The text module is used to create text output. It seamlessly integrates Donald E. Knuths famous TeX typesetting engine[1]. The module is a high-level interface to an extensive stack of TeX and font related functionality in PyX, whose details are way beyond this manual and completely irrelevant for the typical PyX user. However, the basic concept should be described briefly, as it provides important insights into essential properties of the whole machinery.

PyX does not apply any limitations on the text submitted by the user. Instead the text is directly passed to TeX. This has the implication, that the text to be typeset should come from a trusted source or some security measures should have been applied already. PyX just adds a light and transparent wrapper using basic TeX functionality for later identification and output extraction. This procedure enables full access to all TeX features and makes PyX on the other hand dependent on the error handling provided by TeX. However, a detailed and immediate control of the TeX output allows PyX to report problems back to the user as they occur.

While we only talked about TeX so far (and will continue to do so in the rest of this section), it is important to note that the coupling is not limited to plain TeX. Currently, PyX can also use LaTeX for typesetting, and other TeX variants could be added in the future. What PyX really depends on is the ability of the typesetting program to generate DVI[2].

As soon as some text creation is requested or, even before that, a preamble setting or macro definition is submitted, the TeX program is started as a separate process. The input and output is bound to a SingleRunner instance. Typically, the process will be kept alive and will be reused for all future typesetting requests until the end of the PyX process. However, there are certain situations when the TeX program needs to be shutdown early, which are be described in detail in the TeX ipc mode section.

Whenever PyX sends some commands to the TeX interpreter, it adds an output marker at the end, and waits for this output marker to be echoed in the TeX output. All intermediate output is attributed to the commands just sent and will be analysed for problems. This is done by texmessage parsers. Here, a problem could be logged to the PyX logger at warning level, thus be reported to stderr by default. This happens for over- or underful boxes or font warnings emitted by TeX. For other unknown problems (i.e. output not handled by any of the given texmessage parsers), a TexResultError is raised, which creates a detailed error report including the traceback, the commands submitted to TeX and the output returned by TeX.

PyX wraps each text to be typeset in a TeX box and adds a shipout of this box to the TeX code before forwarding it to TeX. Thus a page in the DVI file is created containing just this output. Furthermore TeX is asked to output the box extent. By that PyX will immediately know the size of the text without referring to the DVI. This also allows faking the box size by TeX means, as you would expect it.

Once the actual output is requested, PyX reads the content of the DVI file, accessing the page related to the output in question. It then does all the necessary steps to transform the DVI content to the requested output format, like searching for virtual font files, font metrices, font mapping files, and PostScript Type1 fonts to be used in the final output. Here a present limitation has been mentioned: PyX presently can use PostScript Type1 fonts only to generate text output. While this is a serious limitation, all the default fonts in TeX are available in Type1 nowadays and current TeX installations are alreadily configured to use them by default.

TeX interface

class text.SingleRunner(executable, texenc="ascii", usefiles=[], texipc=config.getboolean("text", "texipc", 0), copyinput=None, dvitype=False, errordetail=errordetail.default, texmessages_start=[], texmessages_end=[], texmessages_preamble=[], texmessages_run=[])

Base class for the TeX interface.


This class cannot be used directly. It is the base class for all texrunners and provides most of the implementation. Still, to the end user the parameters except for executable are important, as they are preserved in derived classes usually.

  • executable (str) – command to start the TeX interpreter
  • texenc (str) – encoding to use in the communication with the TeX interpreter
  • usefiles (list of str) – list of supplementary files to be copied to and from the temporary working directory (see Debugging for usage details)
  • texipc (bool) – TeX ipc mode flag.
  • copyinput (None or str or file) – filename or file to be used to store a copy of all the input passed to the TeX interpreter
  • dvitype (bool) – flag to turn on dvitype-like output
  • errordetail (errordetail) – verbosity of the TexResultError
  • texmessages_start (list of texmessage parsers) – additional message parsers at interpreter startup
  • texmessages_end (list of texmessage parsers) – additional message parsers at interpreter shutdown
  • texmessages_preamble (list of texmessage parsers) – additional message parsers for preamble output
  • texmessages_run (list of texmessage parsers) – additional message parsers for typset output
texmessages_start_default = [texmessage.start]

default texmessage parsers at interpreter startup

texmessages_end_default = [texmessage.end, texmessage.font_warning, texmessage.rerun_warning, texmessage.nobbl_warning]

default texmessage parsers at interpreter shutdown

texmessages_preamble_default = [texmessage.load]

default texmessage parsers for preamble output

texmessages_run_default = [texmessage.font_warning, texmessage.box_warning, texmessage.package_warning, texmessage.load_def, texmessage.load_graphics]

default texmessage parsers for typeset output

preamble(expr, texmessages=[])

Execute a preamble.

