DOC HOME SITE MAP MAN PAGES GNU INFO SEARCH PRINT BOOK
 

Ghostscript PostScript coding guidelines

Table of contents

For other information, see the Ghostscript overview.


Summary of the coding guidelines


Introduction

The many rules that Ghostscript's code follows almost everywhere are meant to produce code that is easy to read. It's important to observe them as much as possible in order to maintain a consistent style, but if you find a rule getting in your way or producing ugly-looking results once in a while, it's OK to break it.


Use of PostScript language features

Restrictions

If you need to store a value temporarily, don't write into a literal in the code, as in this fragment to show a character given the character code:

( ) dup 0 4 -1 roll put show

Instead, allocate storage for it:

1 string dup 0 4 -1 roll put show

Protection

If an object is never supposed to change, use readonly to make it read-only. This applies especially to permanently allocated objects such as constant strings or dictionaries.

During initialization, and occasionally afterwards, it may be necessary to store into a read-only dictionary, or to store a pointer to a dictionary in local VM into a dictionary in global VM. The operators .forceput and .forceundef are available for this purpose. To make these operators inaccessible to ordinary programs, they are removed from systemdict at the end of initialization: system code that uses them should always use bind and odef (or executeonly) to make uses of them inaccessible as well.

Standard constructions

Multi-way conditionals

If you write a block of code with more than about 3 exit points, the usual way to do it would be like this:

{
  ... {
    ...1
  } {
    ... {
      ...2
    } {
      ... {
        ...3
      } {
        ...4
      } ifelse
    } ifelse
  } ifelse
}

However, this causes the 4 logically somewhat parallel code blocks to be indented differently, and as the indentation increases, it becomes harder to see the structure visually. As an alternative, you can do it this way:

{       % The loop doesn't actually loop: it just provides a common exit.
  ... {
    ...1
    exit
  } if
  ... {
    ...2
    exit
  } if
  ... {
    ...3
    exit
  } if
  ...4
  exit
} loop

Don't forget the final exit, to prevent the loop from actually looping.

Switches

Use a dictionary or an array of procedures to implement a 'switch', rather than a series of conditionals, if there are more than about 3 cases. For example, rather than:

dup /a eq {
  pop ...a
} {
  dup /b eq {
    pop ...b
  } {
    dup /c eq {
      pop ...c
    } {
      ...x
    } ifelse
  } ifelse
} ifelse

(or using the loop/exit construct suggested above), consider:

/xyzdict mark
  /a {...a} bind
  /b {...b} bind
  /c {...c} bind
.dicttomark readonly def
...
//xyzdict 1 index .knownget {
  exch pop exec
} {
  ...x
} ifelse

File structuring

Every code file should start with comments containing

  1. a copyright notice;
  2. the name of the file in the form of an RCS Id:
    % Id$: filename.ps $
    
  3. a very brief summary (preferably only one line) of what the file contains.

If you create a file by copying the beginning of another file, be sure to update the copyright year and change the file name.


Commenting

If a file has well-defined functional sections, put a comment at the beginning of each section to describe its purpose or function.

Put a comment before every procedure to describe what the procedure does, unless it's obvious or the procedure's function is defined by the PLRM. In case of doubt, don't assume it's obvious. If the procedure may execute a deliberate 'stop' or 'exit' not enclosed in 'stopped' or a loop respectively, that should be mentioned. However, information about the arguments and results should go in the argument and result comment (described just below) if possible, not the functional comment.

Put a comment on every procedure to describe the arguments and results:

/hypot {	% <num1> <num2> hypot <real>
  dup mul exch dup mul add sqrt
} def

There is another commenting style that some people prefer to the above:

/hypot {	% num1 num2 --> realnum
  dup mul exch dup mul add sqrt
} def

We have adopted the first style for consistency with Adobe's documentation, but we recognize that there are technical arguments for and against both styles, and might consider switching some time in the future. If you have strong feelings either way, please make your opinion known to gs-devel@ghostscript.com.

