Amiga Elysian Archive/AmigaElysianArchive.iso/printer/post17b.lzh/post.doc

Text File | 1992-03-13 | 28KB | 765 lines

Postscript interpreter - user documentation
===========================================

Post V1.7 Copyright Adrian Aylward 1989, 1992

Free use and non-commercial reproduction of the binaries of the interpreter
library (post.library) is permitted, providing that the copyright notices
are not removed.  Distribution by disk libraries is permitted provided only
a nominal copying fee is charged. Upload to commercial bulletin boards is
also permitted, providing charges are only made for connect time, and there
is no specific charge made for the file.

You may freely copy, use and modify the user interface program (post).

For queries and bug reports write to the author:

        Adrian Aylward,
        20 Maidstone Road,
        SWINDON,
        Wiltshire.
        UK.

No responsibilities accepted for bugs, but please let me know so I can fix
them.

Introduction
============

"Post" is a software based PostScript interpreter, presently running on the
Amiga.  The source code is written in C, and should be fairly portable to
other machines.  It supports the full Adobe language, with only minor
variations.

Prerequisites
=============

Post uses the Arp library (V39+), primarily for its file requestor.  For
interactive working under WorkBench 1.3 you also need ConMan (V1.3+).
 Both these programs are widely available; any good BBS or PD/shareware
library should have them. (ConMan is shareware.)  You may well have them
installed on your machine already.

If you prefer not to install ConMan as your standard console handler, you
will need to access it via an alternative device name, which you must
specify by the CONDEV option.

You will need at least a megabyte of memory, more for high density printer
output.

Command line interface
======================

usage:
    post [files...] [IFF file] [SCREEN] [PRINTER] [INTERACTIVE]
         [SIZE xyod..s..p.bc.] [MEM fhlv..] [CLOSEWB] [CONDEV con]

If you forget, type "post ?" to display this.

The input files
---------------

The interpreter reads the input files in turn.  When running interactively
the input files are treated as startup files, and are rerun whenever the
interpreter is restarted.  So you would normally specify just the standard
startup file "init.ps", or an alternative of your own.  When running
non-interactively you should in addition specify all the files you want to
interpret, up to 5 in total; when the last is finished the interpreter will
exit.

The file "*" is treated specially, in that it is always considered
interactive. So it is possible to run Post entirely within a CLI window
and still get the interactive prompts.

Printer output
--------------

This is the default if neither the IFF nor SCREEN options are used, when
running non-interactively.  The output is sent to the printer device as a
graphic dump.

The interpreter will obtain default values for the page size, density, and
number of colors from the current preferences.  If you use the default sizes
make sure you have set the page size: run preferences and on the graphic 2
screen set the width and height limits -  either bounded or absolute or in
pixels.  The output defaults to black and white or 3 colour, according to
the preferences.

IFF file output
---------------

This option will send the output to an IFF file.  Each IFF file contains a
single FORM ILBM.  So that multiple pages may be created in a single run of
the interpreter, sequence numbers are automatically generated:

    IFF path/fred*.pic   will generate  path/fred1.pic, path/fred2.pic ...
and IFF path/fred??.pic  will generate  path/fred01.pic, path/fred02.pic ...

The default page size etc. is the same as for screen output.

SCREEN output
-------------

This option  will send the output to a window on the interactive screen.  It
implies interactive working.

You can generate screen, printer and IFF output at the same time.  However
screen output has to use chip memory, and printer output at high resolution
needs a large page buffer (about a megabyte for an A4 page at 300 dpi.). So
if you attempt the two simultaneously you will probably run out of memory,
unless your machine has 2 megabytes of chip ram.

The default screen size is the same as the workbench screen, but in
interlace mode.  The default page size is A4, with a density of 75 pixels
per inch.  The default number of colours is 3.  The page size is rounded up
to the maximum size of the window.

The INTERACTIVE option
----------------------

This option causes the interpreter to run interactively.  A screen is opened
and on it appears a requestor window for the parameters.  If you click on
OK two windows will be opened, one for console input and output, and the
other to display the page buffer.  Use the menus.