  • expr (str) – expression to be executed
  • texmessages (list of texmessage parsers) – additional message parsers

Preambles must not generate output, but are used to load files, perform settings, define macros, etc. In LaTeX mode, preambles are executed before \begin{document}. The method can be called multiple times, but only prior to SingleRunner.text() and SingleRunner.text_pt().

text_pt(x_pt, y_pt, expr, textattrs=[], texmessages=[], fontmap=None, singlecharmode=False)

Typeset text.

  • x_pt (float) – x position in pts
  • y_pt (float) – y position in pts
  • expr (str) – text to be typeset
  • textattrs (list of textattr, :class:`trafo.trafo_pt, and style.fillstyle) – styles and attributes to be applied to the text
  • texmessages (list of texmessage parsers) – additional message parsers
  • fontmap (None or fontmap) – force a fontmap to be used (instead of the default depending on the output format)
  • singlecharmode (bool) – position each character separately

text output insertable into a canvas.

Return type:



TexDoneError: when the TeX interpreter has been terminated already.

text(x, y, *args, **kwargs)

Typeset text.

This method is identical to text_pt() with the only difference of using PyX lengths to position the output.

  • x (PyX length) – x position
  • y (PyX length) – y position
class text.SingleTexRunner(executable=config.get("text", "tex", "tex"), lfs="10pt", **kwargs)

Plain TeX interface.

This class adjusts the SingleRunner to use plain TeX.

  • executable (str) – command to start the TeX interpreter
  • lfs (str or None) – resemble LaTeX font settings within plain TeX by loading a lfs-file
  • kwargs – additional arguments passed to SingleRunner

An lfs-file is a file defining a set of font commands like \normalsize by font selection commands in plain TeX. Several of those files resembling standard settings of LaTeX are distributed along with PyX in the pyx/data/lfs directory. This directory also contains a LaTeX file to create lfs files for different settings (LaTeX class, class options, and style files).

class text.SingleLatexRunner(executable=config.get("text", "latex", "latex"), docclass="article", docopt=None, pyxgraphics=True, texmessages_docclass=[], texmessages_begindoc=[], **kwargs)

LaTeX interface.

This class adjusts the SingleRunner to use LaTeX.

  • executable (str) – command to start the TeX interpreter
  • docclass (str) – document class
  • docopt (str or None) – document loading options
  • pyxgraphics (bool) – activate graphics bundle support, see Using the graphics-bundle with LaTeX
  • texmessages_docclass (list of texmessage parsers) – additional message parsers at LaTeX class loading
  • texmessages_begindoc (list of texmessage parsers) – additional message parsers at \begin{document}
  • kwargs – additional arguments passed to SingleRunner
texmessages_docclass_default = [texmessage.load]

default texmessage parsers at LaTeX class loading

texmessages_begindoc_default = [texmessage.load, texmessage.no_aux]

default texmessage parsers at \begin{document}

The SingleRunner classes described above do not handle restarts of the interpreter when the actuall DVI result is required and is not available via the TeX ipc mode feature.

The MultiRunner classes below are not derived from SingleRunner even though the provide the same functional interface (MultiRunner.preamble(), MultiRunner.text(), and MultiRunner.text_pt()), but instead wrap a SingleRunner instance, and provide an automatic (or manual by the MultiRunner.reset() function) restart of the interpreter as required.

class text.MultiRunner(cls, *args, **kwargs)

A restartable SingleRunner class

  • cls (SingleRunner class) – the class being wrapped
  • args (list) – args at class instantiation
  • kwargs (dict) – keyword args at at class instantiation
preamble(expr, texmessages=[])

resembles SingleRunner.preamble()

text_pt(*args, **kwargs)

resembles SingleRunner.text_pt()

text(*args, **kwargs)

resembles SingleRunner.text()


