lineedit: partially fix wide and combining chars editing

Signed-off-by: Tomas Heinrich <heinrich.tomas@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

3 files changed, 92 insertions, 45 deletions

diff --git a/include/unicode.h b/include/unicode.h
index 4e2927297..747026abe 100644
--- a/include/unicode.h
+++ b/include/unicode.h
@@ -35,6 +35,16 @@ enum {
 #  define LAST_SUPPORTED_WCHAR CONFIG_LAST_SUPPORTED_WCHAR
 # endif
 
+# if LAST_SUPPORTED_WCHAR < 0x300
+#  undef ENABLE_UNICODE_COMBINING_WCHARS
+#  define ENABLE_UNICODE_COMBINING_WCHARS 0
+# endif
+
+# if LAST_SUPPORTED_WCHAR < 0x1100
+#  undef ENABLE_UNICODE_WIDE_WCHARS
+#  define ENABLE_UNICODE_WIDE_WCHARS 0
+# endif
+
 # if LAST_SUPPORTED_WCHAR < 0x590
 #  undef  ENABLE_UNICODE_BIDI_SUPPORT
 #  define ENABLE_UNICODE_BIDI_SUPPORT 0
@@ -92,6 +102,7 @@ size_t wcrtomb(char *s, wchar_t wc, mbstate_t *ps) FAST_FUNC;
 int iswspace(wint_t wc) FAST_FUNC;
 int iswalnum(wint_t wc) FAST_FUNC;
 int iswpunct(wint_t wc) FAST_FUNC;
+int wcwidth(unsigned ucs) FAST_FUNC;
 #  if ENABLE_UNICODE_BIDI_SUPPORT
 #   undef unicode_bidi_isrtl
 int unicode_bidi_isrtl(wint_t wc) FAST_FUNC;
diff --git a/libbb/lineedit.c b/libbb/lineedit.c
index 7fffe7ba2..9f2d65717 100644
--- a/libbb/lineedit.c
+++ b/libbb/lineedit.c
@@ -42,14 +42,10 @@
 #include "libbb.h"
 #include "unicode.h"
 
-/* FIXME: obsolete CONFIG item? */
-#define ENABLE_FEATURE_NONPRINTABLE_INVERSE_PUT 0
-
 #ifdef TEST
 # define ENABLE_FEATURE_EDITING 0
 # define ENABLE_FEATURE_TAB_COMPLETION 0
 # define ENABLE_FEATURE_USERNAME_COMPLETION 0
-# define ENABLE_FEATURE_NONPRINTABLE_INVERSE_PUT 0
 #endif
 
 
@@ -97,10 +93,10 @@ static bool BB_ispunct(CHAR_T c) { return ((unsigned)c < 256 && ispunct(c)); }
 
 
 # if ENABLE_UNICODE_PRESERVE_BROKEN
-#  define unicode_mark_inv_wchar(wc)   ((wc) | 0x20000000)
-#  define unicode_is_inv_wchar(wc)     ((wc) & 0x20000000)
+#  define unicode_mark_raw_byte(wc)   ((wc) | 0x20000000)
+#  define unicode_is_raw_byte(wc)     ((wc) & 0x20000000)
 # else
-#  define unicode_is_inv_wchar(wc)     0
+#  define unicode_is_raw_byte(wc)     0
 # endif
 
 
@@ -240,7 +236,7 @@ static unsigned save_string(char *dst, unsigned maxsize)
                 wchar_t wc;
                 int n = srcpos;
                 while ((wc = command_ps[srcpos]) != 0
-                    && !unicode_is_inv_wchar(wc)
+                    && !unicode_is_raw_byte(wc)
                 ) {
                         srcpos++;
                 }
@@ -269,15 +265,45 @@ static void BB_PUTCHAR(wchar_t c)
         mbstate_t mbst = { 0 };
         ssize_t len;
 
