Add advance_next32() function to 'unicode' module to encapsulate the escape handling logic #1

src/tools.c

@@ -343,7 +343,7 @@ void btrim(char *text, size_t *len)
 
 void btrim32(uint32_t *text, size_t *len)
 /*
- *  Remove trailing whitespace from line (unicode version).
+ *  Remove trailing whitespace from line (unicode and escape sequence enabled version).
  *
  *      text     string to trim
  *      len      pointer to the length of the string in characters
@@ -353,18 +353,34 @@ void btrim32(uint32_t *text, size_t *len)
 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
  */
 {
-    int idx = (int) (*len - 1);
+    if (text == NULL || len == 0) {
+        return;
+    }
 
-    for (; idx >= 0; --idx) {
-        ucs4_t c = text[idx];
-        if (uc_is_c_whitespace(c) || uc_is_property_white_space(c) || uc_is_property_bidi_whitespace(c)) {
-            set_char_at(text, idx, char_nul);
-        } else {
-            break;
+    const uint32_t *rest = text;
+    int last_char_pos = -1;
+    size_t step_invis;
+
+    for (ucs4_t c = text[0]; c != char_nul; c = rest[0]) {
+        if (c != char_esc) {
+            if (!uc_is_c_whitespace(c) && !uc_is_property_white_space(c) && !uc_is_property_bidi_whitespace(c)) {
+                last_char_pos = (int) (rest - text);
+            }
+        }
+        rest = advance_next32(rest, &step_invis);
+    }
+
+    /* If the last character is followed by an escape sequence, keep it (but only one). */
+    if (last_char_pos >= 0) {
+        rest = text + last_char_pos + 1;
+        if (rest[0] == char_esc) {
+            advance_next32(rest, &step_invis);
+            last_char_pos += step_invis;
         }
     }
 
-    *len = idx + 1;
+    set_char_at(text, (size_t) (last_char_pos + 1), char_nul);
+    *len = (size_t) (last_char_pos + 1);
 }
 
 
@@ -546,7 +562,6 @@ void print_input_lines(const char *heading)
 static size_t count_invisible_chars(const uint32_t *s, size_t *num_esc, char **ascii, size_t **posmap)
 {
     size_t invis = 0;  /* counts invisible column positions */
-    int ansipos = 0;   /* progression of ansi sequence */
     *num_esc = 0;      /* counts the number of escape sequences found */
 
     if (is_empty(s)) {
@@ -562,42 +577,25 @@ static size_t count_invisible_chars(const uint32_t *s, size_t *num_esc, char **a
     (*ascii) = (char *) calloc(buflen, sizeof(char));     /* maybe a little too much, but certainly enough */
     char *p = *ascii;
 
-    ucs4_t c;
     size_t mb_idx = 0;
+    size_t step_invis;
     const uint32_t *rest = s;
-    while ((rest = u32_next(&c, rest))) {
+
+    for (ucs4_t c = s[0]; c != char_nul; c = rest[0]) {
         if (map_idx >= map_size - 4) {
             map_size = map_size * 2 + 1;
             map = (size_t *) realloc(map, map_size * sizeof(size_t));
         }
 
-        if (ansipos == 0 && c == char_esc) {
-            /* Found an ESC char, count it as invisible and move 1 forward in the detection of CSI sequences */
-            ansipos++;
-            invis++;
+        if (c == char_esc) {
             (*num_esc)++;
-        } else if (ansipos == 1 && c == '[') {
-            /* Found '[' char after ESC. A CSI sequence has started. */
-            ansipos++;
-            invis++;
-        } else if (ansipos == 1 && c >= 0x40 && c <= 0x5f) {
-            /* Found a byte designating the end of a two-byte escape sequence */
-            invis++;
-            ansipos = 0;
-        } else if (ansipos == 2) {
-            /* Inside CSI sequence - Keep counting bytes as invisible */
-            invis++;
-
-            /* A char between 0x40 and 0x7e signals the end of an CSI or escape sequence */
-            if (c >= 0x40 && c <= 0x7e) {
-                ansipos = 0;
-            }
-
-        } else if (is_ascii_printable(c)) {
+        }
+        else if (is_ascii_printable(c)) {
             *p = c & 0xff;
             map[map_idx++] = mb_idx;
             ++p;
-        } else {
+        }
+        else {
             int cols = uc_width(c, encoding);
             if (cols > 0) {
                 memset(p, (int) 'x', cols);
@@ -607,8 +605,13 @@ static size_t count_invisible_chars(const uint32_t *s, size_t *num_esc, char **a
                 p += cols;
             }
         }
-        ++mb_idx;
+
+        rest = advance_next32(rest, &step_invis);
+
+        mb_idx += BMAX((size_t) 1, step_invis);
+        invis += step_invis;
     }
+
     *p = '\0';
     (*posmap) = map;
     return invis;

