (1px equal to $\mbox{65536sp}=1$pt). Starting with \PDFTEX\ 1.40.0, \cs{pdfpxdimen} is now a real dimension parameter. \subsection{\cs{pdfinclusioncopyfonts}} \pdftexprimitive{\Syntax{\cs{pdfinclusioncopyfonts} \Whatever{integer}}} If positive, this parameter tells \PDFTEX\ to include fonts that are embedded in a \PDF\ file loaded with \cs{pdfximage}, even if those fonts are also available on disk. Bigger files might be created, but included \PDF\ files are sure to be embedded with the correct fonts; indeed, the fonts on disk might be different from the embedded ones, and glyphs might be missing. \subsection{\cs{pdfsuppresswarningpagegroup}} \pdftexprimitive{\Syntax{\cs{pdfsuppresswarningpagegroup} \Whatever{integer}}} Ordinarily, \PDFTEX\ gives a warning when more than one included \PDF\ file has a so-called ``page group object'' (\type{/Group}), because only one can ``win''---that is, be propagated to the page level. Usually the page groups are identical, but when they are not, the result is unpredictable. It would be ideal if \PDFTEX\ in fact detected whether the page groups were the same and only gave the warning in the problematic case; unfortunately, this is not easy (a patch would be welcome). Nevertheless, often one observes that there is no actual problem. Then seeing the warnings on every run is just noise, and can be suppressed by setting this parameter to a positive number. \introduced{1.40.15} \section{Annotations} \PDF\ 1.4 provides four basic kinds of annotations: \begin{itemize} \item hyperlinks, general navigation \item text clips (notes) \item movies \item sound fragments \end{itemize} The first type differs from the other three in that there is a designated area involved on which one can click, or when moved over some action occurs. \PDFTEX\ is able to calculate this area, as we will see later. All annotations can be supported using the next two general annotation primitives. \subsection{\cs{pdfannot}} \pdftexprimitive{\Syntax{\cs{pdfannot} \Something{annot type spec} \Modelist{h, v, m}}} This command appends a whatsit node corresponding to an annotation to the list being built. The dimensions of the annotation can be controlled via the \Something{rule spec}. The default values are ``running'' for all width, height and depth. When an annotation is written out, running dimensions will take the corresponding values from the box containing the whatsit node representing the annotation. The \Something{general text} is inserted as raw \PDF\ code to the contents of annotation. The annotation is written only if the corresponding whatsit node is searched at shipout time. \subsection{\cs{pdflastannot}} \pdftexprimitive{\Syntax{\cs{pdflastannot} \Whatever{read-only integer}}} This primitive returns the object number of the last annotation created by \cs{pdfannot}. These two primitives allow users to create any annotation that cannot be created by \cs{pdfstartlink} (see below). \section{Destinations and links} \label{sec.linking} The first type of annotation (hyperlinks and navigation), mentioned above, is implemented by three primitives. The first one is used to define a specific location as being referred to. This location is tied to the page, not the exact location on the page. The main reason for this is that \PDFTEX\ maintains a dedicated list of these annotations, and more when optimized, for the sole purpose of speed. \subsection{\cs{pdfdest}} \pdftexprimitive{\Syntax{\cs{pdfdest} \Something{dest spec} \Modelist{h, v, m}}} This primitive appends a whatsit node which establishes a destination for links and bookmark outlines; the link is identified by either a number or a symbolic name, and the way the viewer is to display the page is specified in \Something{dest spec}, which must be one of those mentioned in the table below. \begin{smalltable} \begin{tabular}{ll} \bf keyword & \bf meaning \cr \tt fit & fit the page in the window \cr \tt fith & fit the width of the page \cr \tt fitv & fit the height of the page \cr \tt fitb & fit the \type{BoundingBox} of the page \cr \tt fitbh & fit the width of \type{BoundingBox} of the page \cr \tt fitbv & fit the height of \type{BoundingBox} of the page \cr \tt fitr \Something{rule spec} & fit according to \Something{rule spec} \cr \tt xyz \Optional{zoom \Something{number}} & goto the current position (see below) \cr \tt struct \Optional{\Something{number}} & create structure destination (see below) \cr \end{tabular} \end{smalltable} The \Literal{xyz} keyword can optionally be followed by \Literal{zoom} \Something{integer} to provide a fixed zoom\hyph in. The \Something{integer} is used like the \TEX\ magnification value, \ie\ 1000 is the normal page view. When \Literal{zoom} \Something{integer} is given, the zoom factor changes to 0.001 of the \Something{integer} value, otherwise the current zoom factor is kept unchanged. If \Literal{struct} \Something{number} is used, a structure destination is created instead of a regular destination, referring to the structure element defined in object \Something{number}. Structure destinations use a separate namespace and therefore may have the same identifiers as a regular destination. The destination is written out only if the corresponding whatsit node is searched at shipout time. \subsection{\cs{pdfstartlink}} \pdftexprimitive{\Syntax{\cs{pdfstartlink} \Optional{\Something{rule spec}} \Optional{\Something{attr spec}} \Something{action spec} \Modelist{h, m} }} This primitive is used along with \cs{pdfendlink} and appends a whatsit node corresponding to the start of a hyperlink. The whatsit node representing the end of the hyperlink is created by \cs{pdfendlink}. The dimensions of the link are handled in the similar way as in \cs{pdfannot}. Both \cs{pdfstartlink} and \cs{pdfendlink} must be in the same level of box nesting. A hyperlink with running width can be multi\hyph line or even multi\hyph page, in which case all horizontal boxes with the same nesting level as the boxes containing \cs{pdfstartlink} and \cs{pdfendlink} will be treated as part of the hyperlink. The hyperlink is written out only if the corresponding whatsit node is searched at shipout time. Additional attributes, which are explained in great detail in the \PDFReference, can be given via \Something{attr spec}. Typically, the attributes specify the color and thickness of any border around the link. As a simple example, \verb|/C [0.9 0 0] /Border [0 0 2]| specifies a color (in \RGB) of dark red, and a border thickness of 2~points. While all graphics and text in a \PDF\ document have relative positions, annotations have internally hard\hyph coded absolute positions. Again this is for the sake of speed optimization. The main disadvantage is that these annotations do \emph{not} obey transformations issued by \type{\pdfliteral}. The \Something{action spec} specifies the action that should be performed when the hyperlink is activated, one of (see the syntax rules): \Literal{user} \Something{user-action spec}, \Literal{goto} \Something{goto-action spec}, \Literal{thread} \Something{thread-action spec}. \begin{itemize} \item A \Something{user-action spec} (\type{user{...}}) performs a user\hyph defined action. Examples: a \URL, as in \verb|/S /URI /URI (https://tug.org/)|; a named action, as in \verb|/S /Named /N /NextPage|. \item A \Something{goto-action spec} (\type{goto...}) performs various goto actions, and is by far the most complex action. Here is a copy of the syntax, for easier reference: \Syntax{\small \Something{goto-action spec} \Means % \Optional{\Something{goto-action struct spec}} \Something{numid} \Or \Next \Optional{\Something{file spec}} \Optional{\Something{goto-action struct spec}} \Something{nameid} \Or \Next \Optional{\Something{file spec}} \Optional{\Something{goto-action struct spec}} \Optional{\Something{page spec}} \Something{general text} \Or \Next \Something{file spec} \Optional{\Something{goto-action struct spec}} \Something{nameid} \Something{newwindow spec} \Or \Next \Something{file spec} \Optional{\Something{goto-action struct spec}} \Optional{\Something{page spec}} \Something{general text} \Something{newwindow spec} } \begin{itemize} \item The \Something {numid} and \Something{nameid} alternatives specify a destination identifier. \item The \Something{page spec} alternative specifies a page number for the destination. In this case, the zoom factor is given by \Something{general text}. \item The default behavior of \Something{newwindow spec} depends on the browser setting. \item A destination in another \PDF\ file can be given with \Something{file spec}, in which case \Something{newwindow spec} specifies whether the file should be opened in a new window. A \Something{file spec} can be either a \type{(}\Something{string}\type{)} or a \type{<<}\Something{dictionary}\type{>>}. \item If a \Something{goto-action spec} contains a \Something{goto-action struct spec}, then a structure destination is referenced in addition to the regular destination. The form with \Something{general text} is used if and only if a \Something{file spec} is present; then the \Something{general text} should expand to a literal \PDF\ dictionary describing a structure destination. Otherwise, the \Something{numid} or \Something{nameid} directly after the \Literal{struct} keyword identify a destination which must have been created with \cs{pdfdest} \Literal{struct}\unskip. \end{itemize} \item A \Something{thread-action spec} (\type{thread...