Turn liberty-tui into a terminal/X11 hybrid

Importing code from nncmpp, adjusting it to work with hex as well.

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2014 - 2022, Přemysl Eric Janouch <p@janouch.name>
+Copyright (c) 2014 - 2023, Přemysl Eric Janouch <p@janouch.name>
 
 Permission to use, copy, modify, and/or distribute this software for any
 purpose with or without fee is hereby granted.

liberty-tui.c

@@ -1,270 +0,0 @@
-/*
- * liberty-tui.c: the ultimate C unlibrary: TUI
- *
- * Copyright (c) 2016 - 2017, Přemysl Eric Janouch <p@janouch.name>
- *
- * Permission to use, copy, modify, and/or distribute this software for any
- * purpose with or without fee is hereby granted.
- *
- * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
- * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
- * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
- * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
- * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
- * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
- *
- */
-
-// This file includes some common stuff to build TUI applications with
-
-#include <ncurses.h>
-
-// It is surprisingly hard to find a good library to handle Unicode shenanigans,
-// and there's enough of those for it to be impractical to reimplement them.
-//
-//                         GLib          ICU     libunistring    utf8proc
-// Decently sized            .            .            x            x
-// Grapheme breaks           .            x            .            x
-// Character width           x            .            x            x
-// Locale handling           .            .            x            .
-// Liberal license           .            x            .            x
-//
-// Also note that the ICU API is icky and uses UTF-16 for its primary encoding.
-//
-// Currently we're chugging along with libunistring but utf8proc seems viable.
-// Non-Unicode locales can mostly be handled with simple iconv like in sdtui.
-// Similarly grapheme breaks can be guessed at using character width (a basic
-// test here is Zalgo text).
-//
-// None of this is ever going to work too reliably anyway because terminals
-// and Unicode don't go awfully well together.  In particular, character cell
-// devices have some problems with double-wide characters.
-
-#include <unistr.h>
-#include <uniwidth.h>
-#include <uniconv.h>
-#include <unicase.h>
-
-// --- Configurable display attributes -----------------------------------------
-
-struct attrs
-{
-	short fg;                           ///< Foreground colour index
-	short bg;                           ///< Background colour index
-	chtype attrs;                       ///< Other attributes
-};
-
-/// Decode attributes in the value using a subset of the git config format,
-/// ignoring all errors since it doesn't affect functionality
-static struct attrs
-attrs_decode (const char *value)
-{
-	struct strv v = strv_make ();
-	cstr_split (value, " ", true, &v);
-
-	int colors = 0;
-	struct attrs attrs = { -1, -1, 0 };
-	for (char **it = v.vector; *it; it++)
-	{
-		char *end = NULL;
-		long n = strtol (*it, &end, 10);
-		if (*it != end && !*end && n >= SHRT_MIN && n <= SHRT_MAX)
-		{
-			if (colors == 0) attrs.fg = n;
-			if (colors == 1) attrs.bg = n;
-			colors++;
-		}
-		else if (!strcmp (*it, "bold"))    attrs.attrs |= A_BOLD;
-		else if (!strcmp (*it, "dim"))     attrs.attrs |= A_DIM;
-		else if (!strcmp (*it, "ul"))      attrs.attrs |= A_UNDERLINE;
-		else if (!strcmp (*it, "blink"))   attrs.attrs |= A_BLINK;
-		else if (!strcmp (*it, "reverse")) attrs.attrs |= A_REVERSE;
-#ifdef A_ITALIC
-		else if (!strcmp (*it, "italic"))  attrs.attrs |= A_ITALIC;
-#endif  // A_ITALIC
-	}
-	strv_free (&v);
-	return attrs;
-}
-
-// --- Terminal output ---------------------------------------------------------
-
-// Necessary abstraction to simplify aligned, formatted character output
-
-// This callback you need to implement in the application
-static bool app_is_character_in_locale (ucs4_t ch);
-
-struct row_char
-{
-	ucs4_t c;                           ///< Unicode codepoint
-	chtype attrs;                       ///< Special attributes
-	int width;                          ///< How many cells this takes
-};
-
-struct row_buffer
-{
-	ARRAY (struct row_char, chars)      ///< Characters
-	int total_width;                    ///< Total width of all characters
