PDF Graphics Device

Description

pdf starts the graphics device driver for producing PDF graphics.

Usage

pdf(file = if(onefile) "Rplots.pdf" else "Rplot%03d.pdf",
    width, height, onefile, family, title, fonts, version,
    paper, encoding, bg, fg, pointsize, pagecentre, colormodel,
    useDingbats, useKerning, fillOddEven, compress)

Arguments

Details

All arguments except file default to values given by pdf.options(). The ultimate defaults are quoted in the arguments section.

pdf() opens the file file and the PDF commands needed to plot any graphics requested are sent to that file.

The family argument can be used to specify a PDF-specific font family as the initial/default font for the device. If additional font families are to be used they should be included in the fonts argument.

If a device-independent R graphics font family is specified (e.g., via par(family = ) in the graphics package), the PDF device makes use of the PostScript font mappings to convert the R graphics font family to a PDF-specific font family description. (See the documentation for pdfFonts.)

This device does not embed fonts in the PDF file, so it is only straightforward to use mappings to the font families that can be assumed to be available in any PDF viewer: "Times" (equivalently "serif"), "Helvetica" (equivalently "sans") and "Courier" (equivalently "mono"). Other families may be specified, but it is the user's responsibility to ensure that these fonts are available on the system and third-party software (e.g., Ghostscript) may be required to embed the fonts so that the PDF can be included in other documents (e.g., LaTeX): see embedFonts. The URW-based families described for in section ‘Families’.can be used with viewers, platform dependently:

on Unix-alikes

viewers set up to use URW fonts, which is usual with those based on xpdf or Ghostscript.

on Windows

viewers such as GSView which utilise URW fonts.

Since embedFonts makes use of Ghostscript, it should be able to embed the URW-based families for use with other viewers.

The PDF produced is fairly simple, with each page being represented as a single stream (by default compressed and possibly with references to raster images). The R graphics model does not distinguish graphics objects at the level of the driver interface.

The version argument declares the version of PDF that gets produced. The version must be at least 1.2 when compression is used, 1.4 for semi-transparent output to be understood, and at least 1.3 if CID fonts are to be used: if any of these features are used the version number will be increased (with a warning). (PDF 1.4 was first supported by Acrobat 5 in 2001; it is very unlikely not to be supported in a current viewer.)

Line widths as controlled by par(lwd = ) are in multiples of 1/96 inch. Multiples less than 1 are allowed. pch = "." with cex = 1 corresponds to a square of side 1/72 inch, which is also the ‘pixel’ size assumed for graphics parameters such as "cra".

The paper argument sets the ‘⁠/MediaBox⁠’ entry in the file, which defaults to width by height. If it is set to something other than "special", a device region of the specified size is (by default) centred on the rectangle given by the paper size: if either width or height is less than 0.1 or too large to give a total margin of 0.5 inch, it is reset to the corresponding paper dimension minus 0.5. Thus if you want the default behaviour of postscript use pdf(paper = "a4r", width = 0, height = 0) to centre the device region on a landscape A4 page with 0.25 inch margins.

When the background colour is fully transparent (as is the initial default value), the PDF produced does not paint the background. Most PDF viewers will use a white canvas so the visual effect is if the background were white. This will not be the case when printing onto coloured paper, though.

File specifications

Tilde expansion (see path.expand) is done on the file argument. An input with a marked encoding is converted to the native encoding or an error is given.

For use with onefile = FALSE, give a C integer format such as "Rplot%03d.pdf" (the default in that case) which is expanded using the page number, so this uses files ‘Rplot001.pdf’, ..., ‘Rplot999.pdf’, ‘Rplot1000.pdf’,

A single integer format matching the regular expression "%[#0 +=-]*[0-9.]*[diouxX]" is allowed in file. The character string should not otherwise contain a %: if it is really necessary, use %% in the string for % in the file path.

For pdf, file can be NULL when no external file is created (effectively, no drawing occurs), but the device may still be queried (e.g., for the size of text by (base graphics) strwidth or (grid) stringWidth).

Families

Font families are collections of fonts covering the five font faces, (conventionally plain, bold, italic, bold-italic and symbol) selected by the graphics parameter par(font = ) or the grid parameter gpar(fontface = ). Font families can be specified either as an initial/default font family for the device via the family argument or after the device is opened by the graphics parameter par(family = ) or the grid parameter gpar(fontfamily = ). Families which will be used in addition to the initial family must be specified in the fonts argument when the device is opened.

Font families are declared via a call to pdfFonts or postscriptFonts.

The argument family specifies the initial/default font family to be used. In normal use it is one of "AvantGarde", "Bookman", "Courier", "Helvetica", "Helvetica-Narrow", "NewCenturySchoolbook", "Palatino" or "Times", and refers to the standard Adobe PostScript fonts families of those names which are included (or cloned) in all common PDF/PostScript renderers.

