The relationship between Ghostscript and PostScript

Table of contents

For other information, see the Ghostscript overview.

Ghostscript's capabilities in relation to PostScript

The Ghostscript interpreter, except as noted below, is intended to execute properly any source program written in the (Level 2) PostScript language as defined in the December 1990 printing of the PostScript Language Reference Manual (Second Edition, ISBN 0-201-18127-4) published by Addison-Wesley. However, the interpreter is configurable in ways that can restrict it to various subsets of this language. Specifically, the base interpreter accepts the Level 1 subset of the PostScript language, as defined in the first edition of the PostScript Language Reference Manual, ISBN 0-201-10174-2, Addison-Wesley, 1985, plus the file system, version 25.0 language, and miscellaneous additions listed in sections A.1.6, A.1.7, and A.1.8 of the Second Edition respectively, including allowing a string operand for the "status" operator. The base interpreter may be configured (see the documentation on building Ghostscript for how to configure it) by adding any combination of the following:

Adding all of these produces a full Level 2 PostScript language interpreter.

Ghostscript can also interpret files in the Portable Document Format (PDF) 1.2 format defined in the Portable Document Format Reference Manual Version 1.2 of November 12, 1996, distributed by Adobe Systems Incorporated. This facility is available only if the "pdf" feature was selected when Ghostscript was built.

Ghostscript also includes a number of additional operators defined below that are not in the PostScript language defined by Adobe.

Implementation limits

The implementation limits show here correspond to those in Tables B.1 and B.2 of the Second Edition. Asterisks (*) mark the ones different from the limits in the Second Edition.

Architectural limits

integer       32-bit twos complement integer
real     single-precision IEEE float
array   * 8191 elements (16-bit systems)
      65535 elements (32-bit systems)
dictionary   * 8190 elements (16-bit systems)
      65534 elements (32-bit systems)
string   * 65535 characters
name   * 16383 characters
filename     100 characters
save level   * none (capacity of memory)
gsave level   * none (capacity of memory)

Typical memory limits in Level 1

userdict       200
FontDirectory     100
operand stack   * 800
dictionary stack     20
execution stack     250
interpreter level   * none (capacity of memory)
path   * none (capacity of memory)
dash     11
VM   * capacity of memory
file   * determined by operating system
image   * 65535 values (samples × components) for 1-, 2-, 4-, or 8-bit samples
      32767 values for 12-bit samples

Other differences in VM consumption

Packed array elements occupy either 2 bytes or 8 bytes. The average element size is probably about 5 bytes. Names occupy 12 bytes plus the space for the string.

Additional operators in Ghostscript

Graphics and text operators

Graphics state operators

<bool> .setaccuratecurves -
Sets a graphics state flag that determines whether curves and arcs, when flattened, always start and end with a line that is a segment of the tangent; this also causes butt and square caps to be properly perpendicular to the tangent. initgraphics sets this flag to false, to match other PostScript implementations.
- .currentaccuratecurves <bool>
Returns the current value of the accurate curves flag.
<bool> .setclipoutside -
Sets a graphics state flag that determines whether the effective clipping region is the inside (false) or the outside (true) of the region defined by the clipping operators (clip, rectclip, etc.) initclip sets this flag to false.
- .currentclipoutside <bool>
Returns the current value of the outside clipping flag.
<bool> .setdashadapt -
Sets a graphics state flag that determines whether dash patterns do (true) or do not (false) automatically scale themselves so that each line segment consists of an integral number of pattern repetitions. initgraphics sets this flag to false.
- .currentdashadapt <bool>
Returns the current value of the dash adaptation flag.
<matrix> .setdefaultmatrix -
Sets the default matrix that is returned by defaultmatrix and installed by initmatrix. Ordinary programs should not use this operator.
<num> <bool> .setdotlength -
Sets a graphics state parameter that determines the handling of zero-length lines (dots). If the dot length is zero, dots are painted as circles if round line caps are in effect, otherwise they are not painted at all. If the dot length is non-zero, dots are treated exactly like lines of the given length: the length is specified in user coordinates (like line width) if bool is false, or in default user coordinates of points (units of 1/72in; see the notes on measurements in the documentation on devices) if bool is true. Dots occurring as part of dash patterns will be oriented correctly; isolated dots will be oriented as though they were part of a vertical line. initgraphics sets the dot length to zero.
- .currentdotlength <num> <bool>
Returns the current dot length and dot length mode.
<dx> <dy> .setfilladjust2 -
Sets graphics state parameters that cause all filled and stroked regions to be "fattened" by the given amount relative to an algorithm that only paints pixels whose centers fall within the region to be painted. dx and dy are numbers between 0 and 0.5, measured in device space. The only two values that are likely to be useful are 0, which gives a pure center-of-pixel rule, and 0.5, which gives Adobe's any-part-of-pixel rule. (0.5 is treated slightly specially in order to create half-open pixels per Adobe's specification.)
- .currentfilladjust2 <dx> <dy>
Returns the current fill adjustment values.