}) performs article thread reading. The thread identifier, \Something{numid} or \Something{nameid}, is similar to the destination identifier. A thread in another \PDF\ file can be referenced by specifying a \Something{file spec}. \end{itemize} \subsection{\cs{pdfendlink}} \pdftexprimitive{\Syntax{\cs{pdfendlink} \Modelist{h, m}}} This primitive ends a link started with \cs{pdfstartlink}. All text between \cs{pdfstartlink} and \cs{pdfendlink} will be treated as part of this link. \PDFTEX\ may break the result across lines (or pages), in which case it will make several links with the same content. \subsection{\cs{pdflastlink}} \pdftexprimitive{\Syntax{\cs{pdflastlink} \Whatever{read-only integer}}} This primitive returns the object number of the last link created by \cs{pdfstartlink} (analogous to \cs{pdflastannot}). \introduced{1.40.0} \subsection{\cs{pdflinkmargin}} \pdftexprimitive{\Syntax{\cs{pdflinkmargin} \Whatever{dimen}}} This dimension parameter specifies the margin of the box representing a hyperlink and is read when a page containing hyperlinks is shipped out. \subsection{\cs{pdfdestmargin}} \pdftexprimitive{\Syntax{\cs{pdfdestmargin} \Whatever{dimen}}} Margin added to the dimensions of the rectangle around the destinations. \subsection{\cs{pdfsuppresswarningdupdest}} \pdftexprimitive{\Syntax{\cs{pdfsuppresswarningdupdest} \Whatever{integer}}} Ordinarily, \PDFTEX\ gives a warning when the same destination is used more than once. However, due to problematic macro packages, sometimes a document may end up producing the warning through no fault of its own, and fixing the macros may not be feasible. Then seeing the warnings on every run is just noise, and can be suppressed by setting this parameter to a positive number. \introduced{1.40.13} \subsection{\cs{pdfrunninglinkon}, \cs{pdfrunninglinkoff}} \pdftexprimitive{\Syntax{\cs{pdfrunninglinkoff}}} \pdftexprimitive{\Syntax{\cs{pdfrunninglinkon}}} These commands create corresponding whatsit nodes which turn off/on generation of running links. Their typical usage is to turn off generation of running links in the header or footer of a page. Generation of running links is on when the shipout routine begins. The generation of running links works roughly like this: \PDFTEX\ keeps a stack of links created by \cs{pdfstartlink}, called \type{pdf_link_stack}. When writing out an hbox to \PDF, \PDFTEX\ checks if the nesting level of the box is the same as the nesting level of the top entry in \type{pdf_link_stack}; if so, that box would become a link, too. The whatsit nodes created by the above primitives turn off/on a flag which controls if a hbox being shipped can become a link, in addition to the previous condition. Thus, the commands must be inserted before the hbox in question. For example: \begin{verbatim} % (1) good: \hbox{\pdfrunninglinkoff \hbox{text that would become a link otherwise} } % (2) bad: \hbox{\pdfrunninglinkoff text that would become a link otherwise} % too late; \pdfrunninglinkoff must be inserted before the box \end{verbatim} \section{Bookmarks} \subsection{\cs{pdfoutline}} \pdftexprimitive{\Syntax{\cs{pdfoutline} \Optional{\Something{attr spec}} \Something{action spec} \Optional{\Literal{count} \Something{integer}} \Something{general text} \Modelist{h, v, m} }} This primitive creates an outline (or bookmark) entry. The first parameter, \Something{attr spec}, specifies the action to be taken, and is the same as that allowed for \type{\pdfstartlink}. The \Literal{count} \Something{integer} specifies the number of direct subentries under this entry; specify~0 or omit the clause if this entry has no subentries. If the \Something{integer} is negative, then all subentries will be closed and the absolute value of this number specifies the number of subentries. The \Something{general text} is what will be shown in the outline window. The outline is written to the \PDF\ output immediately. \section{Article threads} \subsection{\cs{pdfthread}} \pdftexprimitive{\Syntax{\cs{pdfthread} \Optional{\Something{rule spec}} \Optional{\Something{attr spec}} \Something{id spec} \Modelist{h, v, m}}} Defines a bead within an article thread. Thread beads with same identifiers (spread across the document) will be joined together. \subsection{\cs{pdfstartthread}} \pdftexprimitive{\Syntax{\cs{pdftstartthread} \Optional{\Something{rule spec}} \Optional{\Something{attr spec}} \Something{id spec} \Modelist{v, m}}} This uses the same syntax as \cs{pdfthread}, apart that it must be followed by a \cs{pdfendthread}. \cs{pdfstartthread} and the corresponding \cs{pdfendthread} must end up in vboxes with the same nesting level; all vboxes between them will be added into the thread. In the output routine, if there are other newly created boxes which have the same nesting level as the vbox(es) containing \cs{pdfstartthread} and \cs{pdfendthread}, they will be also added into the thread, which is probably not what you want. To avoid such undesired behavior, it's often enough to wrap boxes that shouldn't belong to the thread by a box to change their box nesting level. \subsection{\cs{pdfendthread}} \pdftexprimitive{\Syntax{\cs{pdfendthread} \Modelist{v, m}}} This ends an article thread started before by \cs{pdfstartthread}. \subsection{\cs{pdfthreadmargin}} \pdftexprimitive{\Syntax{\cs{pdfthreadmargin} \Whatever{dimen}}} Specifies a margin to be added to the dimensions of a bead within an article thread. \section{Literals and specials} \subsection{\cs{pdfliteral}} \pdftexprimitive{\Syntax{\cs{pdfliteral} \Optional{\Literal{shipout}} \Optional{\Literal{direct} \Or \Literal{page}} \Something{general text} \Modelist{h, v, m} }} Analogous to \cs{special} in the original \TEX, this command inserts raw \PDF\ code into the output, appending a whatsit node to the list being built. This allows support of color and text transformation, among other things. By default, \Something{general text} is expanded immediately, when the whatsit node is created, as with \cs{special}. Starting with \PDFTEX\ 1.40.25, the optional keyword \Literal{shipout} can be used to delay expansion of \Something{general text} until the whatsit node is shipped out, as with non-\cs{immediate} \cs{write}. Normally, \PDFTEX\ ends a text section in the \PDF\ output and sets the transformation matrix to the current location on the page before inserting \Something{general text}; this can be turned off by giving the optional keyword \Literal{direct}. Starting with version 1.30.0, \PDFTEX\ supports the keyword \type{page} in addition to \type{direct}. Both modify the default behavior of \cs{pdfliteral}, avoiding translation of the coordinate space before inserting the literal code. The difference is that the \type{page} keyword instructs \PDFTEX\ to close a \type{BT...ET} text block before inserting anything. This means that the literal code inserted refers to the \PDF\ origin (lower-left corner of the page) and can be safely enclosed with \type{q...Q}. In contrast, using \type{q...Q} operators inside \cs{pdfliteral} with the \type{direct} keyword will produce corrupted \PDF\ output, as the \PDF\ standard doesn't allow doing anything like this within a \type{BT...ET} block. \subsection{\cs{special}} \pdftexprimitive{\Syntax{\cs{special} \Lbrace\Literal{pdf:} \Something{text}\Rbrace}} This is equivalent to \Syntax{\cs{pdfliteral} \Lbrace\Something{text}\Rbrace}. \subsection{\cs{special\ direct}} \pdftexprimitive{\Syntax{\cs{special} \Lbrace\Literal{pdf:direct:} \Something{text}\Rbrace}} This is equivalent to \Syntax{\cs{pdfliteral} \Literal{direct} \Lbrace\Something{text}\Rbrace}. \subsection{\cs{special\ page}} \pdftexprimitive{\Syntax{\cs{special} \Lbrace\Literal{pdf:page:} \Something{text}\Rbrace}} This is equivalent to \Syntax{\cs{pdfliteral} \Literal{page} \Lbrace\Something{text}\Rbrace}. \subsection{\cs{special\ shipout}} \pdftexprimitive{\Syntax{\cs{special} \Optional{\Literal{shipout}} \Lbrace\Something{text}\Rbrace}} Starting with version 1.40.25, \PDFTEX\ extends the \cs{special} primitive to support the optional keyword \Literal{shipout}. This delays expansion of \Something{text} until the page is shipped out, as with non-\cs{immediate} \cs{write}. By default, \Something{text} is expanded immediately. \Syntax{\cs{special} \Literal{shipout} \Lbrace \Literal{pdf:} \Something{text}\Rbrace} is equivalent to \Syntax{\cs{pdfliteral} \Literal{shipout} \Lbrace \Something{text}\Rbrace}. \section{Strings} \subsection{\cs{pdfescapestring}} \pdftexprimitive{\Syntax{\cs{pdfescapestring} \Something{general text} \Whatever{expandable}}} Starting with version 1.30.0, \PDFTEX\ provides a mechanism for converting a general text into \PDF\ string. Many characters that may be needed inside such a text (notably parentheses), have a special meaning inside a \PDF\ string object and thus, can't be used literally. The primitive replaces each special \PDF\ character by its literal representation by inserting a backslash