Start a new SingleRunner instance

Parameters:reinit (bool) – replay preamble() calls on the new instance

After executing this function further preamble calls are allowed, whereas once a text output has been created, preamble() calls are forbidden.

class text.TexRunner(*args, **kwargs)

A restartable SingleTexRunner class

  • args (list) – args at class instantiation
  • kwargs (dict) – keyword args at at class instantiation
class text.LatexRunner(*args, **kwargs)

A restartable SingleLatexRunner class

  • args (list) – args at class instantiation
  • kwargs (dict) – keyword args at at class instantiation
class text.textbox_pt(x_pt, y_pt, left_pt, right_pt, height_pt, depth_pt, do_finish, fontmap, singlecharmode, fillstyles)

Text output.

An instance of this class contains the text output generated by PyX. It is a baseclasses.canvasitem and thus can be inserted into a canvas.


Return the position of a marker.

Parameters:name (str) – name of the marker
Returns:position of the marker
Return type:tuple of two PyX lengths

This method returns the position of the marker of the given name within, previously defined by the \PyXMarker{name} macro in the typeset text. Please do not use the @ character within your name to prevent name clashes with PyX internal uses (although we don’t the marker feature internally right now).

Similar to generating actual output, the marker function accesses the DVI output, stopping. The TeX ipc mode feature will allow for this access without stopping the TeX interpreter.


left extent of the text (PyX length)


right extent of the text (PyX length)


width of the text (PyX length)


height of the text (PyX length)


height of the text (PyX length)

Module level functionality

The text module provides the public interface of the SingleRunner class by module level functions. For that, a module level MultiRunner is created and configured by the set() function. Each time the set() function is called, the existing module level MultiRunner is replaced by a new one.


the current MultiRunner instance for the module level functions


default_runner.preamble (bound method)


default_runner.text_pt (bound method)


default_runner.text (bound method)


default_runner.reset (bound method)

text.set(cls=TexRunner, mode=None, *args, **kwargs)

Setup a new module level MultiRunner

  • cls (MultiRunner object, not instance) – the module level MultiRunner to be used, i.e. TexRunner or LatexRunner
  • mode (str or None) –

    "tex" for TexRunner or "latex" for LatexRunner with arbitraty capitalization, overwriting the cls value

    deprecated:use the cls argument instead
  • args (list) – args at class instantiation
  • kwargs (dict) – keyword args at at class instantiation
text.escapestring(s, replace={" ": "~", "$": "\$", "&": "\&", " "_": "\_", "%": "\%", "^": "\string^", "~": "\string~", "<": "{$<$}", ">": "{$>$}", "{": "{${$}", "}": "{$}$}", "\": "{$setminus$}", "|": "{$mid$}"})

Escapes ASCII characters such that they can be typeset by TeX/LaTeX

TeX output parsers

While running TeX (and variants thereof) a variety of information is written to stdout like status messages, details about file access, and also warnings and errors. PyX reads all the output and analyses it. Some of the output is triggered as a direct response to the TeX input and is thus easy to understand for PyX. This includes page output information, but also workflow control information injected by PyX into the input stream. PyX uses it to check the communication and typeset progress. All the other output is handled by a list of texmessage parsers, an individual set of functions applied to the TeX output one after the other. Each of the function receives the TeX output as a string and return it back (maybe altered). Such a function must perform one of the following actions in response to the TeX output is receives:

  1. If it does not find any text in the TeX output it feals responsible for, it should just return the unchanged string.
  2. If it finds a text it is responsible for, and the message is just fine (doesn’t need to be communicated to the user), it should just remove this text and return the rest of the TeX output.
  3. If the text should be communicated to the user, a message should be written the the pyx logger at warning level, thus being reported to the user on stderr by default. Examples are underfull and overfull box warnings or font warnings. In addition, the text should be removed as in 2 above.
  4. In case of an error, TexResultError should be raised.

This is rather uncommon, that the fourth option is taken directly. Instead, errors can just be kept in the output as PyX considers unhandled TeX output left after applying all given texmessage parsers as an error. In addition to the error message, information about the TeX in- and output will be added to the exception description text by the SingleRunner according to the errordetail setting. The following verbosity levels are available:

class text.errordetail

Constants defining the verbosity of the TexResultError.

none = 0

Without any input and output.

default = 1

Input and parsed output shortend to 5 lines.

full = 2

Full input and unparsed as well as parsed output.

exception text.TexResultError

Error raised by texmessage parsers.