The SIZE option
---------------

The SIZE option sets the page size and density etc..

    SIZE "x..y..s..xo..yo..[x]d..[y]d..p.bc."

Up to 5 SIZE options are allowed; later values override earlier ones.

The page sizes ("x..y..s..") are in pixels, as decimal numbers; "s" sets
the size of the bands, for band rendering to the printer.  The page offsets
(xo..yo..) are in pixels, shifting the location of the bottom left of the
bitmap relative to the PostScript logical page (origin at (0,0) in the
initial user space). Densities ([x]d..[y]d..) are in pixels per inch; "xd"
or "yd" set the x or y density, and "d" sets both.

The printer density can be set by "p"; its value should be in the range
1 to 7.

Printer output uses the Workbench printer device. If you wish to print to
a LaserJet or DeskJet you should use the "postlj" program instead, as it
gives better control over page size and runs faster.

The number of colours is: "b" for black and white (1 bit plane), "c" or "c3"
for 3 colour rgb (3 bit planes), and "c4" for 4 colour cmyk (4 bit planes).

The MEM option
--------------

The MEM option controls the amount of memory allocated for workspace.

    MEM "v..f..l..h.."

Up to 5 MEM options are allowed; later values override earlier ones.

    Default  Minimum
    -------  -------
     f60000     1000     Font cache
     h20000     1000     Halftone cache
     v50000     5000     Virtual machine memory segment size
     l10000     1000     Line drawing workspace segment size

In addition memory is allocated for the page buffer (one bit per pixel)
and for drawing and imaging buffers.

Since most of the workspace is allocated automatically you will not usually
need to adjust these options.  If your program is so large that you do need
to increase any of these values, be aware that it is unlikely that it will
run on any standard PostScript printer.

The font cache is used to store character bitmaps; it saves recalculating
them each time they are needed, so text rendering becomes many times faster.
The default value is reasonable for average sized machines; if you are
trying to shoehorn things into a small machine you could try reducing it. Or
if you are printing a document using a large number of different characters,
fonts, and sizes you may be able to improve performance by increasing it.

The halftone cache holds halftone patterns used for simulating gray values.
The default value will normally be satisfactory, but you might possibly need
to increase it if you are using very large halftone patterns, for example
for complex patterned fills.

The virtual machine memory is automatically allocated in segments as they
are needed, up to a maximum of 63.  For most purposes you will not need
to alter the segment size, but if you are running a gigantic program you
might possibly need to increase it.

The line drawing workspace holds the current path, together with all the
paths saved by a gsave operation.  It is automatically allocated in
segments, up to a maximum of twenty.  The default size is sufficient for
about 4000 path elements, enough for all normal programs.

The CLOSEWB option
------------------

If memory is tight, you can try using this option to close down WorkBench
while Post is running.  This only works if there are no application
windows open on the WorkBench screen.

The CONDEV option
-----------------

The console handlers supplied with Workbench 1.3 did not have the
capability of attaching a console to a window on a custom screen. So if
you try to run Post interactively the console ouput will appear on the
Workbench screen and screen output on its custom screen - which is not
very convenient. So if you are running under 1.3 you should install the
alternative (shareware) console handler called ConMan; you need version
1.3 or greater.

If you have Workbench 2.0 this is built into the standard console handler,
so there si no need to install ConMan.

If you want to use a console handler installed using an alternative device
name you can specify it using the CONDEV option.

Default values for CONDV are

        CON:W               (WorkBench 1.3)
        CON://///WINDOW     (Workbench 2.0)

Post will append to these strings the hexadecimal address of the interactive
window, and use the resulting filename to open the console stream.