before that character. Some characters (\eg\ space) are also converted into 3\hyph digit octal number. For example, \verb|\pdfescapestring{Text (1)}| will be expanded to \type{Text\040\(1\)}. This ensures a literal interpretation of the text by the \PDF\ viewer. \introduced{1.30.0} \subsection{\cs{pdfescapename}} \pdftexprimitive{\Syntax{\cs{pdfescapename} \Something{general text} \Whatever{expandable}}} Analogous to \cs{pdfescapestring}, \cs{pdfescapename} replaces each special \PDF\ character inside the general text by its hexadecimal representation preceded by a \type{#} character. This ensures the proper interpretation of the text if used as a \PDF\ name object. In example, \verb|Text (1)| will be replaced by \type{Text#20#281#29}. \introduced{1.30.0} \subsection{\cs{pdfescapehex}} \pdftexprimitive{\Syntax{\cs{pdfescapehex} \Something{general text} \Whatever{expandable}}} This command converts each character of \Something{general text} into its hexadecimal representation. Each character of the argument becomes a pair of hexadecimal digits. \introduced{1.30.0} \subsection{\cs{pdfunescapehex}} \pdftexprimitive{\Syntax{\cs{pdfunescapehex} \Something{general text} \Whatever{expandable}}} This command treats each character pair of \Something{general text} as a hexadecimal number and returns the corresponding characters. \introduced{1.30.0} \subsection{\cs{pdfstrcmp}} \pdftexprimitive{\Syntax{\cs{pdfstrcmp} \Something{general text} \Something{general text} \Whatever{expandable}}} This command compares two strings and expands to \type{0} if the strings are equal, to \type{-1} if the first string ranks before the second, and to \type{1} otherwise. \introduced{1.30.0} \subsection{\cs{pdfmatch}} \pdftexprimitive{\Syntax{\cs{pdfmatch} \Optional{\Literal{icase}} \Optional{\Literal{subcount} \Something{integer}} \Something{general text} \Something{general text} \Whatever{expandable}}} This command implements pattern matching, using \POSIX\ extended regular expression syntax. The first \Something{general text} is a pattern and the second is a string. The command expands to \type{-1} if the pattern is invalid, to \type{0} if no match is found, and to \type{1} if a match is found. With the \type{icase} option, the matching is case-insensitive. The \type{subcount} option sets the size of the table storing the found (sub)patterns; its default is 10. \introduced{1.30.0} \subsection{\cs{pdflastmatch}} \pdftexprimitive{\Syntax{\cs{pdflastmatch} \Something{integer} \Whatever{expandable}}} The matches found with \cs{pdfmatch} are stored in a table. This command returns the entry for match \Something{integer}, in the format \Something{position}\type{->}\Something{string}; \Something{position} is the position of the match (starting at zero) or \type{-1} if no match was found, and \Something{string} is the matched substring. Entry~0 contains the match as a whole; the subsequent entries contain submatches corresponding to the subpatterns, up to \type{subcount-1}. If \Something{integer} is less than zero, an error is given. For instance: \begin{verbatim} \pdfmatch subcount 3 {ab(cd)*ef(gh)(ij)}{abefghij} \pdflastmatch0 % "0->abefghij" \pdflastmatch1 % "-1->" \pdflastmatch2 % "4->gh" \pdflastmatch3 % "-1->" \end{verbatim} Entry~1 is empty because no match was found for \type{cd}, and entry~3 is empty because it exceeds the table's size as set by \type{subcount}. \introduced{1.30.0} \section{Numbers} \subsection{\cs{ifpdfabsnum}, \cs{ifpdfabsdim}} \pdftexprimitive{\Syntax{\cs{ifpdfabsnum} \Whatever{expandable}}} \pdftexprimitive{\Syntax{\cs{ifpdfabsdim} \Whatever{expandable}}} This conditional works like the standard \cs{ifnum} (resp.\ \cs{ifdim}), except that it compares absolute values of numbers (dimensions). Although it seems to be a trivial shortcut for a couple of regular \cs{if} tests, as primitives they are considerably simpler and faster to use. \introduced{1.40.0} \subsection{\cs{pdfnormaldeviate}} \pdftexprimitive{\Syntax{\cs{pdfnormaldeviate} \Whatever{expandable}}} Generate a normally-distributed random integer with a mean of~0 and standard deviation 65536. That is, about 68\% of the time, the result will be between $-65536$ and $65536$ (one standard deviation away from the mean). About 95\% of results will be within two standard deviations, and 99.7\% within three. This primitive expands to a list of tokens. \introduced{1.30.0} \subsection{\cs{pdfuniformdeviate}} \pdftexprimitive{\Syntax{\cs{pdfuniformdeviate} \Something{number} \Whatever{expandable}}} Generate a uniformly-distributed random integer value between 0 (inclusive) and \Something{number} (exclusive). This primitive expands to a list of tokens. \introduced{1.30.0} \subsection{\cs{pdfrandomseed}} \pdftexprimitive{\Syntax{\cs{pdfrandomseed} \Whatever{read-only integer}}} You can use \cs{pdfrandomseed} to query the current seed value, so you can \eg\ write the value to the log file. The initial value of the seed is derived from the system time, and is not more than 1\,000\,999\,999 (this ensures that the value can be used with commands like \cs{count}). \introduced{1.30.0} \subsection{\cs{pdfsetrandomseed}} \pdftexprimitive{\Syntax{\cs{pdfsetrandomseed} \Something{number}}} Set the random seed (\cs{pdfrandomseed}) to a specific value, allowing you to replay sequences of semi-randoms at a later moment. \introduced{1.30.0} \section{Timekeeping} \subsection{\cs{pdfelapsedtime}} \pdftexprimitive{\Syntax{\cs{pdfelapsedtime} \Whatever{read-only integer}}} Return a number that represents the time elapsed from the moment of the start of the run. The elapsed time is returned in `scaled seconds', meaning seconds divided by 65536, \eg\ \PDFTEX\ has run for \the\pdfelapsedtime\ `scaled seconds' when this paragraph was typeset. The primitive will never return a value greater than the highest number available in \TeX: if the time exceeds 32767 seconds, the constant value $2^{31}-1$ will be returned. \introduced{1.30.0} \subsection{\cs{pdfresettimer}} \pdftexprimitive{\Syntax{\cs{pdfresettimer}}} Reset the internal timer so that \cs{pdfelapsedtime} starts returning the scaled seconds from~0 again. \introduced{1.30.0} \section{Files} \subsection{\cs{pdffiledump}} \pdftexprimitive{\Syntax{\cs{pdffiledump} \Optional{\Literal{offset} \Something{integer}} \Optional{\Literal{length} \Something{integer}} \Something{general text} \Whatever{expandable}}} Expands to the dump of the first \type{length} \Something{integer} bytes of the file specified by \Something{general text}, in uppercase hexadecimal format (same as \cs{pdfescapehex}), starting at offset \type{offset}~\Something{number}, or the beginning of the file if \type{offset} is zero or not specified. If \type{length} is not given, the default is zero, so expands to nothing. Both \Something{integer}s must be $\ge0$. For example, the first ten bytes of the source of this manual are {\tt\pdffiledump length 10 {./\jobname}}. \introduced{1.30.0} \subsection{\cs{pdffilesize}} \pdftexprimitive{\Syntax{\cs{pdffilesize} \Something{general text} \Whatever{expandable}}} Expands to the size of the file specified by \Something{general text}, \eg\ {\tt \pdffilesize{./\jobname}} for the source of this manual. \introduced{1.30.0} \subsection{\cs{pdfmdfivesum\ file}} \pdftexprimitive{\Syntax{\cs{pdfmdfivesum} \Optional{\Literal{file}} \Something{general text} \Whatever{expandable}}} If the keyword \type{file} is given, expands to the \MDFIVE\ checksum of the \Something{general text} in uppercase hexadecimal format (same as \cs{pdfescapehex}). Without \type{file}, expands to the \MDFIVE\ of the \Something{general text} taken as a string. For example, the \MDFIVE\ of the source for this manual is {\tt \pdfmdfivesum file{./\jobname}}. \introduced{1.30.0} \subsection{\cs{pdffilemoddate}} \pdftexprimitive{\Syntax{\cs{pdffilemoddate} \Something{general text} \Whatever{expandable}}} Expands to the modification date of file specified by \Something{general text}, in the same format as for \cs{pdfcreationdate}, \eg\ {\tt \pdffilemoddate {./\jobname}} for the source of this manual. As of \PDFTEX\ 1.40.20, if the \type{SOURCE_DATE_EPOCH} and \type{FORCE_SOURCE_DATE} environment variables are both set, \cs{pdffilemoddate} returns a value in UTC, ending in \type{Z}. \introduced{1.30.0} \subsection{\cs{input}} \pdftexprimitive{\Syntax{\cs{input} \Something{general text} \Whatever{expandable}}} As of \TEXLIVE\ 2020, the \cs{input} primitive in all \TEX\ engines, including \PDFTEX, now also accepts a group-delimited filename argument, as a system-dependent extension, as in \cs{input}\Lbrace\type{foo.tex}\Rbrace. The standard usage of \cs{input} with a space/token-delimited filename is completely unchanged. This group-delimited argument was previously implemented in Lua\TEX; now it is available in all engines. \ASCII\ double quote characters (\type{"}) are removed from the filename, but it is otherwise left unchanged after tokenization. This extension is unlike most others in that it affects a primitive in standard \TEX\ (\cs{input}), rather than being related to a new primitive, command line option, etc. This is allowed because additional methods of recognizing filenames are explicitly