🚀 go-pugleaf

This is a request for comments on an editor which will soon be released
for public abuse. I hope to release the sources to comp.sources.unix
in the near future, but what with taxes and real work (the kind that
pays the bills and causes the taxes) I can't commit to when that will
be. The code is pretty complete (i.e. all the version 1.0 features that
I want are currently in it) and reliable. I use the editor now for most
of my daily editing activies both at home and work and have gotten over
the habit of saving my files every minute or so :-).

I don't want to start an editor war or arguments over which type of editor
is best. I am simply interested in hearing 1) which FOLDED features people
like and 2) what they would like to see added. Also, if anyone sees any obvious
shortfall that I may have missed, now would be a good time to let me know.
To this end, I am including part of the documentation that I am putting
together for FOLDED. Sorry, but this is kind of long.

  -Todd Burkey
   trb@stag.UUCP
   or tburkey@eta.com

-------------Partial Draft of the FOLDED Documentation------------------

			    FOLDED
		       A folding editor
			  (c) 1989
			By Todd Burkey

FOLDED is a dual mode editor that provides several advanced features
not currently found in text and programming editors. FOLDED is primarily
intended to be used as a source code editor. As such, it incorporates
a folding feature similar to that found in thought processor programs
but with added support for the C language (and eventually Pascal). FOLDED
also has some relational capabilities. These allow the user to associate
lines in their source files with sections of their documentation files
and vice-versa. And both folding and relations are incorporated in such a
fashion that they don't impact the original text files in any way.

FOLDED also has the standard set of editor features, with a few bells and
whistles thrown in. It has dynamic 'as-you-type' searching, substition by
block/global/direction of a string or exact word match (i.e. change
variable i to i_var throughout your code), block copy/move/kill in normal
or rectangular region modes, marks that allow intra-buffer movement, easy
movement between buffers, a trash can, variable skipping (point to a
variable, press ^n to skip to the next occurrance of it, ^p for previous),
configuration save (you can leave a multi-buffer edit session and come
back to the exact place you left off with all buffer positions restored),
built in help, prompting for the beginning user, one-key buffer cloning,
and more.

So what doesn't FOLDED do? Well, the first release will not have an UNDO
feature. The trash can will have to do. It also is dual mode, something
which may alienate some of the Emacs users out there (I am bi-editable,
since I use Emacs on some systems, graphic shell editors on others, vi
on still others). With a little work, the editor could actually be made
to work either as a dual mode editor or a single mode editor (hint to
hackers out there). It also could use some better screen optimization,
since I tended to develop it on high speed displays. It is usable at
2400 buad, but any slower than that and you are better off with ed.
The first release also won't have any key binding capability. I still
haven't decided upon the best approach to handle user-defined key maps.
Finally, FOLDED isn't a layered editor. My goals during the development
of FOLDED were to keep it relatively small (it is about 60K executable,
<5000 lines of source) and fast, so layering as implemented in emacs-like
editors was out.

I wrote FOLDED simply to learn.  Early on, I made a choice to 1) put
basic 'must-have' functionality in the editor, 2) put advanced concepts
into the editor that can't be handled by external programs (folding and
relational features), 3) implement a simple to use and understand multi
file editing capability so that moving between files and scratchpads
was easy, 4) design the editor so that I could add enhanced 'wish-list'
functionality as I determined I needed it, 5) ensure that the final
version of the editor would function the same on my Atari ST at home as
it does on the Sun workstations at work and even as it does on the
mainframes, and 6) keep the program small (I hate long loading times).
The latter requirement somewhat dictates the earlier ones...at least
for the first release. I imagine that the editor will continue to
grow unbounded in the future, since it is just too easy to add features
as I feel a need to. One thing that I will eventually look into is a
bindable user macro feature (not interpreted like in emacs, but really
and truly bound into the executable.) The reason will be simply speed,
speed, and more speed :-).

