Skip to content

Instantly share code, notes, and snippets.

@antonijn
Last active March 20, 2017 03:19
Show Gist options
  • Save antonijn/0646f0f172fc3dd1958dc30378cfc2c3 to your computer and use it in GitHub Desktop.
Save antonijn/0646f0f172fc3dd1958dc30378cfc2c3 to your computer and use it in GitHub Desktop.
antonijn, 2017
___ _ _ _ _
|_ _|_ _| |_ _ _ ___ __| |_ _ __| |_(_)___ _ _
| || ' \ _| '_/ _ \/ _` | || / _| _| / _ \ ' \
|___|_||_\__|_| \___/\__,_|\_,_\__|\__|_\___/_||_|
Because all other text editors suck (although
some have at least a few good ideas), I decided
to write my own. Its working name is 'te'. It
combines ideas from vim, acme and sam.
If you have any ideas or suggestions, please
contact me through one of the contact options
given at https://antonijn.github.io/.
_ ___ _ _
/ | ___ | _ ) __ _ __(_)__ __ ___ _ _ __ ___ _ __| |_
| | |___| | _ \/ _` (_-< / _| / _/ _ \ ' \/ _/ -_) '_ \ _|
|_| |___/\__,_/__/_\__| \__\___/_||_\__\___| .__/\__|
|_|
1.1 STRUCTURAL NAVIGATION
It's supposed to be a hybrid structure/text-editor, with
insertion/deletion like a text editor (think: vim), but navigation like
a structure editor (think: acme but with key commands for structure
navigation).
Take for instance the following line:
foo = bar(qux(), quux());
^
You realise that 'quux()' needs an argument, namely 'NULL'. Your cursor
is currently at '^'.
You press a key to go into "parameter selection mode" (say 'p'):
foo = bar(qux(), quux());
^~~~~
You traverse to the next parameter (say '.'):
foo = bar(qux(), quux());
^~~~~~
You press the parameter selection key again:
foo = bar(qux(), quux());
^
You enter insert mode and type 'NULL' (say 'iNULL'):
foo = bar(qux(), quux(NULL));
|
My hypothesis is that this will be much more intuitive than traditional
word/ paragraph/search based navigation.
1.2 PIPING
Instead of providing a myriad of advanced commands like vim (every key has
on average more than two functions) or emacs (https://xkcd.com/378/), most
of te's commands should be based on navigation, not on actual editing.
Editing should be done by external programs, through which you can pipe
selections (akin to vim's «:'<,'>!program-name'» ex-command, but about
five orders of magnitude less clunky).
The editor would come with a variety of programs with short and sweet
names (they won't be added to your PATH by default, don't worry), which
function like traditional vim/emacs commands/functions would. Say you've
been doing too much C#/Java programming and accidentally typed 'null'
instead of 'NULL'. In vim you'd enter visual mode, select the text and
press '~'. In te, you select the text, press something to indicate you
want to type a command (say <C-x>), a small non-intrusive pop-up text
bar appears right below the selection to minimise eye movement *cough* vi
ex-mode *cough*, in which you can write 'up', press enter, and that's it.
There's another program 'lic' which simply spits out license text,
commented in a language of your choice, which you can run at the top
of your document. There would be another program 'u', which outputs
a Unicode character based on its arguments: 'u ae' -> 'æ', etc. For
find/replace there's 'sed'. You could use 'fmt' to format plain text
documents.
This would greatly simplify the core editor, thereby adhering to the Unix
philosophy. It would also allow the user to quickly write the programs
they need themselves, and combine them into suites to redistribute.
Most programs delivered with the program would probably be written
in Python.
___ ___ _ _ _
|_ ) ___ | \ ___| |_ __ _(_) |___
/ / |___| | |) / -_) _/ _` | | (_-<
/___| |___/\___|\__\__,_|_|_/__/
2.1 STRUCURE NAVIGATION
Structure navigation poses a problem on malformed syntax. It could *maybe*
fall-back to an indentation-based syntax system for multiline structures,
but it is in general quite a challenge. Perhaps [debatable] the user would
instinctively revert to line/paragraphs/token/character-based navigation,
because those facilities are also provided.
2.2 UI
There would be the compile-time choice of including a GTK+-3 GUI
interface and/or an NCurses interface. The GTK+-3 interface being my
preferred choice. I think using a GUI would greatly decrease terminal
lag and could remove a lot of weight from the feeling of the editor.
2.2.1 Colours
The colour scheme would of course be customisable, but by default its
insert mode colours would be a light solarized theme and its normal
mode colours would be icy blue. This allows the user to know quickly
what mode their in, even when they've alt-tabbed. Technically speaking
vim also allows you to see your mode by putting a tiny '-- INSERT --'
label to the very bottom of your screen, where you never ever look -
save for when you're in ex-mode when '-- INSERT --' is not displayed...
I've considered more subtle changes like changing the caret shape, but
I think block cursors are unintuitive (a cursor is intuitively between
two characters), and that change would perhaps be too subtle, since you
rarely _really_ look at a cursor anyway. The colour change idea appears