-        if (unicode_is_inv_wchar(c))
-                c = CONFIG_SUBST_WCHAR;
         len = wcrtomb(buf, c, &mbst);
         if (len > 0) {
                 buf[len] = '\0';
                 fputs(buf, stdout);
         }
 }
-#else
+# if ENABLE_UNICODE_COMBINING_WCHARS || ENABLE_UNICODE_WIDE_WCHARS
+static wchar_t adjust_width_and_validate_wc(unsigned *width_adj, wchar_t wc)
+# else
+static wchar_t adjust_width_and_validate_wc(wchar_t wc)
+#  define adjust_width_and_validate_wc(width_adj, wc) \
+        ((*(width_adj))++, adjust_width_and_validate_wc(wc))
+# endif
+{
+        int w = 1;
+
+        if (unicode_status == UNICODE_ON) {
+                if (unicode_is_raw_byte(wc)
+                 || (CONFIG_LAST_SUPPORTED_WCHAR && wc > CONFIG_LAST_SUPPORTED_WCHAR)
+                ) {
+                        goto subst;
+                }
+                w = wcwidth(wc);
+                if ((ENABLE_UNICODE_COMBINING_WCHARS && w < 0)
+                 || (!ENABLE_UNICODE_COMBINING_WCHARS && w <= 0)
+                 || (!ENABLE_UNICODE_WIDE_WCHARS && w > 1)
+                ) {
+ subst:
+                        w = 1;
+                        wc = CONFIG_SUBST_WCHAR;
+                }
+        }
+
+# if ENABLE_UNICODE_COMBINING_WCHARS || ENABLE_UNICODE_WIDE_WCHARS
+        *width_adj += w;
+#endif
+        return wc;
+}
+#else /* !UNICODE */
 static size_t load_string(const char *src, int maxsize)
 {
         safe_strncpy(command_ps, src, maxsize);
@@ -290,6 +316,8 @@ static void save_string(char *dst, unsigned maxsize)
 }
 # endif
 # define BB_PUTCHAR(c) bb_putchar(c)
+/* Should never be called: */
+int adjust_width_and_validate_wc(unsigned *width_adj, int wc);
 #endif
 
 
@@ -300,6 +328,8 @@ static void save_string(char *dst, unsigned maxsize)
 static void put_cur_glyph_and_inc_cursor(void)
 {
         CHAR_T c = command_ps[cursor];
+        unsigned width = 0;
+        int ofs_to_right;
 
         if (c == BB_NUL) {
                 /* erase character after end of input string */
@@ -307,28 +337,23 @@ static void put_cur_glyph_and_inc_cursor(void)
         } else {
                 /* advance cursor only if we aren't at the end yet */
                 cursor++;
-                cmdedit_x++;
+                if (unicode_status == UNICODE_ON) {
+                        IF_UNICODE_WIDE_WCHARS(width = cmdedit_x;)
+                        c = adjust_width_and_validate_wc(&cmdedit_x, c);
+                        IF_UNICODE_WIDE_WCHARS(width = cmdedit_x - width;)
+                } else {
+                        cmdedit_x++;
+                }
         }
 
-#if ENABLE_FEATURE_NONPRINTABLE_INVERSE_PUT
-        /* Display non-printable characters in reverse */
-        if (!BB_isprint(c)) {
-                if (c >= 128)
-                        c -= 128;
-                if (c < ' ')
-                        c += '@';
-                if (c == 127)
-                        c = '?';
-                printf("\033[7m%c\033[0m", c);
-        } else
-#endif
-        {
+        ofs_to_right = cmdedit_x - cmdedit_termw;
+        if (!ENABLE_UNICODE_WIDE_WCHARS || ofs_to_right <= 0) {
+                /* c fits on this line */
                 BB_PUTCHAR(c);
         }
-        if (cmdedit_x >= cmdedit_termw) {
-                /* terminal is scrolled down */
-                cmdedit_y++;
-                cmdedit_x = 0;
+
+        if (ofs_to_right >= 0) {
+                /* we go to the next line */
 #if HACK_FOR_WRONG_WIDTH
                 /* This works better if our idea of term width is wrong
                  * and it is actually wider (often happens on serial lines).
@@ -351,6 +376,14 @@ static void put_cur_glyph_and_inc_cursor(void)
                 BB_PUTCHAR(c);
                 bb_putchar('\b');
 #endif
+                cmdedit_y++;
+                if (!ENABLE_UNICODE_WIDE_WCHARS || ofs_to_right == 0) {
+                        width = 0;
+                } else { /* ofs_to_right > 0 */
+                        /* wide char c didn't fit on prev line */
+                        BB_PUTCHAR(c);
+                }
+                cmdedit_x = width;
         }
 }
 