Path operators

- .dashpath -
If there is no current dash pattern, does nothing. Otherwise, does the equivalent of flattenpath and then chops up the path as determined by the dash pattern.
<x> <y> <width> <height> .rectappend -
<numarray> .rectappend -
<numstring> .rectappend -
Appends a rectangle or rectangles to the current path, in the same manner as rectfill, rectclip, etc. Defined only if the dps or level2 option was selected when Ghostscript was built.

Painting operators

Ghostscript supports an experimental extension of the PostScript imaging model to include RasterOp and some related facilities. This extension is available only if the rasterop option was selected when building Ghostscript.

With the RasterOp extension, imaging operations compute a function D = f(D,S,T) in RGB space, where f is an arbitrary 3-input Boolean function, D is the destination (frame buffer or print buffer), S is the source (described below), and T is the texture (the current PostScript color, which may be a pattern). The source and texture depend on the PostScript imaging operation:

The rasterop option adds the following operators:

<int8> .setrasterop -
Sets the RasterOp function in the graphics state. The default function is 252, Source | Texture.
- .currentrasterop <int8>
Returns the current RasterOp function.
<bool> .setsourcetransparent -
Sets source transparency in the graphics state. When source transparency is true, white source pixels prevent storing into the destination, regardless of what the RasterOp function returns. The default source transparency is false.
- .currentsourcetransparent <bool> -
Returns the current source transparency.
<bool> .settexturetransparent -
Sets texture transparency in the graphics state. When texture transparency is true, white texture pixels prevent storing into the destination, regardless of what the RasterOp function returns. The default texture transparency is false.
- .currenttexturetransparent <bool> -
Returns the current texture transparency.

For more information on RasterOp and transparency, please consult chapter 5 of the "PCL 5 Color Technical Reference Manual", Hewlett-Packard Manual Part No. 5961-0635.

Character operators

<string> <bool> .charboxpath -
For each character C in the rendering of <string>, let the bounding box of C in device space be the four user-space points p1x/y, p2x/y, p3x/y, and p4x/y. For each character in order, .charboxpath appends the following to the current path:

This creates a path whose pathbbox is the bbox of the string.

If the CTM is well-behaved (consists only of reflection, scaling, and rotation by multiples of 90 degrees), this too creates a (simpler) path whose pathbbox is the bbox of the string.

<font> <charname|charcode> <charname> <charstring> .type1execchar -
Does all the work for rendering a Type 1 outline. This operator, like setcharwidth and setcachedevice, is valid only in the context of a show operator -- that is, it must only be called from within a BuildChar or BuildGlyph procedure.
<font> <charcode> %Type1BuildChar -
This is not a new operator: rather, it is a name known specially to the interpreter. Whenever the interpreter needs to render a character (during a, stringwidth, or charpath), it looks up the name BuildChar in the font dictionary to find a procedure to run. If it does not find this name, and if the FontType is 1, the interpreter instead uses the value (looked up on the dictionary stack in the usual way) of the name %Type1BuildChar.

The standard definition of %Type1BuildChar is in the initialization file Users should not need to redefine %Type1BuildChar, except perhaps for tracing or debugging.

<font> <charname> %Type1BuildGlyph -
Provides the Type 1 implementation of BuildGlyph.

Other operators

Mathematical operators

<number> arccos <number>
Computes the arc cosine of a number between -1 and 1.
<number> arcsin <number>
Computes the arc sine of a number between -1 and 1.

String operators

<state> <fromString> <toString> .type1encrypt <newState> <toSubstring>
Encrypts fromString according to the algorithm for Adobe Type 1 fonts, writing the result into toString. toString must be at least as long as fromString, or a rangecheck error occurs. state is the initial state of the encryption algorithm (a 16-bit non-negative integer); newState is the new state of the algorithm.
<state> <fromString> <toString> .type1decrypt <newState> <toSubstring>
Decrypts fromString according to the algorithm for Adobe Type 1 fonts, writing the result into toString. Other specifications are as for type1encrypt.

