PostScript binary encoding

LanguageLevel 2 added two binary encodings of the language: binary tokens, which can be mixed with ordinary text, and binary object sequences. Both are introduced by a byte from 128 to 159, which cannot start a text token. Ordinary .ps files rarely use them, because they are not safe over 7-bit channels, but a reader must recognise them. This page follows the PostScript Language Reference, section 3.14.

Binary token types

Byte	Token	Bytes that follow
128–131	Binary object sequence	See below. 128: high-order byte first, IEEE reals; 129: low-order first, IEEE reals; 130: high-order first, native reals; 131: low-order first, native reals.
132	32-bit integer	4 bytes, high-order first.
133	32-bit integer	4 bytes, low-order first.
134	16-bit integer	2 bytes, high-order first, signed.
135	16-bit integer	2 bytes, low-order first, signed.
136	8-bit integer	1 byte, signed.
137	Fixed-point number	A representation byte `r` (see number representations), then the number.
138	IEEE real	4 bytes, single precision, high-order first.
139	IEEE real	4 bytes, low-order first.
140	Native real	4 bytes in the machine’s own format.
141	Boolean	1 byte: 0 false, 1 true.
142	String	1 byte length `n`, then `n` bytes.
143	Long string	2 bytes length, high-order first, then the bytes.
144	Long string	2 bytes length, low-order first, then the bytes.
145	Literal system name	1 byte index into the system name table.
146	Executable system name	1 byte index.
147–148	Reserved	Display PostScript extension (user names); not part of the standard language.
149	Homogeneous number array	A representation byte, a 2-byte count, then the numbers: see below.
150–159	Unassigned	A `syntaxerror` if met.

Number representations

Used by token 137 and homogeneous number arrays (tables 3.25 and 3.26). The representation byte r; values 48 and 49 (and 176, 177) apply only to homogeneous number arrays:

`r`	Numbers are
0–31	32-bit fixed point, high-order first, with `r` bits after the binary point (0: plain integers).
32–47	16-bit fixed point, high-order first, with `r` − 32 fraction bits.
48	32-bit IEEE reals, high-order first.
49	32-bit native reals.
`r` + 128	The same, low-order byte first.

A fixed-point value whose fraction bits are 0 is an integer object; otherwise it becomes a real.

Homogeneous number arrays

Token 149, then the representation byte, then a 2-byte count of numbers (in the byte order the representation names), then the numbers, all in that representation. The interpreter reads it as a literal array of numbers. Operators that take long lists of coordinates, such as ufill and xyshow, also accept these arrays as strings, an encoded number string.

Binary object sequences

A complete set of objects, usually a whole procedure, in one block that the interpreter can take in without scanning text.

Offset	Size	Field	Description
`0`	1	token type	128 to 131: byte order and real format, as above.
`1`	1	top-level count	1 to 255: the number of top-level objects. 0 means an extended header follows.
`2`	2	overall length	Normal header: the length of the whole sequence in bytes, header included.
Extended header (when byte 1 is 0):
`2`	2	top-level count	The number of top-level objects.
`4`	4	overall length	The length of the whole sequence.

Then come the top-level objects, 8 bytes each, then the objects of any arrays they hold, then the string and name text they point to. The interpreter executes the top-level objects as if they were a procedure.

Byte	Field	Description
`0`	type	Bits 0–6: the object type (below). Bit 7: executable.
`1`	tag	Must be 0 when a program is read, except that the top-level object of a sequence written by `printobject` or `writeobject` carries their one-byte tag here (§3.14.2, §3.14.6).
`2`–`3`	length	Depends on the type.
`4`–`7`	value	Depends on the type.

Type	Object	length	value
0	null	unused	unused
1	integer	unused	the integer
2	real	0 for a real in the sequence’s format; otherwise a scale `s`, the value being a fixed-point number with `s` fraction bits	the number
3	name	A signed 16-bit number: 0, the value is a system name index; −1 (0xFFFF), reserved for Display PostScript; greater than 0, the length of the name text	the index, or the offset of the text from the start of the object list
4	boolean	unused	0 false, 1 true
5	string	the length in bytes	the offset of the text from the start of the object list
6	immediately evaluated name	as name	as name
9	array	the number of elements	the offset of the first element’s 8-byte record
10	mark	unused	unused

Types 7 and 8 are not used. Offsets are counted from the first byte after the header. Bit 7 of the type byte makes names executable (moveto rather than /moveto) and arrays procedures.