FOLDED differs from both emacs and vi in several ways. First, it is an
unconstrained editor. You can move your cursor anywhere on the screen and
when you go into insert mode (or copy a rectangular block of text) the editor
will take care of sticking in needed spaces. Another difference is that the
editor does not 'wrap' lines. Instead, you can scroll/page the entire window
sideways when necessary (the editor will do this for you if you are typing
past the end of the right margin). I got too used to this feature on the
Apollos, I guess. Also, FOLDED does not retain tab characters. Rather, they
are replaced with white space on input and FOLDED will optionally 'optimize'
your file when you write it out (i.e. inserting tabs wherever it can, removing
trailing spaces on lines, etc). Actually, FOLDED doesn't even retain leading
spaces in the way you might expect an editor to store them. All leading spaces
in a line are compressed to an integer variable upon file read-in. The user
still sees the spaces on the screen, but internally this sped up my folding
logic (at the cost of almost going insane over the added code complexity...
there is always a tradeoff).

The first release of FOLDED has been tested on BSD4.2/3 (Suns, Symetric,
and Apollo), SYSVR3 (ETA-10), and the ATARI ST. It is being ported by
others to the Mac and the IBM PC. A preliminary port of the code to the
BELLCORE MGR (window manager) environment took only two hours on a Sun
workstation and provided individual windows for each buffer (with
title bars).

The following sections discuss the concepts of folding, relations, and
a little about the primary commands and user environment of FOLDED.
Although most of the examples will be source code related, I should mention
that I find myself using folding and relations in my documentation and daily
work plans and status reports even more than I do while coding.


FOLDING:

Obviously, a fundamental concept in using FOLDED is that of folding.
Folding in a text editor is a mechanism that allows you to cause
sections of code or text be replaced by a single line of text. This
is useful for 'hiding' away clean and functional sections of code
while you work on and only see the parts that you are currently
debugging. It also proves useful for tucking away big comment blocks.

Worried that folding will change your source code? Don't worry, the
act of folding doesn't affect the edited file at all...even when saved.
When you save a file that has folds in it, a seperate control file
is also saved under your current directory. See the section in the
documentation titled 'CONTROL FILES' for more info on this.

In FOLDED, I implemented three types of folding. In each case, a
segment of your file is reduced such that all you see is the first
line of the segment in reverse image (or bold, depending on your
standout terminal mode). First, is the simple 'fold a marked block'.
This allows you to mark any arbitrary region of text and fold it away.
For example, lets say we had started with the following text (the
numbers are just for reference in this documentation):

 1: main()
 2: /* GID for main - Just a test case
 3:    Author: Todd Burkey
 4:    Date:   1/27/89
 5:    DESC:   Methodology for a test case
 6: */
 7: {
 8:   if(this==that) {
 9:	do_something_that_works();
10:	}
11:   }

By defining a region between lines 2 and 6 and then folding, what you
would see on the screen would be:

 1: main()
 2: /* GID for main - Just a test case	      <-pretend this line is bold
 7: {
 8:   if(this==that) {
 9:	do_something_that_works();
10:	}
11:   }

Another method of folding is that based on 'indent level'. The logic
behind indent level folding is simply to start with the line the user
is currently on, move down in the text until an indent occurs, and
then continue down until an undent occurs that is less than or equal
to the original lines 'indented' position. For example, if you
had been on line 8 in the above 'folded' text when the 'f' key was
pressed, you would see the following:

 1: main()
 2: /* GID for main - Just a test case	      <-pretend this line is bold
 7: {
 8:   if(this==that) {			      <-pretend this line is bold
11:   }

Unfortunately, most source code contains lots of things that tend to
break up the clean indentation levels, so I added a third folding
mechanism. This one is a little smarter in that it does a 'logical
fold'. A logical fold looks for logical block begin/end marks in the
code (the braces in C) and skips around comments, strings, etc. For
example, if you had pressed the 'F' key on line 1 in any of the above
steps, you would end up with:

  1: main()			      <-pretend this line is bold

In all cases, unfolding the text is performed by simply moving to the
folded line and pressing the 'f' or 'F' key.

