summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--src/basic/string-util.c163
-rw-r--r--src/test/test-ellipsize.c41
2 files changed, 172 insertions, 32 deletions
diff --git a/src/basic/string-util.c b/src/basic/string-util.c
index 1eedcb66f7..93049e9820 100644
--- a/src/basic/string-util.c
+++ b/src/basic/string-util.c
@@ -295,6 +295,62 @@ static int write_ellipsis(char *buf, bool unicode) {
return 3;
}
+static size_t ansi_sequence_length(const char *s, size_t len) {
+ assert(s);
+
+ if (len < 2)
+ return 0;
+
+ if (s[0] != 0x1B) /* ASCII 27, aka ESC, aka Ctrl-[ */
+ return 0; /* Not the start of a sequence */
+
+ if (s[1] == 0x5B) { /* [, start of CSI sequence */
+ size_t i = 2;
+
+ if (i == len)
+ return 0;
+
+ while (s[i] >= 0x30 && s[i] <= 0x3F) /* Parameter bytes */
+ if (++i == len)
+ return 0;
+ while (s[i] >= 0x20 && s[i] <= 0x2F) /* Intermediate bytes */
+ if (++i == len)
+ return 0;
+ if (s[i] >= 0x40 && s[i] <= 0x7E) /* Final byte */
+ return i + 1;
+ return 0; /* Bad sequence */
+
+ } else if (s[1] >= 0x40 && s[1] <= 0x5F) /* other non-CSI Fe sequence */
+ return 2;
+
+ return 0; /* Bad escape? */
+}
+
+static bool string_has_ansi_sequence(const char *s, size_t len) {
+ const char *t = s;
+
+ while ((t = memchr(s, 0x1B, len - (t - s))))
+ if (ansi_sequence_length(t, len - (t - s)) > 0)
+ return true;
+ return false;
+}
+
+static size_t previous_ansi_sequence(const char *s, size_t length, const char **ret_where) {
+ /* Locate the previous ANSI sequence and save its start in *ret_where and return length. */
+
+ for (size_t i = length - 2; i > 0; i--) { /* -2 because at least two bytes are needed */
+ size_t slen = ansi_sequence_length(s + (i - 1), length - (i - 1));
+ if (slen == 0)
+ continue;
+
+ *ret_where = s + (i - 1);
+ return slen;
+ }
+
+ *ret_where = NULL;
+ return 0;
+}
+
static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
size_t x, need_space, suffix_len;
char *t;
@@ -354,7 +410,6 @@ static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_le
char *ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
size_t x, k, len, len2;
const char *i, *j;
- char *e;
int r;
/* Note that 'old_length' refers to bytes in the string, while 'new_length' refers to character cells taken up
@@ -378,73 +433,117 @@ char *ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigne
if (new_length == 0)
return strdup("");
- /* If no multibyte characters use ascii_ellipsize_mem for speed */
- if (ascii_is_valid_n(s, old_length))
+ bool has_ansi_seq = string_has_ansi_sequence(s, old_length);
+
+ /* If no multibyte characters or ANSI sequences, use ascii_ellipsize_mem for speed */
+ if (!has_ansi_seq && ascii_is_valid_n(s, old_length))
return ascii_ellipsize_mem(s, old_length, new_length, percent);
- x = ((new_length - 1) * percent) / 100;
+ x = (new_length - 1) * percent / 100;
assert(x <= new_length - 1);
k = 0;
- for (i = s; i < s + old_length; i = utf8_next_char(i)) {
- char32_t c;
- int w;
+ for (i = s; i < s + old_length; ) {
+ size_t slen = has_ansi_seq ? ansi_sequence_length(i, old_length - (i - s)) : 0;
+ if (slen > 0) {
+ i += slen;
+ continue; /* ANSI sequences don't take up any space in output */
+ }
+ char32_t c;
r = utf8_encoded_to_unichar(i, &c);
if (r < 0)
return NULL;
- w = unichar_iswide(c) ? 2 : 1;
- if (k + w <= x)
- k += w;
- else
+ int w = unichar_iswide(c) ? 2 : 1;
+ if (k + w > x)
break;
+
+ k += w;
+ i += r;
}
- for (j = s + old_length; j > i; ) {
+ const char *ansi_start = s + old_length;
+ size_t ansi_len = 0;
+
+ for (const char *t = j = s + old_length; t > i && k < new_length; ) {
char32_t c;
int w;
- const char *jj;
+ const char *tt;
+
+ if (has_ansi_seq && ansi_start >= t)
+ /* Figure out the previous ANSI sequence, if any */
+ ansi_len = previous_ansi_sequence(s, t - s, &ansi_start);
- jj = utf8_prev_char(j);
- r = utf8_encoded_to_unichar(jj, &c);