src/unicode.c

@@ -100,6 +100,48 @@ uint32_t *new_empty_string32()
 
 
 
+uint32_t *advance_next32(const uint32_t *s, size_t *invis)
+{
+    if (is_empty(s)) {
+        return (uint32_t *) s;
+    }
+
+    int ansipos = 0;
+    (*invis) = 0;
+    ucs4_t c;
+    const uint32_t *rest = s;
+    while ((rest = u32_next(&c, rest))) {
+        if (ansipos == 0 && c == char_esc) {
+            /* Found an ESC char, count it as invisible and move 1 forward in the detection of CSI sequences */
+            (*invis)++;
+            ansipos++;
+        } else if (ansipos == 1 && c == '[') {
+            /* Found '[' char after ESC. A CSI sequence has started. */
+            (*invis)++;
+            ansipos++;
+        } else if (ansipos == 1 && c >= 0x40 && c <= 0x5f) {
+            /* Found a byte designating the end of a two-byte escape sequence */
+            (*invis)++;
+            ansipos = 0;
+            break;
+        } else if (ansipos == 2) {
+            /* Inside CSI sequence - Keep counting chars as invisible */
+            (*invis)++;
+
+            /* A char between 0x40 and 0x7e signals the end of an CSI or escape sequence */
+            if (c >= 0x40 && c <= 0x7e) {
+                ansipos = 0;
+                break;
+            }
+        } else {
+            break;
+        }
+    }
+    return (uint32_t *) rest;
+}
+
+
+
 uint32_t *advance32(uint32_t *s, const size_t offset)
 {
     if (is_empty(s)) {
@@ -109,50 +151,28 @@ uint32_t *advance32(uint32_t *s, const size_t offset)
         return s;
     }
 
-    ucs4_t c;                  /* the current character we're looking at */
-    const uint32_t *cStr = s;  /* pointer to c in s */
-    size_t idx = 0;            /* the count of visible characters */
-    const uint32_t *last_esc;  /* pointer to the start of the last escape sequence encountered */
-    const uint32_t *rest = s;  /* pointer to the next character coming up, needed only for u32_next() api */
-    int visible = 1;           /* flag indicating whether the previous char was a visible char */
-    int ansipos = 0;           /* progression of ansi sequence */
+    size_t count = 0;                 /* the count of visible characters */
+    int visible = 1;                  /* flag indicating whether the previous char was a visible char */
+    const uint32_t *last_esc = NULL;  /* pointer to the start of the last escape sequence encountered */
+    const uint32_t *rest = s;         /* pointer to the next character coming up */
+    size_t step_invis = 0;            /* unused, but required for advance_next32() call */
 