mentioned in \type{tex.web} as acceptable system-dependent extensions. Incidentally, this does not directly affect \LATEX's \cs{input} command, as that is a macro redefinition of the standard \cs{input} primitive. \introduced{1.40.21} \section{Color stack} \PDFTEX\ 1.40.0 introduced color stack support; more generally, stacks of any graphic state. \subsection{\cs{pdfcolorstackinit}} \pdftexprimitive{\Syntax{\cs{pdfcolorstackinit} \Optional{\Literal{page}} \Optional{\Literal{direct}} \Something{general text} \Whatever{expandable}}} The primitive initializes a new graphic stack and returns its number. The optional \Literal{page} keyword instructs \PDFTEX\ to restore the graphic at the beginning of every new page. The optional \Literal{direct} keyword has the same effect as for \cs{pdfliteral} primitive. \introduced{1.40.0} \subsection{\cs{pdfcolorstack}} \pdftexprimitive{\Syntax{\cs{pdfcolorstack} \Something{stack number} \Something{stack action} \Something{general text}}} \noindent \Something{stack action} \Means \Literal{set} \Or \Literal{push} \Or \Literal{pop} \Or \Literal{current} The command operates on the stack of a given \Something{stack number}. If \Something{stack action} is: \begin{description} \item[\Literal{push}:] the new value in \Something{general text} is inserted at the top of the graphic stack and becomes the current stack value. \item[\Literal{pop}:] the top value is removed from the stack and the new top value becomes the top of the stack. \item[\Literal{set}:] the current value is replaced with \Something{general text}, without changing the stack size. \item[\Literal{current}:] the current stack value is return without modifying the stack. \end{description} \introduced{1.40.0} \section{Transformations} Since the content of \cs{pdfliteral} is not interpreted, any transformation inserted directly into the content stream, as well as saving and restoring the current transformation matrix, will be unnoticed by \PDFTEX's positioning mechanism. As a consequence, links and other annotations (that are formed in \PDF\ on a different layer then the page content) are not affected by such user-defined transformations. \PDFTEX\ 1.40.0 solves this problem with three new primitives. \subsection{\cs{pdfsetmatrix}} \pdftexprimitive{\Syntax{\cs{pdfsetmatrix}}} Affine transformations are normally expressed with six numbers. The first four (no unit) values defining scaling, rotating and skewing, plus two extra dimensions for translation (shifting). Since translation is handled by \TeX\ itself, the \cs{pdfsetmatrix} primitive expects as an argument a string containing just the first four numbers of the transformation separated by a space and assumes the two position coordinates to be~0. For example, \verb|\pdfsetmatrix{0.87 -0.5 0.5 0.87}| rotates the current space by 30 degrees, inserting \verb|0.87 -0.5 0.5 0.87 0 0 cm| into the output. \introduced{1.40.0} \subsection{\cs{pdfsave}} \pdftexprimitive{\Syntax{\cs{pdfsave}}} The command saves the current transformation by inserting the \type{q} operator into the content stream. \introduced{1.40.0} \subsection{\cs{pdfrestore}} \pdftexprimitive{\Syntax{\cs{pdfrestore}}} The command restores previously saved transformation by inserting the \type{Q} operator into the content stream. \cs{pdfsave} and \cs{pdfrestore} pairs should always be properly nested and should start and end at the same group level. \introduced{1.40.0} \section{Macro programming} Expansion-related commands. \subsection{\cs{expanded}} \pdftexprimitive{\Syntax{\cs{expanded} \Something{tokens} \Whatever{expandable}}} Expands \Something{tokens} in exactly the same way as \cs{message}. In contrast to \cs{edef}, macro parameter characters do not need to be doubled. \cs{protected} macros (q.v.) are not expanded. \introduced{1.40.20} \subsection{\cs{ifincsname}} \pdftexprimitive{\Syntax{\cs{ifincsname} \Whatever{expandable}}} This conditional is true if evaluated inside \cs{csname} \ldots \cs{endcsname}, and false otherwise. \subsection{\cs{ifpdfprimitive}} \pdftexprimitive{\Syntax{\cs{ifpdfprimitive} \Something{control sequence} \Whatever{expandable}}} This conditional checks if the following control sequence has its primitive meaning. If so, \cs{ifpdfprimitive} is true. In any other case (redefined, made \cs{undefined}, has never been primitive), it is false. \introduced{1.40.0} \subsection{\cs{partokencontext}} \pdftexprimitive{\Syntax{\cs{partokencontext} \Something{number}}} The \varname{par-token} (\ie\ the token with the name given by \cs{partokenname}, which is \cs{par} by default; see \cs{partokenname}, next) is inserted into the input stream in different places, according to the \cs{partokencontext} value. This value can be: 0 (the default): {\em par-token} is inserted at empty lines (more exactly, when a token category~5 is seen in state~$N$, reading a line); before \cs{end}, \cs{vskip}, \cs{hrule}, \cs{unvbox}, and \cs{halign}, if \TeX{} is in horizontal mode when those commands are seen; and in various error recovery situations. These are the standard cases in \TeX. 1: {\em par-token} is inserted as above, and also at the end of \cs{vbox}, \cs{vtop}, and \cs{vcenter}, if \TeX{} is in horizontal mode at the time. 2: {\em par-token} is inserted as above, and also at the end of \cs{insert}, \cs{vadjust}, \cs{output}, \cs{noalign}, and items of \cs{valign}, again if \TeX{} is in horizontal mode at the time. With the default \cs{partokencontext=0}, \TEX\ behaves in its normal way: the situations in cases 1 and 2 are processed by a direct call of {\em end-paragraph} routine, with no emitted {\em par-token}. If \cs{partokencontext=1} then \TeX\ inserts the {\em par-token} in additional cases: when vertical boxes are completed but horizontal mode is not finished. Since vboxes are not uncommonly inserted directly by users, with horizontal mode material, this allows macro programmers to control all such boxes being finished by a {\em par-token}. An example: \begin{verbatim} \partokenname\_mypar \partokencontext=1 \def\_mypar{Hi there!\endgraf} \vbox{Vbox text.} \end{verbatim} This will output ``Hi there!''\ after ``Vbox text.''. Finally, with \cs{partokencontext=2}, all cases where classical \TeX\ uses the direct {\em end-paragraph} routine are changed to emit the {\em par-token} instead. In contrast to case~1, these commands are rarely invoked directly by users with horizontal mode material. The setting of the register \cs{partokencontext} is local. \introduced{1.40.24} \subsection{\cs{partokenname}} \pdftexprimitive{\Syntax{\cs{partokenname} \Something{control sequence}}} \TeX{} internally inserts a control sequence, named \cs{par} by default, into the input stream at empty lines, the end of vboxes, and various other places (see \cs{partokencontext}, above). Let's call this control sequence the \varname{par-token}. Executing \cs{partokenname}\Something{control sequence} changes the name of the \varname{par-token} from \cs{par} to the given \Something{control sequence}. The setting performed by \cs{partokenname} is global. This makes it possible to release the name \cs{par} to the ``user's name space'', \ie\ after using \cs{partokenname}, users can define and use \cs{par} as they wish without changing the behavior of anything internal to \TeX{}. For example: \begin{verbatim} \catcode`\_=11 \partokenname\_mypar % use \_mypar at user level \let\_mypar=\par % make \_mypar equivalent to built-in \par % \def\par{some random text} % redefine \par % Hello world. Goodbye. \end \end{verbatim} This will not output ``\verb|some random text|'' (the definition of \cs{par}), due to the \cs{partokenname} setting. By default, the meaning of the \varname{par-token} is to end a paragraph (also named as \cs{endgraf} in the plain \TEX\ format). It can be changed as usual with, for example, \cs{def}. Naturally, it is the control sequence name given to \cs{partokenname} that must be redefined. Continuing the previous example (prior to the \cs{end}): \begin{verbatim} \def\_mypar{Hi there!\endgraf} Paragraph one. Paragraph two.\let\_mypar=\endgraf \end{verbatim} This will output ``Hi there!''\ after ``Paragraph one.'', before ending the paragraph. Another behavior of the \varname{par-token} built into \TEX\ is that macros not defined as \cs{long} cause the error ``runaway argument'' if the \varname{par-token} is scanned as a parameter. After \cs{setpartokenname}, it will be the new control sequence name that triggers this error, not \cs{par}. For instance (still continuing the same example): \begin{verbatim} \def\amac#1{} \amac{long test, no error: \par} \amac{long test, gives error: \_mypar} \end{verbatim} \introduced{1.40.24} \subsection{\cs{pdfprimitive}} \pdftexprimitive{\Syntax{\cs{pdfprimitive} \Something{control sequence}}} This command executes the primitive meaning of the following control sequence, regardless of whether the control sequence has been redefined or made undefined. If the primitive was expandable, \cs{pdfprimitive} expands also. On the other hand, if the following control sequence never was a primitive, nothing happens and no error is raised. (In some versions of \PDFTEX\ prior to 1.40.19, an error was wrongly given.) \introduced{1.40.0} \section{Typesetting} \subsection{\cs{pdfinsertht}} \pdftexprimitive{\Syntax{\cs{pdfinsertht} \Something{integer} \Whatever{expandable}}} If \Something{integer} is the number of an insertion class, this command returns the current height of the corresponding box. For instance, the following returns \type{12pt} in plain \TEX: \begin{verbatim} Abc\footnote*{Whatever.