Examples
--------

To run postscript programs interactively, rendering to the screen:

    post init.ps screen

To print a file with many big characters using a large font cache and
increasing the path limit:

    post init.ps myfile.ps printer mem f500000l200000

To render a series of 640 * 512 pictures at 762 dpi. to iff files t:i01,
t:i02 etc.:

    post init.ps mypics.ps iff t:i?? size x640y512d72

The WorkBench interface
=======================

Just click on the icon.  It always runs interactively.  You can set up
default arguments by setting the ToolTypes; add a string "ARGS=...."
and it will be parsed just like the CLI startup.  For example
"ARGS=init.ps screen" will define the standard startup file and make the
screen the default output.

The Menus
=========

Project
    Restart              Takes you back to the parameters requestor, after
                         which the interpreter is reinitialised.

    Quit                 Quit from the program

File
    Load font            Loads a file from the PSFonts: directory by default

    Load file            Loads a file, without saving or restoring the VM

    Run file             Runs a file, saving VM before and restoring after

    Interactive          Interprets interactive input

Control
    Pause every page     Toggles pause every showpage/copypage

    Continue after pause Does just that

    Interrupt            Sends a break; equivalent to CTRL/C

    Kill                 Generates an interrupt that cannot be trapped

Before and after executing a file the operand stack is cleared.  This is
to prevent invalidrestore errors, particularly if interpretation ended in
error with values on the stack.  If you need to leave things on the stack
you will have to run the files from a driver file, or from the interactive
window.  At the end of the file, (after any restore), the page is erased.

If you use the automatic font loading, as defined in the standard startup
file "init.ps", the font will be discarded when the VM is restored.  To
avoid repeated reloading of common fonts it may be better to use the "Load
font" menu item to preload the ones you want.  N.B. there is no check to
stop you loading the same one twice; it will probably work but it will
waste memory.

Compatability
=============

The interpreter is based upon the language (version "23.0") as described by
the (red) book:

    "PostScript Language Reference Manual"
    Adobe Systems Incorporated,
    Addison-Wesley 1985, ISBN 0-201-10174-2

This is PostScript level 1, in Adobe terminology, as described in the first
edition of the Red Book.

In addition, the 4 colour operators from level 2 are also supported. Level
2 PostScript is defined in the second edition:

    "PostScript Language Reference Manual (second edition)"
    Adobe Systems Incorporated,
    Addison-Wesley 1990, ISBN 0-201-18127-4

File handling
-------------

The files %stdin, %stdout, %stderr are permanently open as far as the
operating system is concerned.  Attempts to close them will invalidate the
file object but not close them to the operating system.  If they are
connected to a terminal they will be unbuffered, so output will be
transmitted immediately, without waiting for a "flush".  The special files
%statementedit and %lineedit are not supported.  There is no end of file
control character.

Error handling
--------------

All the standard Adobe error conditions are supported.  There also some
additional errors:

The error "invalidstop" will be generated if a "stop" operator is executed
not within any encolsing "stopped" context.

The error "memoryallocation" will be executed if the interpreter is unable
to obtain sufficient memory from the operating system.

The error "kill" may be executed in response to an external command.  It
bypasses the error handler in errordict, so that it is always possible to
kill a postscript program, even one that traps its own errors.

The execution stack
-------------------

The exact contents of the execution stack are not defined by the language
reference manual, and cannot therefore be guaranteed to be the same as other
implementations.  Dumping the stack is likely to yield useful debugging
information but is not recommended for normal programming.

Colour mapping
--------------

The interpreter can run with one bit plane (for black and white printers),
3 bit planes (red, green, blue) or 4 bit planes (cyan, magenta, yellow,
black).  For screen output 3 bit planes are appropiate.  For a colour
printer the 3 or 4 colour model can be used.  The output is always in
binary; gray scale is supported only via the halftone mechanism.

For low resolution colour printers the 3 colour model is probably best.
The 4 color model is for high resolution colour printers that can generate
high frequency halftone screens, where a fourth screen is used to improve
the quality of the blacks and grays.