-};
-
-static struct row_buffer
-row_buffer_make (void)
-{
-	struct row_buffer self = {};
-	ARRAY_INIT_SIZED (self.chars, 256);
-	return self;
-}
-
-static void
-row_buffer_free (struct row_buffer *self)
-{
-	free (self->chars);
-}
-
-/// Replace invalid chars and push all codepoints to the array w/ attributes.
-static void
-row_buffer_append (struct row_buffer *self, const char *str, chtype attrs)
-{
-	// The encoding is only really used internally for some corner cases
-	const char *encoding = locale_charset ();
-
-	// Note that this function is a hotspot, try to keep it decently fast
-	struct row_char current = { .attrs = attrs };
-	struct row_char invalid = { .attrs = attrs, .c = '?', .width = 1 };
-	const uint8_t *next = (const uint8_t *) str;
-	while ((next = u8_next (&current.c, next)))
-	{
-		current.width = uc_width (current.c, encoding);
-		if (current.width < 0 || !app_is_character_in_locale (current.c))
-			current = invalid;
-
-		ARRAY_RESERVE (self->chars, 1);
-		self->chars[self->chars_len++] = current;
-		self->total_width += current.width;
-	}
-}
-
-static void
-row_buffer_append_args (struct row_buffer *self, const char *s, ...)
-	ATTRIBUTE_SENTINEL;
-
-static void
-row_buffer_append_args (struct row_buffer *self, const char *s, ...)
-{
-	va_list ap;
-	va_start (ap, s);
-
-	while (s)
-	{
-		row_buffer_append (self, s, va_arg (ap, chtype));
-		s = va_arg (ap, const char *);
-	}
-	va_end (ap);
-}
-
-static void
-row_buffer_append_buffer (struct row_buffer *self, const struct row_buffer *rb)
-{
-	ARRAY_RESERVE (self->chars, rb->chars_len);
-	memcpy (self->chars + self->chars_len, rb->chars,
-		rb->chars_len * sizeof *rb->chars);
-
-	self->chars_len   += rb->chars_len;
-	self->total_width += rb->total_width;
-}
-
-/// Pop as many codepoints as needed to free up "space" character cells.
-/// Given the suffix nature of combining marks, this should work pretty fine.
-static int
-row_buffer_pop_cells (struct row_buffer *self, int space)
-{
-	int made = 0;
-	while (self->chars_len && made < space)
-		made += self->chars[--self->chars_len].width;
-	self->total_width -= made;
-	return made;
-}
-
-static void
-row_buffer_space (struct row_buffer *self, int width, chtype attrs)
-{
-	if (width < 0)
-		return;
-
-	ARRAY_RESERVE (self->chars, (size_t) width);
-
-	struct row_char space = { .attrs = attrs, .c = ' ', .width = 1 };
-	self->total_width += width;
-	while (width-- > 0)
-		self->chars[self->chars_len++] = space;
-}
-
-static void
-row_buffer_ellipsis (struct row_buffer *self, int target)
-{
-	if (self->total_width <= target
-	 || !row_buffer_pop_cells (self, self->total_width - target))
-		return;
-
-	// We use attributes from the last character we've removed,
-	// assuming that we don't shrink the array (and there's no real need)
-	ucs4_t ellipsis = 0x2026; // …
-	if (app_is_character_in_locale (ellipsis))
-	{
-		if (self->total_width >= target)
-			row_buffer_pop_cells (self, 1);
-		if (self->total_width + 1 <= target)
-			row_buffer_append (self, "…",   self->chars[self->chars_len].attrs);
-	}
-	else if (target >= 3)
-	{
-		if (self->total_width >= target)
-			row_buffer_pop_cells (self, 3);
-		if (self->total_width + 3 <= target)
-			row_buffer_append (self, "...", self->chars[self->chars_len].attrs);
-	}
-}
-
-static void
-row_buffer_align (struct row_buffer *self, int target, chtype attrs)
-{
-	row_buffer_ellipsis (self, target);
-	row_buffer_space (self, target - self->total_width, attrs);
-}
-
-static void
-row_buffer_print (uint32_t *ucs4, chtype attrs)
-{
-	// This assumes that we can reset the attribute set without consequences
-	char *str = u32_strconv_to_locale (ucs4);
-	if (str)
-	{
-		attrset (attrs);
-		addstr (str);
-		attrset (0);
-		free (str);
-	}
-}
-
-static void
-row_buffer_flush (struct row_buffer *self)
-{
-	if (!self->chars_len)
-		return;
-
-	// We only NUL-terminate the chunks because of the libunistring API
-	uint32_t chunk[self->chars_len + 1], *insertion_point = chunk;
-	for (size_t i = 0; i < self->chars_len; i++)
-	{
-		struct row_char *iter = self->chars + i;
-		if (i && iter[0].attrs != iter[-1].attrs)
-		{
-			row_buffer_print (chunk, iter[-1].attrs);
-			insertion_point = chunk;
-		}
-		*insertion_point++ = iter->c;
-		*insertion_point = 0;
-	}
-	row_buffer_print (chunk, self->chars[self->chars_len - 1].attrs);
-}