}\par \pdfinsertht\footins \end{verbatim} \subsection{\cs{pdflastxpos}, \cs{pdflastypos}} \pdftexprimitive{\Syntax{\cs{pdflastxpos} \Whatever{read-only integer}}} \pdftexprimitive{\Syntax{\cs{pdflastypos} \Whatever{read-only integer}}} This primitive returns an integer number representing the absolute $x$ resp.\ $y$ coordinate of the last point marked by \cs{pdfsavepos}. The unit is scaled points (sp). \subsection{\cs{pdfsavepos}} \pdftexprimitive{\Syntax{\cs{pdfsavepos} \Modelist{h, v, m}}} This primitive marks the current absolute ($x,y$) position on the media, with the reference point in the lower left corner. It is active only during page shipout, when the final page is assembled. The position coordinates can then be retrieved by the \cs{pdflastxpos} and \cs{pdflastypos} primitives, and \eg\ written out to some auxiliary file. The coordinates can be used only after the current \cs{shipout} has been finalized, therefore normally two \PDFTEX\ runs are required to utilize these primitives. Starting with \PDFTEX\ 1.40.0, this mechanism can also be used while running in \DVI\ mode. \subsection{\cs{quitvmode}} \pdftexprimitive{\Syntax{\cs{quitvmode}}} The primitive instructs \PDFTEX\ to quit vertical mode and start typesetting a paragraph. Thus, \cs{quitvmode} has the same basic effect as the \cs{leavevmode} macro from \type{plain.tex}. However, \cs{leavevmode} expands the \cs{everypar} token list, which may or may not be desired. \cs{quitvmode} does not expand \cs{everypar}. \introduced{1.21a} \subsection{\cs{vadjust}} \pdftexprimitive{\Syntax{\cs{vadjust} \Optional{\Something{pre spec}} \Something{filler} \Lbrace \Something{vertical mode material} \Rbrace \Modelist{h, m} }} In \PDFTEX, the \cs{vadjust} primitive supports an additional optional qualifier \Something{pre spec}, which is simply the string \type{pre}, to the original \TEX. If no \type{pre} is given, \cs{vadjust} behaves exactly as the original (see {\em The \TEX book}, p.~281): it appends an adjustment item created from \Something{vertical mode material} to the current list {\em after} the line in which \cs{vadjust} appears. In contrast, with the qualifier \type{pre}, the adjustment item is put {\em before} the line in which \cs{vadjust pre} appears. \section{Tracing} \subsection{\cs{showstream}} \pdftexprimitive{\Syntax{\cs{showstream} \Whatever{integer}}} If this primitive parameter has a value corresponding to an open output stream (which has been opened with \cs{openout}), then any \cs{show}, \cs{showthe}, \cs{showbox} or \cs{showlists} commands do not write output to the terminal, but instead write only to the referenced output stream, as if they were written with \cs{immediate}\cs{write}. For example: \begin{verbatim} \newwrite\myoutstream \immediate\openout\myoutstream="infofile" \showstream=\myoutstream % From now on, \show... commands are redirected to "infofile.tex". \show\TeX % \showstream=-1 % -1 is never a open file and therefore restores % normal \show... behavior. \immediate\closeout\myoutstream \end{verbatim} This example would not generate any special output to the terminal or log file (except for any logging done by \cs{newwrite}. It writes this text to \type{infofile.tex}, including the initial blank line, since that is what \cs{show} does: \begin{verbatim} > \TeX=macro: ->T\kern -.1667em\lower .5ex\hbox {E}\kern -.125emX \end{verbatim} This primitive is available, with identical behavior, in all \TeX\ engines except the original \TEX\ and \eTeX, where \cs{showstream} remains undefined. \introduced{1.40.24} \subsection{\cs{tracinglostchars}} \pdftexprimitive{\Syntax{\cs{tracinglostchars} \Whatever{integer}}} This primitive parameter has always been part of \TEX, and its operation with values $\le2$ is unchanged. In addition, if its value is $\ge3$, then ``Missing character'' reports become full errors (ordinarily they are only logged), and the character code is reported in hex. For example: \begin{verbatim} \tracinglostchars=3 \font\x=logo10 \x \char99 \end \end{verbatim} \noindent will result in this error message: \begin{verbatim} ! Missing character: There is no c ("63) in font logo10. \end{verbatim} \noindent (The \type{logo10} font only defines the capital letters used in the \METAFONT\ and \METAPOST\ logos, so there is no lowercase.) This new behavior is essentially the same in all \TeX\ engines except the original \TEX\ and \eTeX, where the behavior of \cs{tracinglostchars} remains unchanged. \introduced{1.40.22} \subsection{\cs{tracingstacklevels}} \pdftexprimitive{\Syntax{\cs{tracingstacklevels} \Whatever{integer}}} If this primitive parameter is $>0$, and $\cs{tracingmacros}>0$, macro expansion logging is truncated at the specified depth. Also, and more importantly, each relevant log line is given a prefix beginning with \type{~}, either followed by a \type{.} character for each expansion level or another \type{~} if the expansion was truncated. For example: \begin{verbatim} \tracingmacros=1 % so macro expansion is logged at all \tracingstacklevels=2 % cut off at level 2 \def\a#1{\relax} % argument to show parameter logging is affected too \def\b#1{\a{#1}} \b1 \end{verbatim} \noindent logs the following: \begin{verbatim} ~.\b #1->\a {#1} #1<-1 ~~\a \end{verbatim} Thus, the expansion of \cs{b} is logged normally, with the addition of the \type{~.} prefix. The expansion of \cs{a} is truncated (level~2), hence neither the parameters nor body expansion are shown. Furthermore, an \cs{input} file counts as an expansion level, and the input filename is logged. So, if we add this to our example above: \begin{verbatim} \input anotherfile \end{verbatim} \noindent where \type{anotherfile.tex} simply contains \cs{b2}, the log will get: \begin{verbatim} ~.INPUT anotherfile.tex ~~\b ~~\a \end{verbatim} Now the \cs{b} expansion is not logged either, since the expansion depth is higher than the \cs{tracingstacklevels} value. The intended use of \cs{tracingstacklevels} is not so much to truncate logging as to indicate expansion levels for detailed debugging. Thus normally it would be set to a large number (\cs{maxdimen}), so that everything is fully logged, with the addition of the expansion level indication with the number of dots in the prefix. This primitive is available, with identical behavior, in all \TeX\ engines except the original \TEX\ and \eTeX, where \cs{tracingstacklevels} remains undefined. \introduced{1.40.22} \section{\PDFTEX\ execution environment} \subsection{\cs{pdfdraftmode}} \pdftexprimitive{\Syntax{\cs{pdfdraftmode} \Whatever{integer}}} When set to 1 (or set by the command-line switch \type{-draftmode}), \PDFTEX\ doesn't write the output \PDF\ file and doesn't read any images, but does do everything else (including writing auxiliary files), thus speeding up compilations when you know you need an extra run but don't care about the output, \eg\ just to get cross-references started. If specified, the parameter must appear before any data is written to the \PDF\ output. \introduced{1.40.0} \subsection{\cs{pdfshellescape}} \pdftexprimitive{\Syntax{\cs{pdfshellescape} \Whatever{read-only integer}}} This primitive is~1 if \cs{write18} is enabled, 2 if its operation is restricted to known-safe programs, and 0 otherwise. \introduced{1.30.0} \subsection{\cs{pdftexbanner}} \pdftexprimitive{\Syntax{\cs{pdftexbanner} \Whatever{expandable}}} Returns the \PDFTEX\ banner message, \eg\ for the version used here: {\tt \pdftexbanner}. \introduced{1.20a} \subsection{\cs{pdftexrevision}} \pdftexprimitive{\Syntax{\cs{pdftexrevision} \Whatever{expandable}}} Returns the revision number of \PDFTEX, \eg\ for \PDFTEX\ version \currentpdftex\ (used to produce this document), it returns the number {\tt \pdftexrevision}. \subsection{\cs{pdftexversion}} \pdftexprimitive{\Syntax{\cs{pdftexversion} \Whatever{read-only integer}}} Returns the version of \PDFTEX\ multiplied by 100, excluding the tertiary version number, \eg\ for \PDFTEX\ version \currentpdftex\ (used to produce this document), it returns {\tt \number\pdftexversion}. \chapter{Graphics} \PDFTEX\ supports inclusion of pictures in \PNG, \JPEG, \JBIGTWO, and \PDF\ format; a few differences between these are discussed below. The most common technique with \TEX, the inclusion of \EPS\ figures, is not (and cannot be) supported; \PDF\ inclusion is usually the easiest replacement. \EPS\ files can be converted to \PDF\ by \GHOSTSCRIPT, Adobe Distiller or other \POSTSCRIPT\hyph to\hyph \PDF\ converters. \PDFTEX\ allows inserting selected pages from \PDF\ files, carry their own fonts, graphics, and pixel images, into a document. The cover page of this manual is an example of such an insert, being a one-page document generated by \PDFTEX. By default \PDFTEX\ takes the BoundingBox of a \PDF\ file from its CropBox if available, otherwise from its