Relational operators

<number|string> <number|string> max <number|string>
Returns the larger of two numbers or strings.
<number|string> <number|string> min <number|string>
Returns the smaller of two numbers or strings.

File operators

<string> findlibfile <foundstring> <file> true
<string> findlibfile <string> false
Opens the file of the given name for reading, searching through directories as described in the usage documentation. If the search fails, findlibfile simply pushes false on the stack and returns, rather than causing an error.
<file> <integer> unread -
Pushes back the last-read character onto the front of the file. If the file is open only for writing, or if the integer argument is not the same as the last character read from the file, causes an ioerror error. May also cause an ioerror if the last operation on the file was not a reading operation.
<file> <device> writeppmfile -
Writes the contents of the device, which must be an image device, onto the file, in Portable PixMap (ppm) format. Does not close the file.

Ghostscript also supports the following IODevice in addition to a subset of those defined in the Adobe documentation: %pipe%command, which opens a pipe on the given command. This is supported only on operating systems that provide popen (primarily Unix systems, and not all of those).


Ghostscript supports all the standard PostScript Level 2 filters, except that it does not currently support the EarlyChange key in the LZWEncode filter. Ghostscript also supports the as yet undocumented FlateEncode and FlateDecode filters from PDF 1.2 and (presumably) PostScript Level 3, except for the Effort key in FlateEncode. In addition, Ghostscript supports the following non-standard filters:

<target> /BCPEncode filter <file>
<source> /BCPDecode filter <file>
Create filters that implement the Adobe Binary Communications Protocol. See Adobe documentation for details.
<target> <seed_integer> /eexecEncode filter <file>
Creates a filter for encrypting data into the eexec encrypted format described in the Adobe Type 1 Font Format documentation. The seed_integer must be 55665 for proper operation. This filter produces binary output and does not include the initial 4 (or lenIV) garbage bytes.
<source> <seed_integer> /eexecDecode filter <file>
<source> <dict> /eexecDecode filter <file>
Creates a filter for decrypting data encrypted using eexec encryption as described in the Adobe Type 1 Font Format documentation. The seed_integer must be 55665 for proper operation. Recognized dictionary keys are:
seed <16-bit integer> (required)
lenIV <non-negative integer> (default=4)
<source> /PCXDecode filter <file>
Creates a filter that decodes data in the run-length encoding used in the PCX graphics file format.
<source> <hex_boolean> /PFBDecode filter <file>
Creates a filter that decodes data in .PFB format, the usual semi-binary representation for Type 1 font files on IBM PC and compatible systems. If hex_boolean is true, binary packets are converted to hex; if false, binary packets are not converted.
<target> <dict> /PixelDifferenceEncode filter <file>
<source> <dict> /PixelDifferenceDecode filter <file>
Implements the Predictor=2 pixel-differencing option of the LZW filters. Recognized keys are:
Colors <integer> (1 to 4, default=1)
BitsPerComponent <integer> (1, 2, 4, or 8, default=8)
Columns <integer> (>= 0, required)

See the Adobe Portable Document Format Reference Manual for details.

<target> <dict> /PNGPredictorEncode filter <file>
<source> <dict> /PNGPredictorDecode filter <file>
Implements the "filter" algorithms of the Portable Network Graphics (PNG) graphics format. Recognized keys are:

Keys recognized in PNG filter algorithms
Key    Range    Default

Colors <integer>     1 to 16    16
BitsPerComponent <integer>   1, 2, 4, 8, or 16    8
Columns <integer>   >= 0    1
Predictor <integer>   10 to 15    15

The Predictor is the PNG algorithm number + 10 for the Encoding filter; the Decoding filter ignores Predictor. 15 means the encoder attempts to optimize the choice of algorithm. For more details see the PNG specification
<target> /TBCPEncode filter <file>
<source> /TBCPDecode filter <file>
Create filters that implement the Adobe Tagged Binary Communications Protocol. See Adobe documentation for details.
<target> /zlibEncode filter <file>
<source> /zlibDecode filter <file>
Creates filters that use the zlib data compression method (the same method used by the gzip application). This filter is available only if the fzlib feature was selected when Ghostscript was built.