Many PDF/PostScript renders (including those based on Ghostscript) use the URW equivalents of these fonts, which are "URWGothic", "URWBookman", "NimbusMon", "NimbusSan", "NimbusSanCond", "CenturySch", "URWPalladio" and "NimbusRom" respectively. If your viewer is using URW fonts, you will obtain access to more characters and more appropriate metrics by using these names. To make these easier to remember, "URWHelvetica" == "NimbusSan" and "URWTimes" == "NimbusRom" are also supported. However, if the viewer is not using URW fonts (for example Adobe Acrobat Reader) it may substitute inappropriately or not render at all. (Consider using embedFonts.)

As from R 4.4.0 there is support for URW 2.0 fonts in families "URW2Helvetica" (with ‘⁠Oblique⁠’ fonts), "URW2HelveticaItalic" (with ‘⁠Italic⁠’ fonts), "URW2Times" and "NimbusMonoPS". As recent versions of Ghostscript will render with (and embed) these fonts. these families should be used instead of "URWHelvetica", "URWTimes", "NimbusSan", "NimbusRom", and "NimbusMon"

Another type of family makes use of CID-keyed fonts for East Asian languages – see pdfFonts.

The family argument is normally a character string naming a font family, but family objects generated by Type1Font and CIDFont are also accepted. For compatibility with earlier versions of R, the initial family can also be specified as a vector of four or five afm files.

Note that R does not embed the font(s) used in the PostScript output: see embedFonts for a utility to help do so.

Viewers and embedding applications frequently substitute fonts for those specified in the family, and the substitute will often have slightly different font metrics. useKerning = TRUE spaces the letters in the string using kerning corrections for the intended family: this may look uglier than useKerning = FALSE.

Encodings

Encodings describe which glyphs are used to display the character codes (in the range 0–255). Most commonly R uses ISOLatin1 encoding, and the examples for text are in that encoding. However, the encoding used on machines running R may well be different, and by using the encoding argument the glyphs can be matched to encoding in use. This suffices for European and Cyrillic languages, but not for East Asian languages. For the latter, composite CID fonts are used. These fonts are useful for other languages: for example they may contain Greek glyphs. (The rest of this section applies only when CID fonts are not used.)

None of this will matter if only ASCII characters (codes 32–126) are used as all the encodings (except "TeXtext") agree over that range. Some encodings are supersets of ISOLatin1. However, if accented and special characters do not come out as you expect, you may need to change the encoding. Some other encodings are supplied with R: "ISOLatin2.enc" (Central/Eastern Europe), "ISOLatin7.enc" (ISO 8859-13, ‘Baltic Rim’), "ISOLatin9.enc" (ISO 8859-15, including Euro), "Cyrillic.enc" (ISO 8859-5), "KOI8-R.enc", "KOI8-U.enc", and the Windows encodings "WinAnsi.enc" (also known as "CP1252.enc", "CP1250.enc" (Central/Eastern Europe), "CP1251.enc" (Cyrillic), "Greek.enc" (ISO 8859-7), "CP1253.enc" (modern Greek) and "CP1257.enc" (‘Baltic Rim’). Note that many glyphs in these encodings are not in the fonts corresponding to the standard families. (The Adobe ones for all but Courier, Helvetica and Times cover little more than Latin-1, whereas the URW ones also cover Latin-2, Latin-7, Latin-9 and Cyrillic but no Greek. The Adobe exceptions cover the Latin character sets, but not the Euro.)

NB: support for encodings other than "ISOLatin1.enc" (and the Windows ones on Windows) depends on support by the platform's libiconv in a UTF-8 locale.

If you specify the encoding, it is your responsibility to ensure that the PostScript font contains the glyphs used. One issue here is the Euro symbol which is in several encodings (including WinAnsi and ISOLatin9 encodings) but may well not be in the PostScript fonts. (It is in the URW variants; it is not in the supplied Adobe Font Metric files so will not be centred correctly.)

There is an exception. Character 45 ("-") is always set as minus (its value in Adobe ISOLatin1) even though it is hyphen in the other encodings. Hyphen is available as character 173 (octal 0255) in all the Latin encodings, Cyrillic and Greek. (This can be entered as "\u00ad" in a UTF-8 locale.) There are some discrepancies in accounts of glyphs 39 and 96: the supplied encodings (except CP1250 and CP1251) treat these as ‘quoteright’ and ‘quoteleft’ (rather than ‘quotesingle’/‘acute’ and ‘grave’ respectively), as they are in the Adobe documentation.