+ /* If the sequence extends all the way to the current position, skip it. */
+ if (has_ansi_seq && ansi_len > 0 && ansi_start + ansi_len == t) {
+ t = ansi_start;
+ continue;
+ }
+
+ tt = utf8_prev_char(t);
+ r = utf8_encoded_to_unichar(tt, &c);
if (r < 0)
return NULL;
w = unichar_iswide(c) ? 2 : 1;
- if (k + w <= new_length) {
- k += w;
- j = jj;
- } else
+ if (k + w > new_length)
break;
+
+ k += w;
+ j = t = tt; /* j should always point to the first "real" character */
}
- assert(i <= j);
- /* we don't actually need to ellipsize */
- if (i == j)
+ /* We don't actually need to ellipsize */
+ if (i >= j)
return memdup_suffix0(s, old_length);
- /* make space for ellipsis, if possible */
- if (j < s + old_length)
- j = utf8_next_char(j);
- else if (i > s)
- i = utf8_prev_char(i);
+ if (k >= new_length) {
+ /* Make space for ellipsis, if required and possible. We know that the edge character is not
+ * part of an ANSI sequence (because then we'd skip it). If the last character we looked at
+ * was wide, we don't need to make space. */
+ if (j < s + old_length)
+ j = utf8_next_char(j);
+ else if (i > s)
+ i = utf8_prev_char(i);
+ }
len = i - s;
len2 = s + old_length - j;
- e = new(char, len + 3 + len2 + 1);
+
+ /* If we have ANSI, allow the same length as the source string + ellipsis. It'd be too involved to
+ * figure out what exact space is needed. Strings with ANSI sequences are most likely to be fairly
+ * short anyway. */
+ size_t alloc_len = has_ansi_seq ? old_length + 3 + 1 : len + 3 + len2 + 1;
+
+ char *e = new(char, alloc_len);
if (!e)
return NULL;
/*
- printf("old_length=%zu new_length=%zu x=%zu len=%u len2=%u k=%u\n",
+ printf("old_length=%zu new_length=%zu x=%zu len=%zu len2=%zu k=%zu\n",
old_length, new_length, x, len, len2, k);
*/
- memcpy(e, s, len);
+ memcpy_safe(e, s, len);
write_ellipsis(e + len, true);
- memcpy(e + len + 3, j, len2);
- *(e + len + 3 + len2) = '\0';
+
+ char *dst = e + len + 3;
+
+ if (has_ansi_seq)
+ /* Copy over any ANSI sequences in full */
+ for (const char *p = s + len; p < j; ) {
+ size_t slen = ansi_sequence_length(p, j - p);
+ if (slen > 0) {
+ memcpy(dst, p, slen);
+ dst += slen;
+ p += slen;
+ } else
+ p = utf8_next_char(p);
+ }
+
+ memcpy_safe(dst, j, len2);
+ dst[len2] = '\0';
return e;
}
diff --git a/src/test/test-ellipsize.c b/src/test/test-ellipsize.c
index a96644f617..40edc289e9 100644
--- a/src/test/test-ellipsize.c
+++ b/src/test/test-ellipsize.c
@@ -4,6 +4,7 @@
#include "alloc-util.h"
#include "constants.h"
+#include "escape.h"
#include "string-util.h"
#include "strv.h"
#include "terminal-util.h"
@@ -115,4 +116,44 @@ TEST(ellipsize) {
test_ellipsize_one("shórt");
}
+TEST(ellipsize_ansi) {
+ const char *s = ANSI_HIGHLIGHT_YELLOW_UNDERLINE "yęllow"
+ ANSI_HIGHLIGHT_GREY_UNDERLINE "grěy"
+ ANSI_HIGHLIGHT_BLUE_UNDERLINE "blue"
+ ANSI_NORMAL "nórmął";
+ size_t len = strlen(s);
+
+ for (unsigned percent = 0; percent <= 100; percent += 15)
+ for (ssize_t x = 21; x >= 0; x--) {
+ _cleanup_free_ char *t = ellipsize_mem(s, len, x, percent);
+ printf("%02zd: \"%s\"\n", x, t);
+ assert_se(utf8_is_valid(t));
+
+ if (DEBUG_LOGGING) {
+ _cleanup_free_ char *e = cescape(t);
+ printf(" : \"%s\"\n", e);
+ }
+ }
+}
+
+TEST(ellipsize_ansi_cats) {
+ _cleanup_free_ char *e, *f, *g, *h;
+
+ /* Make sure we don't cut off in the middle of an ANSI escape sequence. */
+
+ e = ellipsize("01" ANSI_NORMAL "23", 4, 0);
+ puts(e);
+ assert_se(streq(e, "01" ANSI_NORMAL "23"));
+ f = ellipsize("ab" ANSI_NORMAL "cd", 4, 90);
+ puts(f);
+ assert_se(streq(f, "ab" ANSI_NORMAL "cd"));
+
+ g = ellipsize("🐱🐱" ANSI_NORMAL "🐱🐱" ANSI_NORMAL, 5, 0);
+ puts(g);
+ assert_se(streq(g, "…" ANSI_NORMAL "🐱🐱" ANSI_NORMAL));
+ h = ellipsize("🐱🐱" ANSI_NORMAL "🐱🐱" ANSI_NORMAL, 5, 90);
+ puts(h);
+ assert_se(streq(h, "🐱…" ANSI_NORMAL "🐱" ANSI_NORMAL));
+}
+
DEFINE_TEST_MAIN(LOG_INFO);
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-02 11:35:34 +0100