tests/fuzz.c

@@ -32,13 +32,6 @@
 #define LIBERTY_WANT_PROTO_MPD
 
 #include "../liberty.c"
-#include "../liberty-tui.c"
-
-static bool
-app_is_character_in_locale (ucs4_t ch)
-{
-	return ch < 128;
-}
 
 // --- UTF-8 -------------------------------------------------------------------
 
@@ -211,19 +204,6 @@ test_config_item_parse (const uint8_t *data, size_t size)
 		config_item_destroy (item);
 }
 
-// --- TUI ---------------------------------------------------------------------
-
-static void
-test_attrs_decode (const uint8_t *data, size_t size)
-{
-	struct str wrap = str_make ();
-	str_append_data (&wrap, data, size);
-
-	attrs_decode (wrap.str);
-
-	str_free (&wrap);
-}
-
 // --- MPD ---------------------------------------------------------------------
 
 static void
@@ -266,7 +246,6 @@ LLVMFuzzerInitialize (int *argcp, char ***argvp)
 	REGISTER (fcgi_parser_push)
 	REGISTER (fcgi_nv_parser_push)
 	REGISTER (config_item_parse)
-	REGISTER (attrs_decode)
 	REGISTER (mpd_client_process_input)
 
 	char **argv = *argvp, *option = "-test=", *name = NULL;

tools/asciiman.awk

@@ -1,6 +1,6 @@
 # asciiman.awk: simplified AsciiDoc to manual page converter
 #
-# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
+# Copyright (c) 2022 - 2023, Přemysl Eric Janouch <p@janouch.name>
 # SPDX-License-Identifier: 0BSD
 #
 # This is not intended to produce great output, merely useful output.
@@ -20,21 +20,17 @@ function fatal(message) {
 	exit 1
 }
 
-function haveattribute(name) {
-	return name in Attrs || ("asciidoc-" name) in ENVIRON
+BEGIN {
+	for (name in ENVIRON)
+		if (match(name, /^asciidoc-/))
+			Attrs[substr(name, RSTART + RLENGTH)] = ENVIRON[name]
 }
 
-function getattribute(name) {
-	if (!(name in Attrs) && ("asciidoc-" name) in ENVIRON)
-		Attrs[name] = ENVIRON["asciidoc-" name]
-	return Attrs[name]
-}
-
-function expand(s,   attr, v) {
-	while (match(s, /[{][^{}]*[}]/)) {
-		attr = substr(s, RSTART + 1, RLENGTH - 2)
-		if (haveattribute(attr))
-			v = v substr(s, 1, RSTART - 1) getattribute(attr)
+function expand(s,   attrname, v) {
+	while (match(s, /[{][^{}]+[}]/)) {
+		attrname = substr(s, RSTART + 1, RLENGTH - 2)
+		if (attrname in Attrs)
+			v = v substr(s, 1, RSTART - 1) Attrs[attrname]
 		else
 			v = v substr(s, 1, RSTART + RLENGTH - 1)
 		s = substr(s, RSTART + RLENGTH)
@@ -49,13 +45,20 @@ function escape(s) {
 	return s
 }
 