(N.B. the preferences printer drivers only handle 3 colour model, so if
you want to use the 4 colour model you will probably need to generate an
IFF file and write your own print dump program.  They will however
automatically use black the ribbon/ink if the printer has one)

The transformations between the RGB and the HSB models are based on those in
the book:

    "Procedural Elements for Computer Graphics"
    David F. Rogers,
    McGraw-Hill 1985, ISBN 0-E07-053534-5.

Band Rendering
--------------

For a high resolution printer the page buffer can be quite large (a megabyte
or so).  If you don't have enough memory to hold the entire page at once you
can render each one in several bands.  Set the "SIZE s..." option to the
largest band size you have space for.  (For matrix printers it will probably
be best to make it a multiple of the number of printer pins times the number
of vertically displaced passes).  Then run the postscript program generating
each page once for each band, using the "setband" operator, which is
described below, to set the base for each band.  Start at a base of zero
and increment the base by the band size until you reach the page height.
Each band is sent to the printer as a separate graphic dump; the last band
generates a form feed to eject the page.  You can use the "currentband"
operator to calculate the number of bands needed:

    currentband 1 sub { setband pageproc } for 0 setband

will execute "pageproc" once for each band (remaining) on the page.  The
procedure should execute copypage (or showpage) exactly once per iteration.

This technique is only suitable for printer output.  There is a special
driver program "postband" that handles this automatically for programs
that conform to the Adobe structuring conventions.

Fonts
-----

There are no fonts built in to the interpreter; they must all be downloaded.
The findfont operator will execute the error invalidfont if the font you
request is not present in the font dictionary.  You can redefine it, perhaps
to search the Unix/Amiga filing system to automatically download fonts or to
substitute a default font.  See the standard startup file.

Both type 1 (Adobe encrypted format) and type 3 (standard PostScript as per
the red book) fonts are now supported.  In addition the IBM font format can
be read directly.

The font caching will work significantly better if all your fonts have
UniqueID's.  Then repeated makes of the same font with the same matrix will
be cached, and cached character bitmaps can be retained even if the font
is removed by a restore operation.  N.B. if fonts are downloaded after a
save they will be removed by the corresponding restore; in the interests of
efficiency it may therefore be better to preload them.

Other features
--------------

Names of the form //name are looked up immediately by the scanner, and
packed arrays have been implemented.  These features were not on the
original Adobe red book, but were added to the language specification later.

When scanning the contents of a string, if there any embedded strings the
escape sequences are interpreted just as they are in a file.  This follows
the specification of the more recent Adobe interpreters (including display
PostScript) and not the original red book.

Operators omitted or changed
----------------------------

    banddevice

Not appropiate for the Unix/Amiga environment.

    bytesavailable

Since we can't tell how many bytes are available we always return -1.

    charpath

Meets the level 2 spec.  At the end of the string a moveto is appended to
move to the correct origin for the next character.  (So the current point
is advanded by the width of the string.)

    copypage

For printer output this works like the standard.  The number of pages
printed is equal to #copies.  For screen output it does not need to do
anything, as the screen is not buffered.  For IFF file output it writes the
page to the next file.

    echo

Not appropiate for the Unix/Amiga environment.

    eexec

This is implemented as per the Adobe type 1 fonts book.  Its operand must be
a file, which must match the object on top of the execution stack.  It can be
terminated only by a closefile (which terminates the eexec but does not
actually close the file).  So the sequences "currentfile eexec" and
"currentfile closefile" as described in the book will work, but other
combinations will likely fail.

    executive

There is no executive, as the interpreter supports interactive usage
directly.

    findfont

Looks in the FontDirectory.  If it can't find the supplied key, it executes
the error invalidfont.

    flushfile

We don't check for errors when flushing an input file, to prevent recursion
in the error handler.

    framedevice

Not appropiate for the Unix/Amiga environment.

    pathbbox

Meets the level 2 spec.  If the path ends with a moveto that is not the only
element then it is ignored when calculating the bounding box.  This means
that the sequence "charpath pathbbox" will determine the bounds of the
character(s), not including the empty space after the final character.  See
also "charpath".

    prompt

Not called by the interpreter.  See "prompts" instead.

    renderbands

Not appropiate for the Unix/Amiga environment.

    resetfile

Does nothing.

    showpage

See the notes for copypage above.

    usertime

Returns the time elapsed since the interpreter started, with a resolution of
one second.