Folding isn't all that novel a concept. It has been used for years now
in thought processor programs on the Mac, IBM PC, and even back in the
CP/M days. Thought processors usually use the 'fold by indent level'
technique. Take the following simple example:

FOLDED Information
  * specifications    <-pretend this line is bold
  * plans	      <-pretend this line is bold
  * problems	      <-pretend this line is bold

This is all I see when I use FOLDED to call up my folded.info file.
Now, if I move to line 3 and press the f key, I will see:

FOLDED Information
  * specifications    <-pretend this line is bold
  * plans
    - Documentation   <-pretend this line is bold
    - Enhancements    <-pretend this line is bold
    - Bullet-proofing <-pretend this line is bold
  * problems	      <-pretend this line is bold

In summary, folding is fully heirarchical (i.e. a fold can contain any number
of folds), it is smart (it can fold based on programming logic blocks and
indentation level), and it is retained from one edit session to the next.
The actual commands (and their associated key bindings) that involve folding
are:

  indent fold (f) : fold based on indent level logic
  smart  fold (F) : fold based on {} logic
  unfold (f or F) : unfold the fold under the cursor
  temp unfold (^f): temporarily unfold all folds (for searching/sub'ing)
  re-fold all (^f): yep, CTRL-f is just a toggle
  block fold  (bf): fold current block region


RELATING:

The word 'relations' in the field of databases conjures up images of opening
up a large database, finding a particular entry, moving to a particular
field, and, with several keystrokes, calling up not only another database,
but a specific record in that database. Now, I have never understood why
we couldn't have similar capabilities in a text editor. We can envision
having a document file, moving to a specific part of the document, calling
up the source code file related to that part of the document, and then
moving to the appropriate spot in the source file. We can even imagine
moving through the source code and finding a structure (or variable) and
wishing we could immediately view the definition of the structure. These
scenarios are fundamentally the same as those handled by a relational
database program.

Relationships as pertains to editors (and, of course, in this context, FOLDED)
primarily involve two concepts.  First, there are word relationships.  These
are something like 'glossary' entries in that you define a word once in a file
and then can immediately go to the exact place you defined it from anywhere in
the file (or any other file you are currently editing.) From a programming
viewpoint, this simply means that you can move your cursor over a structure
name in a source file and with a few keystrokes be placed immediately in an
include file at the place the structure is defined.  A second relationship type
is that of line relations.  These are basically invisible links that relate a
line in one buffer to a line in another buffer.  This type of relationship
allows you to easily move back and forth between source code and program
outlines/documents.  Folded will even open up a file if you try to go to a
relation in a file that you aren't currently editing.

FOLDED supports a number of commands that allow easy creation and use of
these relationships. The following key sequences are used to perform the
most common relation operations:

 Rsl (Relate Set Line): The current line becomes 1/2 of a line relationship.
			The next Rs command will complete the relationship.
 Rsw (Relate Set Word): The current word is indexed for reference.
 Rgl (Relate Goto Line):FOLDED looks at the first line with a relation on or
			above the current line and goes to its related line.
			If the related line is in a file not currently being
			edited it will be opened for viewing.
 Rgw (Relate Goto Word):Read the word under the cursor and move to the correct
			file that it was defined in.
 Rkl (Relate Kill Line):Kill the relationship for the current line.
 Rkw (Relate Kill Word):Kill the relationship for the current word.
 Rp  (Relate Pop):	Pop back to where you were before the last Relate Goto.
			The pop stack is currently 10 levels deep.

As with folding, the relationship information does not affect the actual
files you are editing. In fact, the information is simply saved out as part
of the folding info files. It is also important to note that neither folding
or relations affect the speed of loading, saving, or using the various editor
features to a noticable extent.

Since I added relations very late in the development of the editor, I have a
feeling that this area will see the most improvement over the next year. I've
already decided that I need a few more commands to do things like 'move'
and 'view' relationships.


EDITING COMMANDS:

When you enter FOLDED the first time, press the ? key to see a cheat sheet
that tells you what each key does. Sorry, haven't had time to document the
commands in detail yet...

The following is a roughly what you will see on the cheat sheet screen
(less the standout boxes):


		    Top level cheat sheet for FOLDED.
     Where appropriate, additional prompts will appear in the program.
MOVEMENT: k-up, j-down, h-left, l-right, K/J/H/L for page movement
	  ^/$-beginning/end of line, p-Toggle paging mode, g-goto a line
READING : :r/:v/:i-read/view/insert file, ^r/^v-read/view under cursor
WRITING : :w-write current buffer/block, :s-save changed buffers/config
INFO	: ?-this help, '='-local information, X-extended buffer info
SEARCHES: ^n/^p-find next/prev word, /-forward, \-reverse, ^e-exactmode
BUFFERS : t/T-toggle buffers/trash, ^g/' '-goto/Cycle through buffers
	  c/C  Create scratch/clone buf, ^b-tag buffer for moves/copies
	  '-k'-kill current buffer, y-yank line from trash and insert
BLOCKS	: b-Go to block options (define start/end, moving, copying, etc)
CHANGING: s-change one char, S-substitute a string/word (once or global)
DELETING: d/D-delete a character/line, -v/-d: global delete/vdelete
INSERTS : i/I-insert before char/line, a/A-append after char/line
FOLDING : f-fold by indentation, F-fold by logic, bf-fold a defined block
	  ^f-toggle unfoldall, NOTE: press f or F on a fold to unfold it
RELATES : R-line, word, and error relation options (set, goto, kill, pop)
MARKING : m-mark options (ms-set, mg-goto, md-delete, mc-clear all)
EXTRAS	: Q-quit, !-run program/shell, x-execute macro (unimplemented)
			--Hit any key to continue--


BACKUP PROTECTION:

If FOLDED detects a local directory called febackup, then backup copies of
anything you edit will be placed in that directory. A rename operation is
performed on the original file, so this means that only local partition files
will be backed up in this fashion. Backups only take place at the time you
save a file the first time during an edit session, so no unneccessary backups
are made.


FOLDED CONTROL FILES:

If folded detects a local directory called foldinfo, then FOLDED will store
all its control files there rather than clutter your directory with them. A
control file is created only when the file being edited is saved and has some
folding or relations in it. The control file will be a file of the same name,
but with an 'f' inserted immediately after the final '.' in the name...so
'test.c' has a control file called 'test.fc' and 'junk' would have 'junk.f'.


ENVIRONMENT VARIABLES:

Several Environment variables may be used if you wish to avoid some prompting
by FOLDED. They are:

  FE_ASCII   - Save control files in ASCII (easier to mail that way)
  FE_BINARY  - Save control files in BINARY (a bit faster and more effecient).
  FE_NOTABS  - Always write files without tabs (useful on the Apollos).
  FE_TABS    - Always fully optimize the files when writing.
  MGR_FONT   - Size of the font used under MGR for the windows.


IMPLEMENTATION DETAILS:

  Folded uses a doubly linked list to store lines internally. It mallocs
the exact amount of memory a line needs on read-in, but as you modify lines
it will malloc additional memory in 4 byte chunks...so you can end up with
some wastage if you edit a lot of lines in a file. FOLDED keeps track of a
lot of information about lines in the buffers, so there is quite a bit of
overhead in using FOLDED. I place it pretty close to emacs in the amount
of overhead required for 'data' storage, but since the executable is only
about 60K it wipes emacs away in raw loading time (program+files). I haven't
had time to optimize the code for speed yet, but things like searching and
substituting are significantly faster than emacs and slightly faster than
vi. This is partly due to the design of the editor and also may be due to
FOLDED's lack of regular expression parsing routines.

  The first version of FOLDED probably won't be heavily optimized in the sense
that ints will probably still be used where chars may get by. Also, there
may be tricks I don't know about that could compress the amount of overhead
I incurr in the storage of lines, buffers, etc.

  -Todd Burkey