Put comments describing the stack contents wherever you think they will be helpful; put such a comment at the beginning of every loop body unless you have a good reason not to.

When you change a piece of code, do not include a comment with your name or initials. Also, do not retain the old code in a comment, unless you consider it essential to explain something about the new code; in that case, retain as little as possible. (CVS logs do both of these things better than manual editing.) However, if you make major changes in a procedure or a file, you may put your initials, the date, and a brief comment at the head of the procedure or file respectively.


Formatting

Indentation

Indent 2 spaces per indentation level. You may use tabs at the left margin for indentation, with 1 tab = 8 spaces, but you should not use tabs anywhere else, except to place comments.

Indent { } constructs like this:

... {
  ...
} {
  ...
} ...

If the body of a conditional or loop is no more than about 20 characters, you can put the entire construct on a single line if you want:

... { ... } if
rather than
... {
  ...
} if

There is another indentation style that many people prefer to the above:

...
{ ...
}
{ ...
} ...

We have adopted the first style for consistency with our C code, but we recognize that there are technical arguments for and against both styles, and might consider switching some time in the future. If you have strong feelings either way, please make your opinion known to gs-devel@ghostscript.com.

Spaces

Always put spaces between two adjacent tokens, even if this isn't strictly required. E.g.,

/Halftone /Category findresource

not

/Halftone/Category findresource

Naming

All names should consist only of letters and digits, possibly with an initial ".", except for names drawn from the PostScript or PDF reference manual, which must be capitalized as in the manual. In general, an initial "." should be used for those and only those names that are not defined in a private dictionary but that are meant to be used only in the file where they are defined.

For edits to existing code, names made up of multiple words should not use any punctuation, or capitalization, to separate the words, again except for names that must match a specification. For new code, you may use this convention, or you may use the "Vienna" convention of capitalizing the first letter of words, e.g., readSubrs rather than readsubrs. If you use the Vienna convention, function names should start with an upper case letter, variable names with a lower case letter. Using the first letter of a variable name to indicate the variable's type is optional, but if you do it, you should follow existing codified usage (****** WE NEED A REFERENCE FOR THIS ******).


Miscellany

Some useful non-standard operators

<obj1> <obj2> ... <objn> <n> .execn ...
This executes obj1 through objn in that order, essentially equivalent to
<obj1> <obj2> ... <objn> <n> array astore {exec} forall

except that it doesn't actually create the array.

<dict> <key> .knownget <value> true
<dict> <key> .knownget false
This combines known and get in the obvious way.
<name> <proc> odef -
This defines name as a "pseudo-operator". The value of name will be executable, will have type operatortype, and will be executed if it appears directly in the body of a procedure (like an operator, unlike a procedure), but what will actually be executed will be proc. In addition, if the execution of proc is ended prematurely (by stop, including the stop that is normally executed when an error occurs, or exit) and the operand and dictionary stacks are at least as deep as they were when the "operator" was invoked, the stacks will be cut back to their original depths before the error is processed. Thus, if pseudo-operator procedures are careful not to remove any of their operands until they reach a point in execution beyond which they cannot possibly cause an error, they will behave just like operators in that the stacks will appear to be unchanged if an error occurs.

Some useful procedures

<object> <errorname> signalerror -
Signal an error with the given name and the given "current object". This does exactly what the interpreter does when an error occurs.

Other

If you can avoid it, don't allocate objects (strings, arrays, dictionaries, gstates, etc.) in commonly used operators or procedures: these will need to be garbage collected later, slowing down execution. Instead, keep values on the stack, if you can. The .execn operator discussed above may be helpful in doing this.

If you find yourself writing the same stretch of code (more than about half a dozen tokens) more than once, ask yourself whether it performs a function that could be made into a procedure.


Copyright © 2000 Aladdin Enterprises. All rights reserved.

This software is provided AS-IS with no warranty, either express or implied. This software is distributed under license and may not be copied, modified or distributed except as expressly authorized under the terms of the license contained in the file LICENSE in this distribution.

Ghostscript version 7.07, 17 May 2003