To prevent any unhandled TeX output to be reported as an error, texmessage.warn or texmessage.ignore can be used. To complete the description, here is a list of all available texmessage parsers:

class text.texmessage

Collection of TeX output parsers.

This class is not meant to be instanciated. Instead, it serves as a namespace for TeX output parsers, which are functions receiving a TeX output and returning parsed output.

In addition, this class also contains some generator functions (namely texmessage.no_file and texmessage.pattern), which return a function according to the given parameters. They are used to generate some of the parsers in this class and can be used to create others as well.

static start(msg)

Validate TeX/LaTeX startup message including scrollmode test.

>>> texmessage.start(r'''
... This is e-TeX (version)
... *! Undefined control sequence.
... <*> \raiseerror
...                %
... ''')
static no_file(fileending, qualname=None)

Generator function to ignore the missing file message for fileending.

static no_aux(msg)

Ignore the missing aux file message.

static no_nav(msg)

Ignore the missing nav file message.

static end(msg)

Validate TeX shutdown message.

static load(msg)

Ignore file loading messages.

Removes text starting with a round bracket followed by a filename ignoring all further text until the corresponding closing bracket. Quotes and/or line breaks in the filename are handled as needed for TeX output.

Without quoting the filename, the necessary removal of line breaks is not well defined and the different possibilities are tested to check whether one solution is ok. The last of the examples below checks this behavior.

>>> texmessage.load(r'''other (text.py) things''')
'other  things'
>>> texmessage.load(r'''other ("text.py") things''')
'other  things'
>>> texmessage.load(r'''other ("tex
... t.py" further (ignored)
... text) things''')
'other  things'
>>> texmessage.load(r'''other (t
... ext
... .py
... fur
... ther (ignored) text) things''')
'other  things'
static load_def(msg)

Ignore font definition (*.fd and *.def) loading messages.

static load_graphics(msg)

Ignore graphics file (*.eps) loading messages.

static ignore(msg)

Ignore all messages.

Should be used as a last resort only. You should write a proper TeX output parser function for the output you observe.

static warn(msg)

Warn about all messages.

Similar to ignore, but writing a warning to the logger about the TeX output. This is considered to be better when you need to get it working quickly as you will still be prompted about the unresolved output, while the processing continues.

static pattern(p, warning, qualname=None)

Warn by regular expression pattern matching.

static box_warning(msg)

Warn about overfull/underfull box.

static font_warning(msg)

Warn about font substitutions of NFSS.

static package_warning(msg)

Warn about generic package messages.

static rerun_warning(msg)

Warn about rerun required message.

static nobbl_warning(msg)

Warn about no-bbl message.

TeX/LaTeX attributes

TeX/LaTeX attributes are instances to be passed to a texrunners text() method. They stand for TeX/LaTeX expression fragments and handle dependencies by proper ordering.

class text.halign(boxhalign, flushhalign)

Instances of this class set the horizontal alignment of a text box and the contents of a text box to be left, center and right for boxhalign and flushhalign being 0, 0.5, and 1. Other values are allowed as well, although such an alignment seems quite unusual.

Note that there are two separate classes boxhalign and flushhalign to set the alignment of the box and its contents independently, but those helper classes can’t be cleared independently from each other. Some handy instances available as class members:


Left alignment of the text box, i.e. sets boxhalign to 0 and doesn’t set flushhalign.


Center alignment of the text box, i.e. sets boxhalign to 0.5 and doesn’t set flushhalign.


Right alignment of the text box, i.e. sets boxhalign to 1 and doesn’t set flushhalign.


Left alignment of the content of the text box in a multiline box, i.e. sets flushhalign to 0 and doesn’t set boxhalign.


Identical to flushleft.


Center alignment of the content of the text box in a multiline box, i.e. sets flushhalign to 0.5 and doesn’t set boxhalign.


Identical to flushcenter.


Right alignment of the content of the text box in a multiline box, i.e. sets flushhalign to 1 and doesn’t set boxhalign.


Identical to flushright.


Combines boxleft and flushleft, i.e. halign(0, 0).


Combines boxcenter and flushcenter, i.e. halign(0.5, 0.5).