@@ -389,10 +422,22 @@ static void input_backward(unsigned num)
 
         if (num > cursor)
                 num = cursor;
-        if (!num)
+        if (num == 0)
                 return;
         cursor -= num;
 
+        if ((ENABLE_UNICODE_COMBINING_WCHARS || ENABLE_UNICODE_WIDE_WCHARS)
+         && unicode_status == UNICODE_ON
+        ) {
+                /* correct NUM to be equal to _screen_ width */
+                int n = num;
+                num = 0;
+                while (--n >= 0)
+                        adjust_width_and_validate_wc(&num, command_ps[cursor + n]);
+                if (num == 0)
+                        return;
+        }
+
         if (cmdedit_x >= num) {
                 cmdedit_x -= num;
                 if (num <= 4) {
@@ -412,6 +457,8 @@ static void input_backward(unsigned num)
         }
 
         /* Need to go one or more lines up */
+//FIXME: this does not work correctly if prev line has one "unfilled" screen position
+//caused by wide unicode char not fitting in that one screen position.
         num -= cmdedit_x;
         {
                 unsigned w = cmdedit_termw; /* volatile var */
@@ -765,21 +812,13 @@ static NOINLINE int find_match(char *matchBuf, int *len_with_quotes)
         }
 
         /* mask \+symbol and convert '\t' to ' ' */
-        for (i = j = 0; matchBuf[i]; i++, j++)
+        for (i = j = 0; matchBuf[i]; i++, j++) {
                 if (matchBuf[i] == '\\') {
                         collapse_pos(j, j + 1);
                         int_buf[j] |= QUOT;
                         i++;
-#if ENABLE_FEATURE_NONPRINTABLE_INVERSE_PUT
-                        if (matchBuf[i] == '\t')  /* algorithm equivalent */
-                                int_buf[j] = ' ' | QUOT;
-#endif
                 }
-#if ENABLE_FEATURE_NONPRINTABLE_INVERSE_PUT
-                else if (matchBuf[i] == '\t')
-                        int_buf[j] = ' ';
-#endif
-
+        }
         /* mask "symbols" or 'symbols' */
         c2 = 0;
         for (i = 0; int_buf[i]; i++) {
@@ -1774,7 +1813,7 @@ static int lineedit_read_key(char *read_key_buffer)
 # if !ENABLE_UNICODE_PRESERVE_BROKEN
                                 ic = CONFIG_SUBST_WCHAR;
 # else
-                                ic = unicode_mark_inv_wchar(unicode_buf[0]);
+                                ic = unicode_mark_raw_byte(unicode_buf[0]);
 # endif
                         } else {
                                 /* Valid unicode char, return its code */
@@ -2384,9 +2423,6 @@ int main(int argc, char **argv)
                 "% ";
 #endif
 
-#if ENABLE_FEATURE_NONPRINTABLE_INVERSE_PUT
-        setlocale(LC_ALL, "");
-#endif
         while (1) {
                 int l;
                 l = read_line_input(prompt, buff);
diff --git a/libbb/unicode.c b/libbb/unicode.c
index d1c6167c7..eb0ea6129 100644
--- a/libbb/unicode.c
+++ b/libbb/unicode.c
@@ -418,7 +418,7 @@ static int in_uint16_table(unsigned ucs, const uint16_t *table, unsigned max)
  * This implementation assumes that wchar_t characters are encoded
  * in ISO 10646.
  */
-static int wcwidth(unsigned ucs)
+int FAST_FUNC wcwidth(unsigned ucs)
 {
 # if LAST_SUPPORTED_WCHAR >= 0x300
         /* sorted list of non-overlapping intervals of non-spacing characters */