MediaBox. This can be influenced by the \Something{pdf box spec} option to the \cs{pdfximage} primitive, or by setting the \cs{pdfpagebox} or \cs{pdfforcepagebox} primitives to a value corresponding to the desired box type. To get the right BoundingBox from a \EPS\ file, before converting to \PDF, it is necessary to transform the \EPS\ file so that the start point is at the (0,0)~coordinate and the page size is set exactly corresponding to the BoundingBox. A \PERL\ script (\EPSTOPDF) for this purpose and the \PSTOPDF\ program that comes with \GHOSTSCRIPT\ can do a similar job. (Concerning this conversion, they can handle complete directories, remove some garbage from files, takes precautions against duplicate conversion, etc.) The lossless compressing \PNG\ format is useful for embedding crisp pixel graphics (\eg\ line scans, screenshots). As of \PDFTEX\ 1.30.0, the alpha channel of \PNG\ images is processed if available; this allows embedding of images with simple transparency. The \PNG\ format does not support the CMYK color model, which is sometimes required for print media (this often can be replaced by four component \JPEG\ in high quality or lossless compression mode). Photos in \PNG\ format have rather poor compression; here the \JPEG\ format is preferable. Embedding \PNG\ images in the general case requires \PDFTEX\ to uncompress the pixel array and to re\hyph compress it to the \PDF\ requirements; this may take a noticeable amount of time. Since \PDFTEX\ 1.30.0 there is now a fast \PNG\ embedding mode that goes without uncompressing; the image data are directly copied into the \PDF\ stream, resulting in a much higher embedding speed. This direct mode is only activated if the image array structure of the \PNG\ file is compatible with the \PDF\ image structure; \eg\ an interlaced \PNG\ image requires uncompressing to re\hyph arrange the image lines. Luckily, it seems that in practice \PNG\ files commonly allow fast copying. The use of gamma correction disables fast copying, as it requires calculations with individual pixels. Whether the fast copy mode is used for a \PNG\ image can be seen from the log file, which then shows the string `(PNG copy)' after the \PNG\ file name. The \JPEG\ format is normally used in lossy mode; then it's ideal for embedding photos. It's not recommended for crisp images from synthetic sources with a limited amount of colors. Both \JFIF\ and \EXIF\ are supported for additional information. The \JBIGTWO\ format works only for bi-tonal (black and white) pixel images like scanned line and text documents, but for these it has typically a much higher compression ratio than the other two pixel image formats. The \JBIGTWO\ format is part of the \PDF\ standard since version 1.5; native \JBIGTWO\ image inclusion is available in \PDFTEX\ since version 1.40.0. A \JBIGTWO\ file might contain many images, which gives an even better compression ratio than with a single image per file, as \JBIGTWO\ encoders can exploit similarities between bit patterns over several images. Encoders for \JBIGTWO\ can operate in lossy as well as lossless modes. A free \JBIGTWO\ encoder is available: see \useurl{jbigenc} (perhaps there are others). Other options for graphics in \PDFTEX\ are: \begin{description} \item[\LATEX\ picture mode] Since this is implemented simply in terms of font characters, it works in exactly the same way as usual. \item[Xy-pic] If the \POSTSCRIPT\ backend is not requested, Xy-pic uses its own Type~1 fonts, and needs no special attention. \item[tpic] The tpic \cs{special} commands (used in some macro packages) can be redefined to produce literal \PDF, using some macros written by Hans Hagen. \item[\METAPOST] Although the output of \METAPOST\ is \POSTSCRIPT, it is in a highly simplified form, and a \METAPOST\ to \PDF\ conversion (\MPTOPDF, available as \type{supp-pdf.tex} and \type{supp-pdf.mkii}, written by Hans Hagen and Tanmoy Bhattacharya) is implemented as a set of macros which reads \METAPOST\ output and supports all of its features. \end{description} The inclusion of raw \POSTSCRIPT\ commands, a technique utilized by for instance the \type{pstricks} package, cannot directly be supported. Although \PDF\ is a direct descendant of \POSTSCRIPT, it lacks any programming language commands, and cannot deal with arbitrary \POSTSCRIPT. \chapter{Additional \PDF\ keys} \label{sec.addpdfkeys} {\em This section is based on the manual on keys written by Martin Schr\"oder.} A \PDF\ document should contain only the structures and attributes defined in the \PDF\ specification. However, the specification allows applications to insert additional keys, provided they follow certain rules. The most important rule is that developers have to register with Adobe prefixes for the keys they want to insert. Hans Hagen has registered the prefix \type{PTEX} for \PDFTEX. \PDFTEX\ generates an XObject for every included \PDF\ image. The dictionary of this object contains these additional keys: \begin{smalltable} \begin{tabular}{lll} \bf key & \bf type & meaning \cr \tt PTEX.FileName & string & The name of the included file as seen by \PDFTEX. \cr \tt PTEX.InfoDict & dictionary & The \type{InfoDict} of the included \PDF\ (an indirect object). \cr \tt PTEX.PageNumber & integer & The page number of the included file. \cr \end{tabular} \end{smalltable} The \PDFReference\ says: ``Although viewer applications can store custom metadata in the document information dictionary, it is inappropriate to store private content or structural information there; such information should be stored in the document catalog instead.'' Although it would seem more natural to put this information in the document information dictionary, we have to obey the rules laid down in the \PDFReference. The following key ends up in the document catalog. \begin{smalltable} \begin{tabular}{llp{.66\hsize}} \bf key & \bf type & meaning \cr \tt PTEX.Fullbanner & string & The full version of the \PDFTEX\ binary that produced the file as displayed by {\tt pdftex \type{--version}}, a.k.a.\ \cs{pdftexbanner} ({\tt\pdftexbanner}). This is necessary because the string in the \type{Producer} key in the info dictionary is rather short, namely {\tt pdfTeX-\currentpdftex}. \cr \end{tabular} \end{smalltable} Any or all of these keys can be suppressed with the \cs{pdfsuppressptexinfo} primitive, described in section~\ref{sec.docinfocatalog}. \chapter{Character translation} Characters that are input to \PDFTEX\ are subject to optional \TEX\ character translation (\TCX) under control of a \TCX\ file. The \TCX\ maps the input character codes (\eg\ from \cs{input} or \cs{read}) to the character codes as seen by \PDFTEX. This mapping takes place before the characters enter \PDFTEX's `mouth'. If no \TCX\ file is read, the input characters enter \PDFTEX\ directly; no mapping is done. \TCX\ files consist of lines each containing one or two integer numbers in the range 0..255, either in decimal or hex notation. A comment sign \type{%} in a \TCX\ line starts a comment until the end of line. The first number in each line is for matching the input character code, the second, optional number is the corresponding \TEX\ character code. If a line contains only one number, characters with this code enter \PDFTEX\ unchanged; no mapping is done. \TCX\ mapping also influences \PDFTEX\ output streams for \cs{message} and \cs{write}. Without \TCX\ mapping, only characters that are within the range 32..126 are flagged as `printable', meaning that these characters are output directly by \cs{message} and \cs{write} primitives. Characters outside the range 32..126 are instead output in escaped form, \eg\ as \type{^^A} for a character with code \type{0x01}. When a character code is mentioned in the 2nd column of the \TCX\ file, or as the only value in a line, it is flagged as `printable'. During \cs{message} and \cs{write}, output characters are mapped in reverse direction: they are looked up in the 2nd column of the \TCX\ file and the corresponding values from the 1st column are output. Again, if a \PDFTEX\ character code is found as the only number in a line, no mapping is done. Mentioning a character code as the only number on a line has the sole purpose to flag this code `printable'. The characters output into the \PDF\ file, \eg\ by the \cs{pdfliteral} or \cs{special} primitives, are not subject to \TCX\ output remapping. Beware: Character translation interferes with the \ENCTEX\ primitives; to avoid surprises, don't use \ENCTEX\ and \TCX\ mapping at the same time. Further details about \TCX\ file loading can be found in the \WEBC\ manual, \useurl{tcxfiles}. Nowadays \TCX\ files are rarely used. The \type{-8bit} command line option makes all characters printable, when that is desired. \appendix \chapter{Installation} This section describes the steps needed to get \PDFTEX\ running on a system where \PDFTEX\ is not yet installed. Nowadays all known \TEX\ distributions, such as \TEXLIVE, \MIKTEX, and \MACTEX, include \PDFTEX\ already. For example, the \TEXLIVE\ distribution comes with \PDFTEX\ versions for many \UNIX, \WIN, and Mac systems; more information can be found at \hbox{\useurl{texlive}.} When you use any of these distributions, you don't need to bother with the \PDFTEX\ installation procedure in this chapter. If there is no precompiled \PDFTEX\ binary for your system, or the version coming with a distribution is not the current one and you would like to try out a fresh \PDFTEX\ immediately, you will need to build \PDFTEX\ from sources; read on. You should already have a working \TEX\ system, \eg\ \TEXLIVE, into which the freshly compiled \PDFTEX\ will be integrated. (Numerous support files are required in order to have a working binary.) Note that the installation description in this manual is \WEBC\hyph specific. \section{Getting sources and binaries} The primary home page is \useurl{ptexorg}, where you also find bug tracking information. Development sources are at \useurl{ptexdevel}. Precompiled binaries for some platforms are available in subdirectories below \useurl{ctansystems}, although these are unlikely to be up to date. Your best bet is to use a current \TEX\ distribution. \section{Compiling} Given a directory where the sources have been downloaded or checked out, some working on a \UNIX\ system the following steps are needed to compile \PDFTEX: \begin{verbatim} cd source # from branches/stable! ./build-pdftex.sh \end{verbatim} The binary \filename{pdftex} will hopefully then end up in the subdirectory \filename{build-pdftex/texk/web2c}. The additional utilities \filename{pdftosrc} and \filename{ttf2afm} are in the \PDFTEX\ source repository, but are not built by default. If they are needed, edit \filename{build-pdftex.sh} to \type{--enable} them. For \PDFTEX\ maintainers: a sibling script to \type{build-pdftex.sh} is in the repository, namely \type{sync-pdftex.sh}, which syncs changes from a \TEXLIVE\ source repository to a \PDFTEX\ source repository. Read and understand the script before using it! \section{Placing files} The next step is to put the newly-compiled \filename{pdftex} (and, if enabled, \filename{pdftosrc} and/or \filename{ttf2afm}) binaries into the installed distribution. \Eg\ for a typical \TEXLIVE\ system:\\ \texttt{/usr/local/texlive/\rcsyear/bin/}\type{x86_64-linux}, replacing \type{x86_64-linux} with the appropriate platform name. (As always, it is sensible to rename the installed \filename{pdftex} binary first and not just overwrite it.) If you're doing this into a live hierarchy, run \verb|fmtutil-sys refresh| afterwards, so that all formats are regenerated system-wide with the new \filename{pdftex} binary. If what you want is to test a new version of \filename{pdftex}, and not replace the installed version, the best approach is to copy it into the binary directory as, say, \filename{pdftex.new}. The different executable name can sometimes change the files found and other behavior, but usually this won't matter. \PDFTEX\ uses the Kpathsea library to search for supporting files; many variables and configuration files can come into play. See the Kpathsea manual: \useurl{kpathsea}. \section{Configuration of \PDFTEX} \label{sec.cfg} As opposed to \TEX\ with its \DVI\ output, the \PDFTEX\ program does not have a separate postprocessing stage to transform the \TEX\ input into final \PDF. As a consequence, all data needed for building the \PDF\ must be available during the \PDFTEX\ run, in particular information on media dimensions and offsets, graphics files for embedding, and font information (font files, encodings). When \TEX\ builds a page, it places items relative to the (1in,1in) offset from the top left page corner (the \DVI\ reference point). Separate \DVI\ postprocessors allow specifying the paper size (\eg\ A4 or letter), so that this reference point is moved to the correct position on the paper, and the text ends up at the right place. In \PDF, the paper dimensions are part of the page definition, and \PDFTEX\ therefore requires that they be defined at the beginning of the \PDFTEX\ run. As with pages described by \POSTSCRIPT, the \PDF\ reference point is in the lower left corner. Formerly, these dimensions and other \PDFTEX\ parameters were read in from a configuration file named \filename{pdftex.cfg}, which had a special (non-\TEX) format, at the start of processing. Nowadays such a file is ignored by \PDFTEX. Instead, the page dimensions and offsets, as well as many other parameters, can be set by \PDFTEX\ primitives during the \PDFTEX\ format building process, so that the settings are dumped into the fresh format and consequently will be used when \PDFTEX\ is later called with that format. All settings from the format can still be overridden during a \PDFTEX\ run by using the same primitives. This configuration concept is a more unified approach, as it avoids a configuration file with a special format. A list of \PDFTEX\ primitives likely relevant to setting up the \PDFTEX\ engine is given in the table below. All primitives are described in detail in previous sections. \begin{smalltable} \begin{tabular}{@{}llll@{}} \bf primitive name & \bf type & \bf default & \bf comment\cr \cs{pdfoutput} & integer & 0 & \DVI\ by default \cr \cs{pdfadjustspacing} & integer & 0 & off \cr \cs{pdfcompresslevel} & integer & 9 & best \cr \cs{pdfobjcompresslevel} & integer & 0 & no object streams \cr \cs{pdfdecimaldigits} & integer & 3 & \cr \cs{pdfimageresolution} & integer & 72 & dpi \cr \cs{pdfpkresolution} & integer & 0 & \cr \cs{pdfpkmode} & tokens & empty & mode per \type{mktex.cnf}\cr \cs{pdfuniqueresname} & integer & 0 & \cr \cs{pdfprotrudechars} & integer & 0 & \cr \cs{pdfgentounicode} & integer & 0 & \cr \cs{pdfmajorversion} & integer & 1 & output \PDF\ 1.x \cr \cs{pdfminorversion} & integer & 4 & specifically \PDF\ 1.4 \cr \cs{pdfpagebox} & integer & 0 & \cr \cs{pdfforcepagebox} & integer & 0 & \cr \cs{pdfinclusionerrorlevel} & integer & 0 & \cr %---------------------------------------------------------------------------- \cs{pdfhorigin} & dimension & 1in & \cr \cs{pdfvorigin} & dimension & 1in & \cr \cs{pdfpagewidth} & dimension & 0pt & \cr \cs{pdfpageheight} & dimension & 0pt & \cr \cs{pdflinkmargin} & dimension & 0pt & \cr \cs{pdfdestmargin} & dimension & 0pt & \cr \cs{pdfthreadmargin} & dimension & 0pt & \cr \cs{pdfmapfile} & text & \filename{pdftex.map} & not dumped\cr \end{tabular} \end{smalltable} Figure~\ref{fig.configfile} shows a recent \PDFTEX\ configuration file (\type{pdftexconfig.tex}) from \TEXLIVE\ (part of the \type{tex-ini-files} package), overriding some of these settings. It is read when a format is built. It enables \PDF\ output, sets paper dimensions and the default pixel density for \PK\ font inclusion. The default values are chosen so that \PDFTEX\ often can be used (\eg\ in \type{-ini} mode) without setting any additional parameters. \begin{figure}[t] \myverbatiminput{incl/pdftexconfig.txt} \caption{\PDFTEX\ configuration file for \TEXLIVE\ (\texttt{pdftexconfig.tex}).} \label{fig.configfile} \end{figure} Independent of whether such a configuration file is read, the first action in a \PDFTEX\ run is reading the global \WEBC\ configuration file (\filename{texmf.cnf}), which is common to all programs in the \WEBC\ system. This file mainly defines file search paths, the memory layout (\eg\ string pool and hash size), and a few other general parameters. \section{Creating format files} The \PDFTEX\ engine supports building separate formats for either \DVI\ or \PDF\ output in the same way as the classical \TEX\ engine does for \DVI. Format generation (and other \type{initex} features) is enabled by the \type{-ini} option. The default mode (\DVI\ or \PDF) can be chosen either on the command line by setting the option \type{-output-format} to \type{dvi} or \type{pdf}, or by setting the \cs{pdfoutput} parameter. The format file inherits this setting, so that a later invocation of \PDFTEX\ with this format starts in the preselected mode (which can still be overridden). A format file can be read in only by the engine that has generated it; a format incompatible with an engine leads to a fatal error. It is customary to package the configuration and macro file input into a \type{.ini} file. \Eg\ the file \type{etex.ini} in figure~\ref{fig.etex.ini} is for generating an \ETEX\ format with \DVI\ output. It has been traditional for many years to generate \type{etex.fmt} with \PDFTEX\ rather than the original \ETEX, because \PDFTEX\ contains useful additional programming and other non-\PDF-related features. The \type{pdfetex.ini} file figure~\ref{fig.pdfetex.ini} shows the corresponding format with \PDF\ output by default; this is what creates the format file read when \type{pdftex} is normally invoked. The corresponding \PDFTEX\ invocations for format generation are: \begin{verbatim} pdftex -ini *etex.ini pdftex -ini *pdfetex.ini \end{verbatim} These calls produce format files \filename{etex.fmt}, \filename{pdfetex.fmt}, as the default format file name is taken from the input file name. You can override this with the \type{-jobname} option. The asterisk \type{*} before the file name is an unusual flag, only used in \type{-ini} mode, which causes the \PDFTEX\ engine to enable \ETEX's features. To reiterate, the distribution (\TEXLIVE, \MIKTEX) usually takes care of format (re)generation. The above is if you need to do such things manually for testing, debugging, development, etc. Incidentally, as of \PDFTEX\ 1.40.21 (\TEX\ Live 2020), \filename{.fmt} files are compressed with \type{zlib}. This makes for a considerable savings in space, and consequently in time. \begin{figure}[t] \myverbatiminput{incl/ini-etex.txt} \caption{The \type{etex.ini} file to dump the plain \ETEX\ format with \DVI\ output.