Operators added
---------------

    currentband

        "currentband" base size height

Returns the base of the current band, size of each band, and total height
of the page.

    setband

        base "setband"

Sets the base height of the band being rendered.  The value must be greater
than or equal to zero and less than the total height of the page.

    fontfile

        (filename) "fontfile"

Opens a file for reading.  If the first byte is hex "80" the file is assumed
to be in IBM font format, and its contents are converted to standard
ascii/binary as they are read.  The standard input file (%stdin) is never in
IBM font format.

    invalidstop

        (error)

A "stop" has been executed for which there is no dynamically enclosing
"stopped" context.

    prompts

        string string "prompts"

The first string is used as the prompt when the interpreter is scanning
terminal input and executing it immediately; the second string is used when
execution is defered while scanning the contents of a procedure "{ ... }".
No prompt is given when a newline is encountered within a string.

    vmhwm

        "vmhwm" hwmused maximum

Returns two integers: the high water mark of the amount virtual memory used
since the program was first started or the last "vmhwm", and the maximum
available amount of memory.

    callextfunc

        result arg1 arg2 ... argc func "callextfunc" result

Calls an external function using a C calling sequence.  This operator is not
available for use by ordinary PostScript programs; it can only be used when
the library is called with an external function table supplied as an argument
(see the library interface documentation for details).  The "func" argument
is the index of the function in the external call table, and "argc" is the
number of parameters; both must be non-negative integers.  The result is
the object in which the result will be placed.  If it a null then no result
will be returned and the null object will be discarded; otherwise it must
match the type of the result returned from the function.  For both result and
parameters bool and integer types will work OK; real types will work if your
C compiler passes them just like ints (use ANSI prototypes so it does not
expand them to double).  For string parameters the address is passed.  Arrays
may be passed as parameters, but they are only useful if you know their
PostScript object representation.  The maximum number of parameters
permitted is 20.

Miscellaneous
-------------

The system dictionary is left writeable so that standard preludes can add
things to it - possibly removing write permission when they have finished.
Similarly FontDirectory is writeable, so that it is possible to add entries
without using definefont.

Device space
------------

The largest device that the interpreter can handle is 30000 by 30000
pixels.  In practice you will probably run out of memory long before
reaching this limit.

Known Bugs
==========

If you close down WorkBench the menus sometimes get truncated and slightly
corrupted.  I suspect this is a Workbench/Intuition problem.

If you type CTRL/C to abort if the printer is not ready then strange things
may happen on subsequent attempts to print.

N.B. if you run very low on memory, Intuition may behave strangely, and
refuse to resize windows etc..  This is not a bug in Post.

If you type CTRL/backslash in the interactive window you won't be able to
run interactively again until a restart.

Post tries to keep the console window active, so that all keyboard input
goes there.  It will probably be incompatible with mouse handlers that
automatically activate the window under the cursor.

Versions
========

    V0.0  14-Nov-89 (prerelease)

The original.

    V0.1  06-Dec-89 (prerelease)

Image routines added.

Range checks eased on gray levels and colours etc..

Area filling speeded up, matrix manipulation and colour mapping rewritten,
numeric conditioning improved for path fill.

Bugs fixed: syntax error, == string escapes, error handling and quit when
recursing, vm error handling, integer overflow in path fill, matrix save,
interrupt after printing page, dictionary second save, name table restore,
boolean type checking, multiple halftone screens memory allocation.

    V0.2  07-Jan-90 (prerelease)

Font and character routines added.  Null device added.  Band rendering of
printer output added.

Matrix manipulation rewritten again.  Interpreter recursion rewritten.
Allocate stacks dynamically.

