notcurses

blingful TUI library for modern terminal emulators. definitely not curses.

• birthed screaming into this world by nick black (nickblack@linux.com)
• C++ wrappers by marek habersack (grendel@twistedcode.net)

for more information, see dankwiki and the man pages. There's also a reference in this repo.

In addition, there is Doxygen output. There is a mailing list which can be reached via notcurses@googlegroups.com.

I wrote a coherent guidebook, which is available for free download, or paperback purchase.

Notcurses is available in the Arch AUR, and from Debian Unstable. Packages for Fedora 33 and Ubuntu Focal are available from DSSCAW.

• Introduction
• Requirements
  • Building
• Included tools
• Differences from NCURSES
  • Features missing relative to NCURSES
  • Adapting NCURSES programs
• Environment notes
  • TrueColor detection
  • Fonts
  • FAQs
• Supplemental material
  • Useful links
  • Other TUI libraries
  • History
  • Thanks

Introduction

• What it is: a library facilitating complex TUIs on modern terminal emulators, supporting vivid colors and Unicode to the maximum degree possible. Many tasks delegated to Curses can be achieved using Notcurses (and vice versa).

• What it is not: a source-compatible X/Open Curses implementation, nor a replacement for NCURSES on existing systems.

Notcurses abandons the X/Open Curses API bundled as part of the Single UNIX Specification. The latter shows its age, and seems not capable of making use of terminal functionality such as unindexed 24-bit color ("TrueColor", not to be confused with the 8-bit indexed 24-bit "extended color" of NCURSES). For some necessary background, consult Thomas E. Dickey's superb and authoritative NCURSES FAQ. As such, Notcurses is not a drop-in Curses replacement. It is almost certainly less portable, and definitely tested on less hardware. Sorry about that. Ultimately, I hope to properly support all terminals supporting the features necessary for complex TUIs. I would argue that teletypes etc. are fundamentally unsuitable. Most operating systems seem reasonable targets, but I only have Linux and FreeBSD available for testing.

Whenever possible, Notcurses makes use of the Terminfo library shipped with NCURSES, benefiting greatly from its portability and thoroughness.

Notcurses opens up advanced functionality for the interactive user on workstations, phones, laptops, and tablets, at the expense of e.g. some industrial and retail terminals.

Why use this non-standard library?

• Thread safety, and efficient use in parallel programs, has been a design consideration from the beginning.

• A svelter design than that codified by X/Open:

  • Exported identifiers are prefixed to avoid common namespace collisions.
  • The library object exports a minimal set of symbols. Where reasonable, static inline header-only code is used. This facilitates compiler optimizations, and reduces loader time.

• All APIs natively support the Universal Character Set (Unicode). The cell API is based around Unicode's Extended Grapheme Cluster concept.

• Visual features including images, fonts, video, high-contrast text, sprites, and transparent regions. All APIs natively support 24-bit color, quantized down as necessary for the terminal.

• It's Apache2-licensed in its entirety, as opposed to the drama in several acts that is the NCURSES license (the latter is summarized as "a restatement of MIT-X11").

Much of the above can be had with NCURSES, but they're not what NCURSES was designed for. The most fundamental advantage in my mind, though, is that Notcurses is of the multithreaded era. On the other hand, if you're targeting industrial or critical applications, or wish to benefit from the time-tested reliability and portability of Curses, you should by all means use that fine library.

Requirements

• (build) A C11 and a C++17 compiler
• (build) CMake 3.14.0+
• (build+runtime) From NCURSES: terminfo 6.1+
• (OPTIONAL) (build+runtime) From QR-Code-generator: libqrcodegen 1.5.0+
• (OPTIONAL) (build+runtime) From FFMpeg: libswscale 5.0+, libavformat 57.0+, libavutil 56.0+
• (OPTIONAL) (build+runtime) OpenImageIO 2.1+
• (OPTIONAL) (testing) Doctest 2.3.5+
• (OPTIONAL) (documentation) pandoc 1.19.2+
• (OPTIONAL) (python bindings): Python 3.7+, CFFI 1.13.2+
• (OPTIONAL) (rust bindings): rust 1.40.0+, cargo 0.40.0+, bindgen 0.53.0+
• (runtime) Linux 5.3+

Building

• Create a subdirectory, traditionally build. Enter the directory.
• cmake ... You might want to set e.g. CMAKE_BUILD_TYPE.
• make
• make test

The default multimedia engine is FFmpeg. You can select a different engine using USE_MULTIMEDIA. Valid values are ffmpeg, oiio (for OpenImageIO), or none.

If you have unit test failures, please file a bug including the output of ./notcurses-tester > log 2>&1 (make test also runs notcurses-tester, but hides important output).

To watch the bitchin' demo, run ./notcurses-demo -p ../data. More details can be found on the notcurses-demo(1) man page.

Included tools

Six binaries are installed as part of notcurses:

• notcurses-demo: some demonstration code
• notcurses-view: renders visual media (images/videos)
• notcurses-input: decode and print keypresses
• notcurses-planereels: play around with ncreels
• notcurses-tester: unit testing
• notcurses-tetris: a tetris clone

To run notcurses-demo from a checkout, provide the tests/ directory via the -p argument. Demos requiring data files will otherwise abort. The base delay used in notcurses-demo can be changed with -d, accepting a floating-point multiplier. Values less than 1 will speed up the demo, while values greater than 1 will slow it down.

notcurses-tester expects ../tests/ to exist, and be populated with the necessary data files. It can be run by itself, or via make test.

Differences from NCURSES

The biggest difference, of course, is that notcurses is not an implementation of X/Open (aka XSI) Curses, nor part of SUS4-2018.

The detailed differences between notcurses and NCURSES probably can't be fully enumerated, and if they could, no one would want to read them. With that said, some design decisions might surprise NCURSES programmers:

• There is no distinct PANEL type. The z-buffer is a fundamental property, and all drawable surfaces are ordered along the z axis. There is no equivalent to update_panels().
• Scrolling is disabled by default, and cannot be globally enabled.
• The Curses cchar_t has a fixed-size array of wchar_t. The notcurses cell instead supports a UTF-8 encoded extended grapheme cluster of arbitrary length. The only supported charsets are ANSI_X3.4-1968 and UTF-8.
• The hardware cursor is disabled by default, when supported (civis capability).
• Echoing of input is disabled by default, and cbreak mode is used by default.
• Colors are always specified as 24 bits in 3 components (RGB). If necessary, these will be quantized for the actual terminal. There are no "color pairs".
• There is no distinct "pad" concept (these are NCURSES WINDOWs created with the newpad() function). All drawable surfaces can exceed the display size.
• Multiple threads can freely call into notcurses, so long as they're not accessing the same data. In particular, it is always safe to concurrently mutate different ncplanes in different threads.
• NCURSES has thread-ignorant and thread-semi-safe versions, trace-enabled and traceless versions, and versions with and without support for wide characters. notcurses is one library: no tracing, UTF-8, thread safety.
• There is no ESCDELAY concept; notcurses expects that all bytes of a keyboard escape sequence arrive at the same time. This improves latency and simplifies the API.
• It is an error in NCURSES to print to the bottommost, rightmost coordinate of the screen when scrolling is disabled (because the cursor cannot be advanced). Failure to advance the cursor does not result in an error in notcurses (but attempting to print at the cursor when it has been advanced off the plane does).

Features missing relative to NCURSES

This isn't "features currently missing", but rather "features I do not intend to implement".

• There is no support for soft labels (slk_init(), etc.).
• There is no concept of subwindows which share memory with their parents.
• There is no tracing functionality ala trace(3NCURSES). Superior external tracing solutions exist, such as bpftrace.

Adapting NCURSES programs

Do you really want to do such a thing? NCURSES and the Curses API it implements are far more portable and better-tested than notcurses is ever likely to be. Will your program really benefit from notcurses's advanced features? If not, it's probably best left as it is.

Otherwise, most NCURSES concepts have clear partners in notcurses. Any functions making implicit use of stdscr ought be replaced with their explicit equivalents. stdscr ought then be replaced with the result of notcurses_stdplane() (the standard plane). PANELs become ncplanes; the Panels API is otherwise pretty close. Anything writing a bare character will become a simple cell; multibyte or wide characters become complex cells. Color no longer uses "color pairs". You can easily enough hack together a simple table mapping your colors to RGB values, and color pairs to foreground and background indices into said table. That'll work for the duration of a porting effort, certainly.

I have adapted two large (~5k lines of C UI code each) programs from NCURSES to notcurses, and found it a fairly painless process. It was helpful to introduce a shim layer, e.g. compat_mvwprintw for NCURSES's mvwprintw:

static int
compat_mvwprintw(struct ncplane* nc, int y, int x, const char* fmt, ...){
  va_list va;
  va_start(va, fmt);
  if(ncplane_vprintf_yx(nc, y, x, fmt, va) < 0){
    va_end(va);
    return ERR;
  }
  va_end(va);
  return OK;
}

These are pretty obvious, implementation-wise.

Environment notes

• If your terminal has an option about default interpretation of "ambiguous-width characters" (this is actually a technical term from Unicode), ensure it is set to Wide, not narrow. If that doesn't work, ensure it is set to Narrow, heh.

• If you can disable BiDi in your terminal, do so while running notcurses applications, until I have that handled better. notcurses doesn't recognize the BiDi state machine transitions, and thus merrily continues writing left-to-right. Likewise, ultra-wide glyphs will have interesting effects. ﷽!

• The unit tests assume dimensions of at least 80x24. They might work in a smaller terminal. They might not. Don't file bugs on it.

TrueColor detection

notcurses aims to use only information found in the terminal's terminfo entry to detect capabilities, TrueColor being one of them. Support for this is indicated by terminfo having a flag, added in NCURSES 6.1, named RGB set to true. However, as of today there are few and far between terminfo entries which have the capability in their database entry and so TrueColor won't be used in most cases. Terminal emulators have had for years a kludge to work around this limitation of terminfo in the form of the COLORTERM environment variable which, if set to either truecolor or 24bit does the job of indicating the capability of sending the escapes 48 and 38 together with a tripartite RGB (0 ≤ c ≤ 255 for all three components) to specify fore- and background colors. Checking for COLORTERM admittedly goes against the goal stated at the top of this section but, for all practical purposes, makes the detection work quite well today.

Fonts

Fonts end up being a whole thing, little of which is pleasant. I'll write this up someday FIXME.

FAQs

• Q: Why didn't you just use Sixel?

• A: Many terminal emulators don't support Sixel. Sixel doesn't work well with mouse selection. With that said, I do intend to support Sixel soon, as a backend, when available, for certain types of drawing.

• Q: I'm not seeing NCKEY_RESIZE until I press some other key.

• A: You've almost certainly failed to mask SIGWINCH in some thread, and that thread is receiving the signal instead of the thread which called notcurses_getc_blocking(). As a result, the poll() is not interrupted. Call pthread_sigmask() before spawning any threads.

• Q: One of the demos claimed to spend more than 100% of its runtime rendering. Do you know how to count?

• A: Runtime is wall clock time. A multithreaded demo can spend more than the wall-clock time rendering if the threads contend.

• Q: Using the C++ wrapper, how can I ensure that the NotCurses destructor is run when I return from main()?

• A: As noted in the C++ FAQ, wrap it in an artificial scope (this assumes your NotCurses is scoped to main()).

• Q: How do I hide a plane I want to make visible later?

• A: Either move it above and to the left of the screen (preventing resizes from making it visible), or place it underneath another (opaque) plane (the latter performs better).

• Q: Why isn't there an ncplane_box_yx()? Do you hate orthogonality, you dullard?

• A: ncplane_box() and friends already have far too many arguments, you monster.

• Q: Why doesn't Notcurses support 10- or 16-bit color?

• A: Notcurses supports 24 bits of color, spread across three eight-bit channels. You presumably mean 10-bit-per-channel color. Notcurses will support it when a terminal supports it.

• Q: You seem a creative guy. Why the least creative name ever?

• A: I really didn't expect this to go anywhere.

Supplemental material

Useful links

• BiDi in Terminal Emulators
• The Xterm FAQ
  • XTerm Control Sequences
• The NCURSES FAQ
• ECMA-35 Character Code Structure and Extension Techniques (ISO/IEC 2022)
• ECMA-43 8-bit Coded Character Set Structure and Rules
• ECMA-48 Control Functions for Coded Character Sets (ISO/IEC 6429)
• Unicode 12.1 Full Emoji List
• Unicode Standard Annex #29 Text Segmentation
• Unicode Standard Annex #15 Normalization Forms
• The TTY demystified
• Dark Corners of Unicode
• UTF-8 Decoder Capability and Stress Test

Useful man pages

• Linux: console_codes(4)
• Linux: termios(3)
• Linux: ioctl_tty(2)
• Linux: ioctl_console(2)
• Portable: terminfo(5)
• Portable: user_caps(5)

History

• 2020-05-10: Notcurses 1.4.0 "the saga continues".)
• 2020-04-19: Notcurses is accepted into Debian.
• 2020-04-12: Notcurses 1.3.0 "hypnotize".
  • A new hype video is never really publicized.
• 2020-04-08: The Notcurses book is published.
• 2020-03-23: Notcurses is featured on Linux World News.
• 2020-02-17: Notcurses 1.2.0 "check the résumé, my record's impeccable".
• 2019-01-19: Notcurses 1.1.0 "all the hustlas they love it just to see one of us make it". Much better video support, pulsing planes, palette256.
  • The new hype video gets a lot of attention.
• 2019-01-04: Notcurses 1.0.0 "track team, crack fiend, dying to geek" is released, six days ahead of schedule. 147 issues closed. 702 commits.
• 2019-12-18: Notcurses 0.9.0 "You dig in! You dig out! You get out!", and also the first contributor besides myself (@grendello). Last major pre-GA release.
• 2019-12-05: Notcurses 0.4.0 "TRAP MUSIC ALL NIGHT LONG", the first generally usable notcurses.
  • I prepare a demo, and release it on YouTube.
• November 2019: I begin work on Outcurses. Outcurses is a collection of routines atop NCURSES, including ncreels. I study the history of NCURSES, primarily using Thomas E. Dickey's FAQ and the mailing list archives.
  • 2019-11-14: I file Outcurses issue #56 regarding use of TrueColor in outcurses. This is partially inspired by Lexi Summer Hale's essay everything you ever wanted to know about terminals. I get into contact with Thomas E. Dickey and confirm that what I'm hoping to do doesn't really fit in with the codified Curses API.
  • 2019-11-16: I make the first commit to Notcurses.
• September 2019: I extracted fade routines from Growlight and Omphalos, and offered them to NCURSES as extensions. They are not accepted, which is understandable. I mention that I intend to extract ncreels, and offer to include them in the CDK (Curses Development Kit). Growlight issue #43 is created regarding this extraction. A few minor patches go into NCURSES.
• 2011, 2013: I develop Growlight and Omphalos, complicated TUIs making extensive use of NCURSES.

Thanks

• Notcurses could never be what it is without decades of tireless, likely thankless work by Thomas E. Dickey on NCURSES. His FAQ is a model of engineering history. He exemplifies documentation excellence and conservative, thoughtful stewardship. The free software community owes Mr. Dickey a great debt.
• Robert Edmonds provided tremendous assistance Debianizing the package, and David Cantrell did likewise for Fedora. Both are hella engineers.
• Justine Tunney, one of my first friends at Google NYC, was always present with support, and pointed out the useful memstream functionality of POSIX, eliminating the need for me to cons up something similar.
• I one night read the entirety of Lexi Summer Hale's essays, and woke up intending to write notcurses.
• NES art was lifted from The Spriters Resource and NES Sprite, the kind of sites that make the Internet great. It probably violates any number of copyrights. C'est la vie.
• Mark Ferrari, master of the pixel, for no good reason allowed me to reproduce his incredible and groundbreaking color-cycling artwork. Thanks Mark!
• Finally, the demoscene and general l33t scene of the 90s and early twenty-first century endlessly inspired a young hax0r. There is great joy in computing; no one will drive us from this paradise Turing has created!

“Our fine arts were developed, their types and uses were established, in times very different from the present, by men whose power of action upon things was insignificant in comparison with ours. But the amazing growth of our techniques, the adaptability and precision they have attained, the ideas and habits they are creating, make it a certainty that profound changes are impending in the ancient craft of the Beautiful.” —Paul Valéry