Combines boxright and flushright, i.e. halign(1, 1).


valign example

class text.valign(valign)

Instances of this class set the vertical alignment of a text box to be top, center and bottom for valign being 0, 0.5, and 1. Other values are allowed as well, although such an alignment seems quite unusual. See the left side of figure valign example for an example.

Some handy instances available as class members:








Identical to clearing the vertical alignment by clear to emphasise that a baseline alignment is not a box-related alignment. Baseline alignment is the default, i.e. no valign is set by default.

class text.parbox(width, baseline=top)

Instances of this class create a box with a finite width, where the typesetter creates multiple lines in. Note, that you can’t create multiple lines in TeX/LaTeX without specifying a box width. Since PyX doesn’t know a box width, it uses TeXs LR-mode by default, which will always put everything into a single line. Since in a vertical box there are several baselines, you can specify the baseline to be used by the optional baseline argument. You can set it to the symbolic names top, parbox.middle, and parbox.bottom only, which are members of valign. See the right side of figure valign example for an example.

Since you need to specify a box width no predefined instances are available as class members.

class text.vshift(lowerratio, heightstr="0")

Instances of this class lower the output by lowerratio of the height of the string heigthstring. Note, that you can apply several shifts to sum up the shift result. However, there is still a clear class member to remove all vertical shifts.

Some handy instances available as class members:


vshift(0) (this doesn’t shift at all)






This is a special vertical shift to lower the output by the height of the mathematical axis. The mathematical axis is used by TeX for the vertical alignment in mathematical expressions and is often usefull for vertical alignment. The corresponding vertical shift is less than middlezero and usually fits the height of the minus sign. (It is the height of the minus sign in mathematical mode, since that’s that the mathematical axis is all about.)

There is a TeX/LaTeX attribute to switch to TeXs math mode. The appropriate instances mathmode and clearmathmode (to clear the math mode attribute) are available at module level.


Enables TeXs mathematical mode in display style.

The size class creates TeX/LaTeX attributes for changing the font size.

class text.size(sizeindex=None, sizename=None, sizelist=defaultsizelist)

LaTeX knows several commands to change the font size. The command names are stored in the sizelist, which defaults to ["normalsize", "large", "Large", "LARGE", "huge", "Huge", None, "tiny", "scriptsize", "footnotesize", "small"].

You can either provide an index sizeindex to access an item in sizelist or set the command name by sizename.

Instances for the LaTeXs default size change commands are available as class members:





















There is a TeX/LaTeX attribute to create empty text boxes with the size of the material passed in. The appropriate instances phantom and clearphantom (to clear the phantom attribute) are available at module level.


Skip the text in the box, but keep its size.

Using the graphics-bundle with LaTeX

The packages in the LaTeX graphics bundle (color.sty, graphics.sty, graphicx.sty, ...) make extensive use of \\special commands. PyX defines a clean set of such commands to fit the needs of the LaTeX graphics bundle. This is done via the pyx.def driver file, which tells the graphics bundle about the syntax of the \\special commands as expected by PyX. You can install the driver file pyx.def into your LaTeX search path and add the content of both files color.cfg and graphics.cfg to your personal configuration files[3]. After you have installed the cfg files, please use the text module with unset pyxgraphics keyword argument which will switch off a convenience hack for less experienced LaTeX users. You can then import the LaTeX graphics bundle packages and related packages (e.g. rotating, ...) with the option pyx, e.g. \\usepackage[pyx]{color,graphicx}. Note that the option pyx is only available with unset pyxgraphics keyword argument and a properly installed driver file. Otherwise, omit the specification of a driver when loading the packages.

When you define colors in LaTeX via one of the color models gray, cmyk, rgb, RGB, hsb, then PyX will use the corresponding values (one to four real numbers). In case you use any of the named colors in LaTeX, PyX will use the corresponding predefined color (see module color and the color table at the end of the manual). The additional LaTeX color model pyx allows to use a PyX color expression, such as color.cmyk(0,0,0,0) directly in LaTeX. It is passed to PyX.

When importing Encapsulated PostScript files (eps files) PyX will rotate, scale and clip your file like you expect it. Other graphic formats can not be imported via the graphics package at the moment.

For reference purpose, the following specials can be handled by PyX at the moment:

PyX:color_begin (model) (spec)
starts a color. (model) is one of gray, cmyk, rgb, hsb, texnamed, or pyxcolor. (spec) depends on the model: a name or some numbers
ends a color.
PyX:epsinclude file= llx= lly= urx= ury= width= height= clip=0/1
includes an Encapsulated PostScript file (eps files). The values of llx to ury are in the files’ coordinate system and specify the part of the graphics that should become the specified width and height in the outcome. The graphics may be clipped. The last three parameters are optional.
PyX:scale_begin (x) (y)
begins scaling from the current point.
ends scaling.
PyX:rotate_begin (angle)
begins rotation around the current point.
ends rotation.


While the PyX configuration technically has nothing to do with the text module, we mention it here as part of the text module since its main purpose is the configuration of various aspects related to the typesetting of text.

PyX comes with reasonable defaults which should work out of the box on most TeX installations. The default values are defined in the PyX source code itself and are repeated in the system-wide config file in INI file format located at pyx/data/pyxrc. This file also contains a description of each of the listed config values and is read at PyX startup. Thus the system-wide configuration can be adjusted by editing this file.

In addition, a user-specific configuration can be setup by a ~/.pyxrc on unix-like Systems (including OS X) or pyxrc in the directory defined by the environment variable APPDATA on MS Windows. This user-specific configuration will overwrite the system-wide settings.

TeX ipc mode

For output generation of typeset text and to calculate the positions of markers (see textbox_pt.marker()) the DVI output of the TeX interpreter must be read. In contrast, the text extent (textbox_pt.left, textbox_pt.right, textbox_pt.width, textbox_pt.height, textbox_pt.depth) is available without accessing the DVI output, as the TeX interpreter is instructed by PyX to output it to stdout, which is read and analysed at the typesetting step immediately.

Since TeX interpreters usually buffer the DVI output, the interpreter itself needs to be terminated to get the DVI output. As MultiRunner instances can start a new interpreter when needed, this does not harm the functionality and happens more or less unnoticable. Still it generates some penalty in terms of execution speed, which can become huge for certain situations (alternation between typesetting and marker access).

One of the effects of the texipc option available in almost all present TeX interpreters is to flush the DVI output after each page. As PyX reads the DVI output linearily, it can successfully read all output whithout stopping the TeX interpreter. It is suggested to enable the texipc feature in the system-wide configuration if available in the TeX interpreter being used.


PyX provides various functionality to collect details about the typesetting process. First off all, PyX reads the output generated by the TeX interpreter while it processes the text provided by the user. If the given texmessage parsers do not validate this output, an TexResultError is raised immediately. The verbosity of this output can be adjusted by the errordetail setting of the SingleRunner. This might help in some cases to identify an error in the text passed for typesetting, but for more complicated problems, other help is required.

One possibility is to output the actual code passed to the TeX interpreter. For that you can pass a file name or a file handle to the copyinput argument of the SingleRunner. You can then process the text by the TeX interpreter yourself to reproduce the issue outside of PyX.

Similarily you can also save the log output from the TeX interpreter. For that you need to pass a log file name (with the suffix .log) in the usefiles argument (which is a list of files) of the SingleRunner. This list of files are saved and restored in the temporary directory used by the TeX interpreter. While originally it is meant to share, for example, a .aux file between several runs (for which the temporary directory is different and removed after each run), it can do the same for the .log file (where the restore feature is needless, but does not harm). PyX takes care of the proper \jobname, hence you can choose the filename arbitrarily with the exception of the suffix, as the suffix is kept during the save and restore.

Still, all this might not help to fully understand the problem you’re facing. For example there might be situations, where it is not clear which TeX interpreter is actually used (when several executables are available and the path setup within the Python interpreter differs from the one used on the shell). In those situations it might help to enable some additional logging output created by PyX. PyX uses the logging module from the standard library and logs to a logger named "pyx". By default, various information about executing external programms and locating files will not be echoed, as it is written at info level, but PyX provides a simple convenience function to enable the output of this logging level. Just call the pyxinfo() function defined on the PyX package before actually start using the package in your Python program:


Make PyX a little verbose (for information or debugging)

This function enables info level on the "pyx" logger. It also adds some general information about the Python interpreter, the PyX installation, and the PyX configuration to the logger.


[3]If you do not know what this is all about, you can just ignore this paragraph. But be sure that the pyxgraphics keyword argument is always set!