to be subtle enough in practice [debatable].
The whole editor should feel intuitive and shouldn't feel like everything
takes a lot of effort, as is my feeling about vim. One extremely important
principle is therefore minimising eye movement. This is respresented
through the "colour change indicating mode switch" idea, and by placing
the command field pop-up as close to the cursor as possible - or at
least to the place where the user is probably looking [debatable].
The command field pop-up will also expand to show stderr output (stylized
as a warning if status=0, or as an error otherwise). The pop-up can
be exited either by entering a command that executes flawlessly, or
by pressing the same key combination that gets you from insert mode to
normal mode.
The text editor will provide syntax highlighting [semi-longterm].
2.2.2 Tiling
I think the text editor should implement a tiling system very similar
to vim's [semi-longterm].
2.2.3 Built-in Terminal Tiles
Not for now. I think that <C-z> and <Alt-\t> should do for now. Besides,
built-in terminals tend not to mix well with the concept of a modal editor
(look at NeoVim) [debatable].
2.3 PIPING
Of course one command should allow piping of the current selection (and
I see no reason why that shouldn't also be allowed in insert mode for
output-only commands). However, some terminal commands could possibly
also be used for information or navigation. For instance: grep could
show the location of a regular expression - which the text editor could
use to navigate to, wc could show the amount of characters currently in
the document.
It would be cumbersome to have your entire selection replaced by the
output of these commands. It would instead be nicer if these commands
could use another command prompt, using a key combination different
from but similar to the regular piping prompt command. How the output
of these commands could then be used for navigation, I don't know yet.
2.4 CUSTOMIZATION
Internal editor settings should be minimal, but colour schemes and
possibly key combinations etc. would be modified either through Lua code
or simple config files.
2.5 PLUGINS
Despite the use of external programs through piping, some plugins may be
desired [debatable]. For instance, a plugin may be written to connect
the editor to a syntax-completion engine. I haven no idea how I would
go about implementing such a plugin system [longterm].
____ ___ _ _ _ _
|__ / ___ |_ _|_ __ _ __| |___ _ __ ___ _ _| |_ __ _| |_(_)___ _ _
|_ \ |___| | || ' \| '_ \ / -_) ' \/ -_) ' \ _/ _` | _| / _ \ ' \
|___/ |___|_|_|_| .__/_\___|_|_|_\___|_||_\__\__,_|\__|_\___/_||_|
|_|
3.1 LANGUAGE & TOOLKIT
te will be written in C. The helper programs will be written in Python/Lua
and Bash/Sh. The POSIX APIs will be used for piping etc.
GTK+-3.0 is used for basic cross-platform windows, Cairo for widget
drawing and Pango for text rendering. There is of course also the
NCurses interface.
3.2 LICENSE
The core editor will be GPLv2 licensed. Most text manipulation programs
will probably be ISC/MIT licensed.
3.3 BUFFER
Text buffers are done using the gap buffer idea as is done in emacs
(https://www.wikipedia.org/wiki/Gap_buffer). This means okay memory
consumption and hopefully unnoticable logical cursor movement (the screen
cursor and logical cursor may not always coincide).
Piping can be done in place, piping text to a program while simultaneously
trying to catch up with said programs output.
3.4 UNDO/REDO
There will be an infinite capacity undo buffer like in vim. This can be
done space-efficiently in the following way [debatable].
The undo/redo tree is split in two sections, the past and the future:
the past being the undo tree and the future being the redo tree. The
past tree's nodes would merely contain the _locations_ of where text was
entered, and the full text that was deleted, in a diff-like format. The
future tree's nodes are exactly the opposite: the locations of where
text was removed and the precide text that was inserted. This is all
that is logically needed to reconstruct any past/future from the present.
Note that two past nodes (or two future nodes for that matter) may be
merged together (not by the user, but internally, when a new past node
is being constructed).
3.5 SYNTAX HIGHLIGHTING
Syntax highlighting by regular expressions simply doesn't work, and is
probably very slow. Hand-writing parsers on the other hand takes an awful
lot of effort, and may inhibit customization. Perhaps just delimiter/word
based highlighting to start off with.
@Masood-Lapeh
Copy link

Hey!
Interesting ideas..
I have one advice.. Try to be mindful of Bi-directional texts (and context-sensitive shaping) from beginning.. some people need it..
Vim doesn't support it AFAIK,
Emacs support is not good (many little annoying bugs),
Atom support is not good (you can see bi-di text but it's very hard to edit right to left parts).
Some simple text editors like gedit or Notepad support BiDi texts nicely..
I think pango supports bidirectional text automatically..
Good luck, have fun!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment