QED for Unix, UTF-8/Unicode aware.

This port is now obsolete. All development has moved to the new port phonologus/qed-new.

This is a port of the University of Toronto Unix version of the QED editor, which was originally written by Rob Pike, David Tilbrook, Hugh Redelmeier and Tom Duff. I have updated it to work with UTF-8 and Unicode.

The behaviour of this qed when displaying Unicode is entirely dependent on the capabilities of the terminal it is running under. In particular, qed outputs Unicode in Logical Order, leaving the handling of the complexities of rendering bi-directional text and wide-characters on-screen to the terminal. On macOS, the Terminal.app program is quite successful with displaying bi-directional text, whereas xterm on Linux sticks to Logical Order.

This port was developed on a modern Linux (Fedora 32), and macOS. It depends on just a few low-level Unix system calls, so should be reasonably portable across any Unix-like OS. I have built this qed successfully on:

• Linux (Fedora 32, gcc)
• macOS Intel (Catalina 10.15 with Commandline Developer Tools, clang)
• macOS Apple Silicon (Big Sur 11.2 with Commandline Developer Tools, clang)
• OpenBSD (6.7, both gcc and clang)

The sources that this port began with were scavenged from Arnold Robbins's qed-archive. Specifically they are from the unix-1992 sub-directory, which is from a tarball of Rob Pike's original sources of the University of Toronto version of QED.

The updated Qed tutorial and manpage in the doc subdirectory are essential reading.

The tutorial is a heavily edited an updated version of Rob Pike's original tutorial which was already very out-of-date, even before this port began. Rob's original tutorial was written in troff using the manpage macros. The updated tutorial has been translated into asciidoc, which makes for a very pretty html and pdf runoff, as well as a traditional manpage format.

A brand new tutorial for a modern audience would be a nice thing to have. Conspicuously missing from the tutorial here are: introduction to line-oriented editing; browsing buffers; Unicode input; the xform edit mode; up-to-date examples (qed programs to streamline a git work-flow, for instance). I will be writing a new Qed Book from scratch out-of-tree, so the updated tutorial here will not receive any more revisions.

Unlike the original tutorial, the original manpage was pretty much up-to-date, except for the new Unicode functionality. I have updated the manpage so that it accurately describes this qed. The new functionality is also described below.

The Makefile provides a simple way to build a qed binary. make clean && make will build a fully stand-alone binary qed, which can be installed anywhere. make install will install the binary, the manpage, and the tutorial. The tutorial is installed as qed-tutorial(7), and can be summoned with man qed-tutorial, or man 7 qed-tutorial.

Edit the Makefile to change where this all goes.

To regenerate the html, pdf, and manpage versions of the tutorial, cd doc/src && make && make release. This depends on an asciidoctor and asciidoctor-pdf toolchain being installed. The pdf version of the actual manpage, qed(1), requires ps2pdf to regenerate.

The doc/historical/read.me file is Rob Pike's, accompanying his original tarball. It provides an interesting snapshot into the work-in-progress at the time, and the relationship between the QED and ED editors.

The q subdirectory contains Rob's library of useful QED programs. Again a master-class in themselves, and they can be fruitfully studied after working through the tutorial.

Differences introduced in this port

• This qed consumes and outputs UTF-8 Unicode. Internally, all "characters" are stored and manipulated as Unicode code-points represented as int-s. The full Unicode range is supported, thus even emojis and other exotica beyond the Basic Multilingual Plain can appear, even in regexes.

• In list mode, this qed displays non-printing codepoints in the ASCII range as \xhh. Additionally, it displays non-ASCII code-points in the Basic Multilingual Plane as \uhhhh, and code-points beyond the BMP as \Uhhhhhh.

• This qed adds the special characters \x, \u, \U which can be used to enter (valid) Unicode codepoints directly, using (lowercase) hexadecimal. For example \x53, will insert ASCII 0x53 (S); \u05d0 will insert Hebrew Aleph (א); and \U01f600 will insert a grinning emoji (😀). The format corresponds to the new list mode format described above. If the wrong number of digits is supplied, or the resulting codepoint falls outside the legitimate range, this qed generates a ?U error.

• The program qedbufs referred to in the original manpage is not ported, and likely won't be.

• When reading in a file which does not terminate in a newline, this qed supplies the newline, and notifies the user with the message ?N: appended. This matches the Research Unix Version 10 ed and the Plan9 ed behaviour. Previously QED would throw an ?N error, and discard any input from the last newline to the end of file.

• QED's xform command (x, described in the manpage), allows a novel form of character-by-character line editing. It is very impressive with single-width, left-to-right, text. Any other text is likely to have very unpredictable, terminal-dependent, and unintended results.