Color models

The default color model ("srgb") is sRGB. Model "gray" (or "grey") maps sRGB colors to greyscale using perceived luminosity (biased towards green). "cmyk" outputs in CMYK colorspace. The simplest possible conversion from sRGB to CMYK is used (https://en.wikipedia.org/wiki/CMYK_color_model#Mapping_RGB_to_CMYK), and raster images are output in RGB.

Also available for backwards compatibility is model "rgb" which uses uncalibrated RGB and corresponds to the model used with that name in R prior to 2.13.0. Some viewers may render some plots in that colorspace faster than in sRGB, and the plot files will be smaller.

Conventions

This section describes the implementation of the conventions for graphics devices set out in the ‘R Internals’ manual.

• The default device size is 7 inches square.

• Font sizes are in big points.

• The default font family is Helvetica.

• Line widths are as a multiple of 1/96 inch, with a minimum of 0.01 enforced.

• Circles of any radius are allowed. If useDingbats = TRUE, opaque circles of less than 10 big points radius are rendered using char 108 in the Dingbats font: all semi-transparent and larger circles using a Bézier curve for each quadrant.

• Colours are by default specified as sRGB.

At very small line widths, the line type may be forced to solid.

Printing

Except on Windows it is possible to print directly from pdf by something like (this is appropriate for a CUPS printing system):

    pdf("|lp -o landscape", paper = "a4r")

This forces onefile = TRUE.

Note

If you have drawn any typeset glyphs (see glyphInfo) then it is highly recommended that you use embedGlyphs to embed the relevant fonts.

Note

If you see problems with PDF output, do remember that the problem is much more likely to be in your viewer than in R. Try another viewer if possible. Symptoms for which the viewer has been at fault are apparent grids on image plots (turn off graphics anti-aliasing in your viewer if you can) and missing or incorrect glyphs in text (viewers silently doing font substitution).

Unfortunately the default viewers on most Linux and macOS systems have these problems, and no obvious way to turn off graphics anti-aliasing.

Acrobat Reader does not use the fonts specified but rather emulates them from multiple-master fonts. This can be seen in imprecise centering of characters, for example the multiply and divide signs in Helvetica. This can be circumvented by embedding fonts where possible. Most other viewers substitute fonts, e.g. URW fonts for the standard Helvetica and Times fonts, and these too often have different font metrics from the true fonts.

Acrobat Reader can be extended by ‘font packs’, and these will be needed for the full use of encodings other than Latin-1 (although they may be offered for download as needed).

On some Unix-alike systems:

If useDingbats = TRUE, the default plotting character pch = 1 was displayed in some PDF viewers incorrectly as a "q" character. (These seem to be viewers based on the ‘⁠poppler⁠’ PDF rendering library). This may be due to incorrect or incomplete mapping of font names to those used by the system. Adding the following lines to ‘~/.fonts.conf’ or ‘/etc/fonts/local.conf’ may circumvent this problem, although this has largely been corrected on the affected systems.

<fontconfig>
<alias binding="same">
  <family>ZapfDingbats</family>
  <accept><family>Dingbats</family></accept>
</alias>
</fontconfig>

Some further workarounds for problems with symbol fonts on viewers using ‘fontconfig’ are given in the ‘Cairo Fonts’ section of the help for X11.

On Windows:

The TeXworks PDF viewer was one of those which has been seen to fail to display Dingbats (used by e.g. pch = 1) correctly. Whereas on other platforms the problems seen were incorrect output, on Windows points were silently omitted: however recent versions seem to manage to display Dingbats.

There was a different font bug in the pdf.js viewer included in Firefox: that mapped Dingbats to the Symbol font and so displayed symbols such pch = 1 as lambda.

See Also

pdfFonts, pdf.options, embedFonts, glyphInfo, Devices, postscript.

cairo_pdf and (on macOS only) quartz for other devices that can produce PDF.

More details of font families and encodings and especially handling text in a non-Latin-1 encoding and embedding fonts can be found in

Paul Murrell and Brian Ripley (2006). “Non-standard fonts in PostScript and PDF graphics.” R News, 6(2), 41–47. https://www.r-project.org/doc/Rnews/Rnews_2006-2.pdf.

Examples

## Test function for encodings
TestChars <- function(encoding = "ISOLatin1", ...)
{
    pdf(encoding = encoding, ...)
    par(pty = "s")
    plot(c(-1,16), c(-1,16), type = "n", xlab = "", ylab = "",
         xaxs = "i", yaxs = "i")
         title(paste("Centred chars in encoding", encoding))
    grid(17, 17, lty = 1)
    for(i in c(32:255)) {
        x <- i %% 16
        y <- i %/% 16
        points(x, y, pch = i)
    }
    dev.off()
}
## there will be many warnings.
TestChars("ISOLatin2")
## this does not view properly in older viewers.
TestChars("ISOLatin2", family = "URWHelvetica")
## works well for viewing in gs-based viewers, and often in xpdf.