} \label{fig.etex.ini} \end{figure} \begin{figure}[t] \myverbatiminput{incl/ini-pdfetex.txt} \caption{The \type{pdfetex.ini} file to dump plain \ETEX\ with \PDF\ output.} \label{fig.pdfetex.ini} \end{figure} \section{Testing the installation} When everything is set up, you can test the installation. A simple test of plain \PDFTEX\ is: \begin{verbatim} pdftex story \\end \end{verbatim} This should typeset the famous one-page short story by A.U. Thor, generating a \PDF\ file. A more thorough and descriptive test is the plain \TEX\ test file \filename{samplepdf.tex}, available in the distribution in the \type{samplepdftex/} directory. Process this file by typing: \begin{verbatim} pdftex samplepdf \end{verbatim} If the installation is ok, this should produce a file called \filename{samplepdf.pdf}. The file \filename {samplepdf.tex} is also a good place to look for examples of how to use \PDFTEX's primitives. \section{Common problems} The most common problem with installations is that \PDFTEX\ complains that some file cannot be found. In such cases, first make sure that all \TEX-related environment variables are unset. For detailed debugging, set the environment variable \type{KPATHSEA_DEBUG=255} before running \PDFTEX\ or use the option \verb|-kpathsea-debug 255|. More options can be found in the \WEBC\ documentation. Variables in \filename{texmf.cnf} can be overwritten by environment variables. Here are some of the most common problems you can encounter when getting started: \begin{itemize} \item \verb|I can't find the format file `pdftex.fmt'!| \newline \verb|I can't find the format file `pdflatex.fmt'!| The format file is not created (see above how to do that) or is not properly placed. Make sure that \type{TEXFORMATS} in \filename {texmf.cnf} contains the path to \filename {pdftex.fmt} or \filename {pdflatex.fmt}. \item \verb|Fatal format file error; I'm stymied| This typically appears if you forgot to regenerate the \type{.fmt} files after installing a new version of the \PDFTEX\ binary. The first line tells by which engine the offending format was generated. \item \PDFTEX\ cannot find one or more map files (\type{*.map}), encoding vectors (\type{*.enc}), virtual fonts, Type~1 fonts, TrueType or OpenType fonts, or some image file. Make sure that the required file exists and the corresponding variable in \filename{texmf.cnf} contains a path to the file. When you have installed new fonts, and your \PDF\ viewer complains about missing fonts, you should take a look at the log file produced by \PDFTEX. Missing fonts, map files, encoding vectors as well as missing characters (glyphs) are reported there. \end{itemize} For more help resources of all kinds, see \useurl{tughelp}. \chapter{Formal syntax specification} \label{sec.syntax} This appendix formally specifies the \PDFTEX-specific extensions to the \TEX\ macro programming language. Most primitive names are prefixed by `\type{pdf}'. General definitions and syntax rules follow after the list of primitives. Two new units of measure were introduced in \PDFTEX\ 1.30.0: the new Didot ($\mbox{1nd}=0.375$mm) and the new Cicero ($\mbox{1nc}=12$nd). The former was proposed by \ISO\ in 1975. \begingroup \parindent=0pt % because the summary is nothing but a long string of \Syntax % commands, we have to allow breaks, in contrast to the main text. \def\aftersyntax{\par\vskip 1.5pt plus1pt minus1pt\relax} % Generated list of primitives (see Makefile). \input{incl/pdftex-syntax.tex} \section{General definitions and syntax rules} \Syntax{ \Something{general text} \Means \Lbrace \Something{balanced text} \Rbrace } \Syntax{ \Something{attr spec} \Means \Literal{attr} \Something{general text} } \Syntax{ \Something{resources spec} \Means \Literal{resources} \Something{general text} } \Syntax{ \Something{rule spec} \Means (\Literal{width} \Or \Literal{height} \Or \Literal{depth}) \Something{dimen} \Optional{\Something{rule spec}} } %\Syntax{\Something{object type spec} \Means % \Optional{\Literal{stream} % \Optional{\Something{attr spec}}} \Something{object contents}} \Syntax{ \Something{object type spec} \Means \Literal{reserveobjnum} \Or \Next \Optional{\Literal{useobjnum} \Something{number}} \Next \Optional{\Literal{stream} \Optional{\Something{attr spec}}} \Something{object contents} } \Syntax{ \Something{annot type spec} \Means \Literal{reserveobjnum} \Or \Next \Optional{\Literal{useobjnum} \Something{number}} \Optional{\Something{rule spec}} \Something{general text} } \Syntax{ \Something{object contents} \Means \Something{file spec} \Or \Something{general text} } \Syntax{ \Something{xform attr spec} \Means \Optional{\Something{attr spec}} \Optional{\Something{resources spec}} } \Syntax{ \Something{image attr spec} \Means \Optional{\Something{rule spec}} \Optional{\Something{attr spec}} \Optional{\Something{page spec}} \Optional{\Something{colorspace spec}} \Next \Optional{\Something{pdf box spec}} } \Syntax{ \Something{outline spec} \Means \Optional{\Something{attr spec}} \Something{action spec} \Optional{\Literal{count} \Something{number}} \Something{general text} } \Syntax{ \Something{action spec} \Means \Literal{user} \Something{user-action spec} \Or \Literal{goto} \Something{goto-action spec} \Or \Next \Literal{thread} \Something{thread-action spec} } \Syntax{ \Something{user-action spec} \Means \Something{general text} } \Syntax{ \Something{goto-action struct spec} \Means \Literal{struct} (\Something{numid} \Or \Literal{nameid} \Or \Something{general text}) } % keep in sync with redundant copy below. \Syntax{ \Something{goto-action spec} \Means \Optional{\Something{goto-action struct spec}} \Something{numid} \Or \Next \Optional{\Something{file spec}} \Optional{\Something{goto-action struct spec}} \Something{nameid} \Or \Next \Optional{\Something{file spec}} \Optional{\Something{goto-action struct spec}} \Optional{\Something{page spec}} \Something{general text} \Or \Next \Something{file spec} \Optional{\Something{goto-action struct spec}} \Something{nameid} \Something{newwindow spec} \Or \Next \Something{file spec} \Optional{\Something{goto-action struct spec}} \Optional{\Something{page spec}} \Something{general text} \Something{newwindow spec} } \Syntax{ \Something{thread-action spec} \Means \Optional{\Something{file spec}} \Something{numid} \Or \Optional{\Something{file spec}} \Something{nameid} } \Syntax{ \Something{colorspace spec} \Means \Literal{colorspace} \Something{number} } \Syntax{ \Something{pdf box spec} \Means \Literal{mediabox} \Or \Literal{cropbox} \Or \Literal{bleedbox} \Or \Literal{trimbox} \Or \Literal{artbox} } \Syntax{ \Something{map spec} \Means \Lbrace \Optional{\Something{map modifier}} \Something{balanced text} \Rbrace } \Syntax{ \Something{map modifier} \Means \Literal{+} \Or \Literal{=} \Or \Literal{-} } \Syntax{ \Something{numid} \Means \Literal{num} \Something{number} } \Syntax{ \Something{nameid} \Means \Literal{name} \Something{general text} } \Syntax{ \Something{newwindow spec} \Means \Literal{newwindow} \Or \Literal{nonewwindow} } \Syntax{ \Something{dest spec} \Means \Optional{\Literal{struct} \Something{number}} (\Something{numid} \Or \Something{nameid}) \Something{dest type} } \Syntax{ \Something{dest type} \Means \Literal{xyz} \Optional{\Literal{zoom} \Something{number}} \Or \Literal{fitr} \Something{rule spec} \Or \Next \Literal{fitbh} \Or \Literal{fitbv} \Or \Literal{fitb} \Or \Literal{fith} \Or \Literal{fitv} \Or \Literal{fit} } \Syntax{ \Something{thread spec} \Means \Optional{\Something{rule spec}} \Optional{\Something{attr spec}} \Something{id spec} } \Syntax{ \Something{id spec} \Means \Something{numid} \Or \Something{nameid} } \Syntax{ \Something{file spec} \Means \Literal{file} \Something{general text} } \Syntax{ \Something{page spec} \Means \Literal{page} \Something{number} \Or% \Literal{named} \Something{general text} } \Syntax{ \Something{expand spec} \Means \Something{stretch} \Something{shrink} \Something{step} \Optional{\Literal{autoexpand}} } \Syntax{ \Something{stretch} \Means \Something{number} } \Syntax{ \Something{shrink} \Means \Something{number} } \Syntax{ \Something{step} \Means \Something{number} } \Syntax{ \Something{pre spec} \Means \Literal{pre} } \Syntax{ \Something{stack action} \Means \Literal{set} \Or \Literal{push} \Or \Literal{pop} \Or \Literal{current} } \endgroup % of syntax definitions \medskip A \Something{general text} is expanded immediately, like \cs{special} in traditional \TEX, unless explicitly mentioned otherwise. Some of the object and image-related primitives can be prefixed by \cs{immediate}. \chapter{Abbreviations used in this manual} In this document we use numerous abbreviations. For convenience we give their meanings here. \abbrevlist \twocolumn {\small \emergencystretch=1.5em \input{incl/fdl-1.2.tex}} \end{document}