-function readattribute(line,    attrname, attrvalue) {
-	if (match(line, /^:[^:]*: /)) {
+function readattribute(line,    attrname) {
+	if (match(line, /^:[^:]+:$/)) {
+		Attrs[substr(line, RSTART + 1, RLENGTH - 2)] = ""
+	} else if (match(line, /^:[^:]+!:$/)) {
+		delete Attrs[substr(line, RSTART + 1, RLENGTH - 3)]
+	} else if (match(line, /^:![^:]+:$/)) {
+		delete Attrs[substr(line, RSTART + 2, RLENGTH - 3)]
+	} else if (match(line, /^:[^:]+: /)) {
 		attrname = substr(line, RSTART + 1, RLENGTH - 3)
-		attrvalue = substr(line, RSTART + RLENGTH)
-		Attrs[attrname] = expand(attrvalue)
-		return 1
+		Attrs[attrname] = expand(substr(line, RSTART + RLENGTH))
+	} else {
+		return 0
 	}
+	return 1
 }
 
 NR == 1 {
@@ -80,15 +83,56 @@ NR == 1 {
 	# Requesting tbl(1), even though we currently do not support tables.
 	print "'\\\" t"
 	printf ".TH \"%s\" \"%s\" \"\" \"%s\"",
-		toupper(name), section, getattribute("mansource")
-	if (getattribute("manmanual"))
-		printf " \"%s\"", getattribute("manmanual")
+		toupper(name), section, Attrs["mansource"]
+	if ("manmanual" in Attrs)
+		printf " \"%s\"", Attrs["manmanual"]
 	print ""
 
 	# Hyphenation is indeed rather annoying, in particular with long links.
 	print ".nh"
 }
 
+function readattrlist(line, posattrs, namedattrs,    name, value, n) {
+	if (!match(line, /^\[.*\]$/))
+		return 0
+
+	line = expand(substr(line, RSTART + 1, RLENGTH - 2))
+	while (line) {
+		name = ""
+		if (match(line, /^[[:alnum:]][[:alnum:]-]*/)) {
+			value = substr(line, RSTART, RLENGTH)
+			if (match(substr(line, RSTART + RLENGTH),
+					/^[[:space:]]*=[[:space:]]*/)) {
+				name = value
+				line = substr(line, 1 + length(name) + RLENGTH)
+			}
+		}
+
+		# The quoting syntax actually is awful like this.
+		if (match(line, /^"(\\.|[^"\\])*"/)) {
+			value = substr(line, RSTART + 1, RLENGTH - 2)
+			gsub(/\\"/, "\"", value)
+		} else if (match(line, /^'(\\.|[^'\\])*'/)) {
+			value = substr(line, RSTART + 1, RLENGTH - 2)
+			gsub(/\\'/, "'", value)
+		} else {
+			match(line, /^[^,]*/)
+			value = substr(line, RSTART, RLENGTH)
+			sub(/[[:space:]]*$/, "", value)
+		}
+
+		line = substr(line, RSTART + RLENGTH)
+		sub(/^[[:space:]]*,[[:space:]]*/, "", line)
+		if (!name)
+			posattrs[++n] = value
+		else if (value == "None")
+			delete namedattrs[name]
+		else
+			namedattrs[name] = value
+	}
+	return 1
+}
+
 function format(line,    v) {
 	# Pass-through, otherwise useful for hacks, is a bit of a lie here,
 	# and formatting doesn't fully respect word boundaries.
@@ -151,7 +195,7 @@ function inline(line) {
 }
 
 # Returns 1 iff the left-over $0 should be processed further.
-function process(firstline) {
+function process(firstline,     posattrs, namedattrs) {
 	if (readattribute(firstline))
 		return 0
 	if (getline <= 0) {
@@ -159,6 +203,17 @@ function process(firstline) {
 		return 0
 	}
 
+	# Block attribute list lines.
+	delete posattrs[0]
+	delete namedattrs[0]
+	while (readattrlist(firstline, posattrs, namedattrs)) {
+		firstline = $0
+		if (getline <= 0) {
+			inline(firstline)
+			return 0
+		}
+	}
+
 	# mandoc(1) automatically precedes section headers with blank lines.
 	if (length(firstline) == length($0) && /^-+$/) {
 		print ".SH \"" escape(toupper(expand(firstline))) "\""
@@ -196,7 +251,7 @@ function process(firstline) {
 		return 1
 	}
 
-	# We generally assume these block end with a blank line.
+	# We generally assume these blocks end with a blank line.
 	if (match(firstline, /^[[:space:]]*[*][[:space:]]+/)) {
 		flushspace()