Bugs fixed: missing access checks on strings, scan token negative chars,
depth check in path flattening, pathbbox, strings or null objects as
dictionary keys, shade after grestore(all), fill/image with clip region,
clip unclosed subpath, dictstack/execstack.

    V1.0  12-Feb-90

Support for big scrollable windows added.  Menu handler added with port
name option.  Workbench startup added with parameters requestor window.

Hex string scan optimised.  Break now CTRL/C only, not CTRL/D.

Bugs fixed: aspect ratio in iff file, clipping pictures much wider than the
page.

    V1.1  27-Mar-90

Interpreter made into shared library.  Message port interface removed.
Band printing driver now included.

Immediate lookup of //name added.  Packed arrays added.  CR and CRLF now
converted to LF for readline and when scaning tokens from a file.  Copypage
now takes account of #copies.

Bugs fixed: slow images where strings are not rectangles, attributes after
copy and bind, charpath with fonts using gsave, iff run compression,
setundercolorremoval now conforms to Adobe spec, make errorname literal,
slow images repositioned by 1/2 pixel.

    V1.2  14-Oct-90

Allow multiple SIZE and MEM options.

Page offset option added, page size defaults to A4, integer gadgets checked
for +ve values.

Type 1 fonts, IBM font format, encrypted files, eexec operator added.

Bugs fixed: location zero not zero, printer device signals, proc arrays not
executable, bind packed arrays, restore closing files, makefont matrices not
commutative, makefont with translation, error handling during recursion,
graphics state within buildchar proc, miterlimit less that sqrt(2), cvs
copying too many bytes, solid lines after dashes, floating point error trap
recursion (2620 version on 68882), check for fpu present, charpath with null
character path, error handling loading file, invalid restore check,
writehexstring.

    V1.3  27-Oct-90

LaserJet driver added.

Interface fixes: Menus no longer ghost after startup file error and restart,
pause status is now retained over a restart, printer preferences page size
is now handled correctly, no longer crashes if arp.library is missing.

Bugs fixed: IBM binary eexec sections beginning with white space, funny
characters in error message command names, hints with stems of negative
width, font character cache sizing, tune baseline alignment and flex,
eexec lines ending in CR (not CRLF), aload packed arrays, zero length
charstrings, font character cache hash chains.

    V1.4  22-Jan-91

External function call interface added.  New flexible memory allocation for
vm and paths; new error "memoryallocation".  More flexible font cache limit.

Interface fixes: Memory Fonts/Paths gadgets transposed, drag scroll bars
under 2.0.

Bugs fixed: writehexstring, zero height clipped paths, $error in userdict,
undocumented charstring codes, halftone cacheing, mixed type 1 and 3 fonts.

    V1.5  01-Mar-91

Type 1 font character rendering improved (though slower).

Fix printer page size selection (inches).  Printer output now truncates to
maximum dump size, to prevent scaling.

Bugs fixed: path fill at top of clip path, free first vm segment, rotate
scale and translate with matrix operands, cached characters within clip path,
long clip paths, clippath not default clip path, hollow clip paths, setband.

    V1.6  19-Apr-91

Version defined as real number string in init.ps.  Define ISOLatin1 encoding
vector.

New menu and error "kill".

New error name string library entry.

Pathbbox now meets level 2 spec.  Allow Encoding length less than 256.  Use
.notdef for type 1 buildchar.  Increase $error dictionary size; move $error
to systemdict.  Allow name operands to length operator.  New operator
cleardictstack.  Change cvrs to correct spec.

Bugs fixed: fast image positioning when ydir negative.

    V1.7  12-Mar-92

Post no treats "*" as interactive. New specification for CONDEV option,
should work with Workbench 2.0.

PostLJ new page size options and now works with DeskJet.


Type 1 font rendering now interpolates between hints.

Bugs fixed: Strings as dictionary keys, type 1 fonts etc. with white space
after eexec, arcs of zero radius, stroking lines of zero length, roll of
zero items, image and kshow procs with strange stack discipline, setflat
values out of range.