Some versions of Ghostscript may also support other non-standard filters for experimental purposes. The current version includes the following non-standard filters, which are not documented further. No guarantee is made that these filters will exist in compatible form, or at all, in future versions.

<target/source> <string> ByteTranslateEncode/Decode <target> <int> BigStringEncode <target/source> <dict> BoundedHuffmanEncode/Decode
FirstBitLowOrder <bool> false MaxCodeLength <int> 16 EndOfData <bool> true EncodeZeroRuns <int> 256 Tables <int_array>
<target/source> <dict> BWBlockSortEncode/Decode
BlockSize <int> 16384
<target/source> MoveToFrontEncode/Decode

Ghostscript also supports additional keys in the optional dictionary operands for some filters. For the LZWDecode filter:

InitialCodeLength <integer> (default 8)
An integer between 2 and 11 specifying the initial number of data bits per code. Note that the actual initial code length is 1 greater than this, to allow for the reset and end-of-data code values.
FirstBitLowOrder <boolean> (default false)
If true, codes appear with their low-order bit first.
BlockData <boolean> (default false)
If true, the data is broken into blocks in the manner specified for the GIF file format.

For the CCITTFaxEncode and CCITTFaxDecode filters:

DecodedByteAlign <integer> (default 1)
An integer N with the value 1, 2, 4, 8, or 16, specifying that decoded data scan lines are always a multiple of N bytes. The encoding filter skips data in each scan line from Columns to the next multiple of N bytes; the decoding filter pads each scan line to a multiple of N bytes.

Virtual memory operators

<save> .forgetsave -
Cancels the effect of a save, making it as though the save never happened.

Miscellaneous operators

<string> getenv <string> true
<string> getenv false
Looks up a name in the shell environment. If the name is found, returns the corresponding value and true; if the name is not found, returns false.
<name> <array> .makeoperator <operator>
Constructs and returns a new operator that is actually the given procedure in disguise. The name is only used for printing. The operator has the executable attribute.

Operators defined in this way do one other thing besides running the procedure: if an error occurs during the execution of the procedure, and there has been no net reduction in operand or dictionary stack depth, the operand or dictionary stack pointer respectively is reset to its position at the beginning of the procedure.

<string> <boolean> .setdebug -
If the Ghostscript interpreter was built with the DEBUG flag set, sets or resets any subset of the debugging flags normally controlled by -Z in the command line. Has no effect otherwise.
- .oserrno <errno>
Returns the error code for the most recent operating system error.
- .oserror <string>
Returns the error string for the most recent operating system error.

Device operators

<device> copydevice <device>
Copies a device. The copy is writable and installable.
<index> .getdevice <device>
Returns a device from the set of devices known to the system. The first device, which is the default, is numbered 0. If the index is out of range, causes a rangecheck error. This device is actually a prototype, not a directly usable device, and is marked read-only; it cannot have its parameters changed or be installed as the current device.
<matrix> <width> <height> <palette> makeimagedevice <device>
Makes a new device that accumulates an image in memory. matrix is the initial transformation matrix: it must be orthogonal (that is, [a 0 0 b x y] or [0 a b 0 x y]). palette is a string of 2^N or 3 × 2^N elements, specifying how the 2^N possible pixel values will be interpreted. Each element is interpreted as a gray value, or as RGB values, multiplied by 255. For example, if you want a monochrome image for which 0=white and 1=black, the palette should be <ff 00>; if you want a 3-bit deep image with just the primary colors and their complements (ignoring the fact that 3-bit images are not supported), the palette might be <000000 0000ff 00ff00 00ffff ff0000 ff00ff ffff00 ffffff>. At present, the palette must contain exactly 2, 4, 16, or 256 entries, and must contain an entry for black and an entry for white; if it contains any entries that aren't black, white, or gray, it must contain at least the six primary colors (red, green, blue, and their complements cyan, magenta, and yellow); aside from this, its contents are arbitrary.

Alternatively, palette can be 16, 24, 32, or null (equivalent to 24). These are interpreted as:

Palette     Bits allocated per color

16   5 red, 6 green, 5 blue
24   8 red, 8 green, 8 blue
32   8C, 8M, 8Y, 8K

Note that one can also make an image device (with the same palette as an existing image device) by copying a device using the copydevice operator.