-    while ((rest = u32_next(&c, rest))) {
-        if (ansipos == 0 && c == char_esc) {
-            /* Found an ESC char, count it as invisible and move 1 forward in the detection of CSI sequences */
-            last_esc = cStr;
+    for (ucs4_t c = s[0]; c != char_nul; c = rest[0]) {
+        if (c == char_esc) {
+            last_esc = rest;
             visible = 0;
-            ansipos++;
-        } else if (ansipos == 1 && c == '[') {
-            /* Found '[' char after ESC. A CSI sequence has started. */
-            ansipos++;
-            visible = 0;
-        } else if (ansipos == 1 && c >= 0x40 && c <= 0x5f) {
-            /* Found a char designating the end of a two-byte escape sequence */
-            visible = 0;
-            ansipos = 0;
-        } else if (ansipos == 2) {
-            /* Inside CSI sequence - Keep counting chars as invisible */
-            visible = 0;
-
-            /* A char between 0x40 and 0x7e signals the end of an CSI or escape sequence */
-            if (c >= 0x40 && c <= 0x7e) {
-                ansipos = 0;
-            }
         } else {
-            /* a visible char */
-            if (idx == offset) {
-                if (!visible) {
+            if (count++ == offset) {
+                if (!visible && last_esc != NULL) {
                     return (uint32_t *) last_esc;
                 }
-                return (uint32_t *) cStr;
+                break;
             }
-            ++idx;
             visible = 1;
         }
-        cStr = rest;
+        rest = advance_next32(rest, &step_invis);
     }
-    return new_empty_string32();  /* offset too large, not enough characters in string */
+    return (uint32_t *) rest;         /* may point to zero terminator when offset too large */
 }
 
 

src/unicode.h

@@ -50,10 +50,28 @@ int is_ascii_printable(const ucs4_t c);
 /** Return a freshly allocated empty UTF-32 string. */
 uint32_t *new_empty_string32();
 
+/**
+ * Return the next position in <s> in accordance with escape sequences. The result can be the next normal character,
+ * or again an escape sequence, if it directly follows the first.
+ *
+ * @param <s> The pointer to the start position. Is assumed to point either at the ESC at the start of an escape
+ *      sequence, or to be positioned outside an escape sequence.
+ * @param <invis> Will contain the number of invisible characters skipped in order to get to the new position.
+ *      This will be 0 unless <s> pointed to an ESC char, in which case it contains the length in characters of that
+ *      escape sequence.
+ * @return The next position, or 0 if the end of the string was reached
+ */
+uint32_t *advance_next32(const uint32_t *s, size_t *invis);
+
 /**
  * Determine a new position in the given string s with the given offset of visible characters.
  * If the character right in front of the target character is invisible, then the pointer is moved to the start of
  * that invisible sequence. The purpose is to catch any escape sequences which would for example color the character.
+ *
+ * @param <s> The pointer to the start position. Is assumed to point either at the ESC at the start of an escape
+ *      sequence, or to be positioned outside an escape sequence.
+ * @param <offset> the number of visible character positions to advance the pointer
+ * @return a pointer to the new position in s, or 0 if the end of the string was reached
  */
 uint32_t *advance32(uint32_t *s, const size_t offset);
 

test/083_list_design_info.txt

@@ -10,7 +10,7 @@ Original Designer:      (public domain)
 Creation Date:          March 18, 1999 (Thursday, 15:25h)
 Current Revision:       1.0 as of March 18, 1999 (Thursday, 15:25h)
 Indentation Mode:       box (indent box)
-Replacement Rules:      1. (glob) "\*/" WITH "*\/"
+Replacement Rules:      1. (glob) "\*/" WITH "*\\/"
 Reversion Rules:        1. (glob) "\*\\/" TO "*/"
 Minimum Box Dimensions: 5 x 3  (width x height)
 Default Padding:        left 1, right 1

test/104_ansi_lolcat.txt

@@ -6,7 +6,7 @@
 :OUTPUT-FILTER
 :EXPECTED
 /********************************************/
-/* There is no complete theory of anything. */
+/* There is no complete theory of anything. */
 /*                      Robert Anton Wilson */
 /********************************************/
 :EOF

test/105_ansi_lolcat_within_leading_space.txt

@@ -5,7 +5,7 @@
 :OUTPUT-FILTER
 :EXPECTED
     /********************************************/
-    /* There is no complete theory of anything. */
+    /* There is no complete theory of anything. */
     /*   Robert Anton Wilson                    */
     /********************************************/
 :EOF