Category : Word Processors
Archive   : STEVIE.ZIP
Filename : SOURCE.DOC

Release Notes for STEVIE - Version 3.10

Source Notes

Tony Andrews

3/6/88


Overview
--------

This file provides a brief description of the source code for
Stevie. The data structures are described later as well. For information
specific to porting the editor, see the file 'porting.doc'. This document
is more relevant to people who want to hack on the editor apart from doing
a simple port.

Most of this document was written some time ago so a lot of the
discussion centers on problems related to the Atari ST environment and
compilers. Most of this can be ignored for other systems.

Things You Need
---------------

Stevie has been compiled with both the Alcyon (4.14A) and the
Megamax C compilers. For the posted binary, Megamax was used because
it's less buggy and provides a reasonable malloc(). Ports to other
compilers should be pretty simple. The current ifdef's for ALCYON and
MEGAMAX should point to the potential trouble areas. (See 'porting.doc'
for more information.)

The search code depends on Henry Spencer's regular expression
code. I used a version I got from the net recently (as part of a 'grep'
posting) and had absolutely no trouble compiling it on the ST. Thanks,
Henry!

The file 'getenv.c' contains a getenv routine that may or may
not be needed with your compiler. My version works with Alcyon and
Megamax, under either the Beckemeyer or Gulam shells.

Lastly, you need a decent malloc. Lots of stuff in stevie is
allocated dynamically. The malloc with Alcyon is problematic because
it allocates from the heap between the end of data and the top of stack.
If you make the stack big enough to edit large files, you wind up
wasting space when working with small files. Mallocs that get their memory
from GEMDOS (in fairly large chunks) are much better.


Cruft
-----

Some artifacts from Tim Thompson's original version remain. In
some cases, code has been re-written, with the original left in place
but ifdef'd out. This will all be cleaned up eventually, but for now
it's sometimes useful to see how things used to work.


Data Structures
---------------

A brief discussion of the evolution of the data structures will
do much to clarify the code, and explain some of the strangeness you may
see.

In the original version, the file was maintained in memory as a
simple contiguous buffer. References to positions in the file were simply
character pointers. Due to the obvious performance problems inherent in
this approach, I made the following changes.

The file is now represented by a doubly linked list of 'line'
structures defined as follows:

struct line {
struct line *prev, *next; /* previous and next lines */
char *s; /* text for this line */
int size; /* actual size of space at 's' */
unsigned int num; /* line "number" */
};

The members of the line structure are described more completely here:

prev - pointer to the structure for the prior line, or NULL for the
first line of the file

next - like 'prev' but points to the next line

s - points to the contents of the line (null terminated)

size - contains the size of the chunk of space pointed to by s. This
is used so we know when we can add text to a line without getting
more space. When we DO need more space, we always get a little
extra so we don't make so many calls to malloc.

num - This is a pseudo line number that makes it easy to compare
positions within the file. Otherwise, we'd have to traverse
all the links to tell which line came first.


Since character pointers served to mark file positions in the
original, a similar data object was needed for the new data structures.
This purpose is served by the 'lptr' structure which is defined as:

struct lptr {
struct line *linep; /* line we're referencing */
int index; /* position within that line */
};


The member 'linep' points to the 'line' structure for the line containing
the location of interest. The integer 'index' is the offset into the line
data (member 's') of the character to be referenced.

The following typedef's are more commonly used:

typedef struct line LINE;
typedef struct lptr LPTR;

Many operations that were trivial with character pointers had to be
implemented by functions to manipulate LPTR's. Most of these are in the
file 'ptrfunc.c'. There you'll find functions to increment, decrement,
and compare LPTR's.

This was the biggest change to the editor. Fortunately, I made this
change very early on, while I was still doing the work on a UNIX system.
Using 'sdb' made it much easier to debug this code than if I had done it
on the ST.


Summary
-------