<matrix> <width> <height> <palette> <word?> makewordimagedevice <device>
Makes an image device as described above. word? is a Boolean value indicating whether the data should be stored in a word-oriented format internally. No ordinary PostScript programs should use this operator.
<device> <index> <string> copyscanlines <substring>
Copies one or more scan lines from an image device into a string, starting at a given scan line in the image. The data is in the same format as for the image operator. It is an error if the device is not an image device or if the string is too small to hold at least one complete scan line. Always copies an integral number of scan lines.
<device> setdevice -
Sets the current device to the specified device. Also resets the transformation and clipping path to the initial values for the device. Signals an invalidaccess error if the device is a prototype.
- currentdevice <device>
Gets the current device from the graphics state.
<device> getdeviceprops <mark> <name1> <value1> ... <namen> <valuen>
Gets the properties of a device. See the section on device parameters below for details.
<mark> <name1> <value1> ... <namen> <valuen> <device> putdeviceprops <device>
Sets properties of a device. May cause undefined, typecheck, rangecheck, or limitcheck errors.
- flushpage -
On displays, flushes any buffered output, so that it is guaranteed to show up on the screen; on printers, has no effect.

Device parameters

Ghostscript supports the concept of device parameters for all devices, not just page devices. (For non-page devices, these are accessible through getdeviceprops and putdeviceprops, as indicated above.) Here are the currently defined parameters for all devices:
BitsPerPixel <integer> (usually read-only)
Number of bits per pixel.
.HWMargins [<four floats>]
Size of non-imageable regions around the edges of the page, in points (units of 1/72in; see the notes on measurements in the documentation on devices).
HWSize [<integer> <integer>]
X and Y size in pixels.
Name <string> (read-only)
The device name. Currently the same as OutputDevice.
Colors, GrayValues, RedValues, GreenValues, BlueValues, ColorValues (usually read-only)
As for the "deviceinfo" operator of Display PostScript. Red, Green, Blue, and ColorValues are only defined if Colors > 1.
TextAlphaBits, GraphicsAlphaBits (usually read-only)
The number of bits of anti-aliasing information for text or graphics respectively. Legal values are 1 (no anti-aliasing, the default for most devices), 2, or 4.

In addition, the following are defined per Adobe's documentation for the setpagedevice operator:

Duplex (if supported)
NumCopies (for printers only)
Orientation (if supported)
PageOffset (write-only)

Some devices may only allow certain values for HWResolution and PageSize. The null device ignores attempts to set PageSize; its size is always [0 0].

For printers these are also defined:

BufferSpace <integer>
Buffer space for band lists, if the bitmap is too big to fit in memory.
MaxBitmap <integer>
Maximum space for a full bitmap in memory.
OutputFile <string>
() means "send to printer directly", otherwise specifies the file name for output; "%d" is replaced by the page number; on Unix systems "(|command)" writes to a pipe.
OpenOutputFile <boolean>
If true, open the device's output file when the device is opened, rather than waiting until the first page is ready to print.
PageCount <integer> (read-only)
Counts the number of pages printed on the device.

The following parameters are for use only by very specialized applications that separate band construction from band rasterization. Improper use may cause unpredictable errors. In particular, if you only want to allocate more memory for banding, to increase band size and improve performance, use the BufferSpace parameter, not BandBufferSpace.

BandHeight <integer>
The height of bands when banding. 0 means use the largest band height that will fit within the BandBufferSpace (or BufferSpace, if BandBufferSpace is not specified).
BandWidth <integer>
The width of bands in the rasterizing pass, in pixels. 0 means use the actual page width.
BandBufferSpace <integer>
The size of the band buffer in the rasterizing pass, in bytes. 0 means use the same buffer size as for the interpretation pass.

In addition, Ghostscript supports the following parameter for setpagedevice and currentpagedevice that is not a device parameter per se:

ViewerPreProcess procedure
Specifies a procedure to be applied to the page device dictionary before any other processing is done. The procedure may not alter the dictionary, but it may return a modified copy. This "hook" is provided for use by viewing programs such as GSview.

Miscellaneous additions

run can take either a string or a file as its argument. In the latter case, it just runs the file, closing it at the end, and trapping errors just as for the string case.

Copyright © 1996, 1997, 1998 Aladdin Enterprises. All rights reserved.

This file is part of GNU Ghostscript. See the GNU General Public License (the "License") for full details of the terms of using, copying, modifying, and redistributing GNU Ghostscript.

Ghostscript version 5.50, 16 September 1998