Name tables

The system name table is a fixed list in which each index stands for a common name, so that a frequent name costs one byte (tokens 145 and 146 reach indexes 0–255; binary object sequences reach all of them). An index with no name is an undefined error, reported as systemn. The table, from appendix F of the reference; each row starts at the index in the first column and continues one by one:

From	Names
0	`abs add aload anchorsearch and arc arcn arct arcto array`
10	`ashow astore awidthshow begin bind bitshift ceiling charpath clear cleartomark`
20	`clip clippath closepath concat concatmatrix copy count counttomark currentcmykcolor currentdash`
30	`currentdict currentfile currentfont currentgray currentgstate currenthsbcolor currentlinecap currentlinejoin currentlinewidth currentmatrix`
40	`currentpoint currentrgbcolor currentshared curveto cvi cvlit cvn cvr cvrs cvs`
50	`cvx def defineusername dict div dtransform dup end eoclip eofill`
60	`eoviewclip eq exch exec exit file fill findfont flattenpath floor`
70	`flush flushfile for forall ge get getinterval grestore gsave gstate`
80	`gt identmatrix idiv idtransform if ifelse image imagemask index ineofill`
90	`infill initviewclip inueofill inufill invertmatrix itransform known le length lineto`
100	`load loop lt makefont matrix maxlength mod moveto mul ne`
110	`neg newpath not null or pathbbox pathforall pop print printobject`
120	`put putinterval rcurveto read readhexstring readline readstring rectclip rectfill rectstroke`
130	`rectviewclip repeat restore rlineto rmoveto roll rotate round save scale`
140	`scalefont search selectfont setbbox setcachedevice setcachedevice2 setcharwidth setcmykcolor setdash setfont`
150	`setgray setgstate sethsbcolor setlinecap setlinejoin setlinewidth setmatrix setrgbcolor setshared shareddict`
160	`show showpage stop stopped store string stringwidth stroke strokepath sub`
170	`systemdict token transform translate truncate type uappend ucache ueofill ufill`
180	`undef upath userdict ustroke viewclip viewclippath where widthshow write writehexstring`
190	`writeobject writestring wtranslation xor xshow xyshow yshow FontDirectory SharedFontDirectory Courier`
200	`Courier-Bold Courier-BoldOblique Courier-Oblique Helvetica Helvetica-Bold Helvetica-BoldOblique Helvetica-Oblique Symbol Times-Bold Times-BoldItalic`
210	`Times-Italic Times-Roman execuserobject currentcolor currentcolorspace currentglobal execform filter findresource globaldict`
220	`makepattern setcolor setcolorspace setglobal setpagedevice setpattern` (226–255 unassigned)
256	`= == ISOLatin1Encoding StandardEncoding [ ] atan banddevice bytesavailable cachestatus`
266	`closefile colorimage condition copypage cos countdictstack countexecstack cshow currentblackgeneration currentcacheparams`
276	`currentcolorscreen currentcolortransfer currentcontext currentflat currenthalftone currenthalftonephase currentmiterlimit currentobjectformat currentpacking currentscreen`
286	`currentstrokeadjust currenttransfer currentundercolorremoval defaultmatrix definefont deletefile detach deviceinfo dictstack echo`
296	`erasepage errordict execstack executeonly exp false filenameforall fileposition fork framedevice`
306	`grestoreall handleerror initclip initgraphics initmatrix instroke inustroke join kshow ln`
316	`lock log mark monitor noaccess notify nulldevice packedarray quit rand`
326	`rcheck readonly realtime renamefile renderbands resetfile reversepath rootfont rrand run`
336	`scheck setblackgeneration setcachelimit setcacheparams setcolorscreen setcolortransfer setfileposition setflat sethalftone sethalftonephase`
346	`setmiterlimit setobjectformat setpacking setscreen setstrokeadjust settransfer setucacheparams setundercolorremoval sin sqrt`
356	`srand stack status statusdict true ucachestatus undefinefont usertime ustrokepath version`
366	`vmreclaim vmstatus wait wcheck xcheck yield defineuserobject undefineuserobject UserObjects cleardictstack`
376	`A` to `Z` (376–401), then `a` to `z` (402–427)
428	`setvmthreshold << >> currentcolorrendering currentdevparams currentoverprint currentpagedevice currentsystemparams currentuserparams defineresource`
438	`findencoding gcheck glyphshow languagelevel product pstack resourceforall resourcestatus revision serialnumber`
448	`setcolorrendering setdevparams setoverprint setsystemparams setuserparams startjob undefineresource GlobalFontDirectory ASCII85Decode ASCII85Encode`
458	`ASCIIHexDecode ASCIIHexEncode CCITTFaxDecode CCITTFaxEncode DCTDecode DCTEncode LZWDecode LZWEncode NullEncode RunLengthDecode`
468	`RunLengthEncode SubFileDecode CIEBasedA CIEBasedABC DeviceCMYK DeviceGray DeviceRGB Indexed Pattern Separation`
478	`CIEBasedDEF CIEBasedDEFG DeviceN`

Some entries name Display PostScript operators (viewclip, fork, yield and others) that ordinary interpreters lack.

Sources

Adobe, PostScript Language Reference, third edition (1999) – §3.14 and tables 3.25, 3.26 and 3.27 (binary tokens, number representations, object sequences), appendix F (system name encodings)