cat-file: split --batch input lines on whitespace
[gitweb.git] / utf8.c
diff --git a/utf8.c b/utf8.c
index ddfdc5e2b88d346886f64e39a93ced85cc10a78e..0d20e0acb2b6fb2dd1d63abed676a036b2b7ec2f 100644 (file)
--- a/utf8.c
+++ b/utf8.c
@@ -1,4 +1,5 @@
 #include "git-compat-util.h"
+#include "strbuf.h"
 #include "utf8.h"
 
 /* This code is originally from http://www.cl.cam.ac.uk/~mgk25/ucs/ */
@@ -8,6 +9,20 @@ struct interval {
   int last;
 };
 
+size_t display_mode_esc_sequence_len(const char *s)
+{
+       const char *p = s;
+       if (*p++ != '\033')
+               return 0;
+       if (*p++ != '[')
+               return 0;
+       while (isdigit(*p) || *p == ';')
+               p++;
+       if (*p++ != 'm')
+               return 0;
+       return p - s;
+}
+
 /* auxiliary function for binary search in interval table */
 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
 {
@@ -162,7 +177,7 @@ static int git_wcwidth(ucs_char_t ch)
  * If the string was not a valid UTF-8, *start pointer is set to NULL
  * and the return value is undefined.
  */
-ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
+static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
 {
        unsigned char *s = (unsigned char *)*start;
        ucs_char_t ch;
@@ -251,18 +266,26 @@ int utf8_width(const char **start, size_t *remainder_p)
  * string, assuming that the string is utf8.  Returns strlen() instead
  * if the string does not look like a valid utf8 string.
  */
-int utf8_strwidth(const char *string)
+int utf8_strnwidth(const char *string, int len, int skip_ansi)
 {
        int width = 0;
        const char *orig = string;
 
-       while (1) {
-               if (!string)
-                       return strlen(orig);
-               if (!*string)
-                       return width;
+       if (len == -1)
+               len = strlen(string);
+       while (string && string < orig + len) {
+               int skip;
+               while (skip_ansi &&
+                      (skip = display_mode_esc_sequence_len(string)) != 0)
+                       string += skip;
                width += utf8_width(&string, NULL);
        }
+       return string ? width : len;
+}
+
+int utf8_strwidth(const char *string)
+{
+       return utf8_strnwidth(string, -1, 0);
 }
 
 int is_utf8(const char *text)
@@ -279,14 +302,27 @@ int is_utf8(const char *text)
        return 1;
 }
 
-static void print_spaces(int count)
+static void strbuf_addchars(struct strbuf *sb, int c, size_t n)
 {
-       static const char s[] = "                    ";
-       while (count >= sizeof(s)) {
-               fwrite(s, sizeof(s) - 1, 1, stdout);
-               count -= sizeof(s) - 1;
+       strbuf_grow(sb, n);
+       memset(sb->buf + sb->len, c, n);
+       strbuf_setlen(sb, sb->len + n);
+}
+
+static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
+                                    int indent, int indent2)
+{
+       if (indent < 0)
+               indent = 0;
+       while (*text) {
+               const char *eol = strchrnul(text, '\n');
+               if (*eol == '\n')
+                       eol++;
+               strbuf_addchars(buf, ' ', indent);
+               strbuf_add(buf, text, eol - text);
+               text = eol;
+               indent = indent2;
        }
-       fwrite(s, count, 1, stdout);
 }
 
 /*
@@ -295,51 +331,142 @@ static void print_spaces(int count)
  * If indent is negative, assume that already -indent columns have been
  * consumed (and no extra indent is necessary for the first line).
  */
-int print_wrapped_text(const char *text, int indent, int indent2, int width)
+void strbuf_add_wrapped_text(struct strbuf *buf,
+               const char *text, int indent1, int indent2, int width)
 {
-       int w = indent, assume_utf8 = is_utf8(text);
-       const char *bol = text, *space = NULL;
+       int indent, w, assume_utf8 = 1;
+       const char *bol, *space, *start = text;
+       size_t orig_len = buf->len;
 
+       if (width <= 0) {
+               strbuf_add_indented_text(buf, text, indent1, indent2);
+               return;
+       }
+
+retry:
+       bol = text;
+       w = indent = indent1;
+       space = NULL;
        if (indent < 0) {
                w = -indent;
                space = text;
        }
 
        for (;;) {
-               char c = *text;
+               char c;
+               size_t skip;
+
+               while ((skip = display_mode_esc_sequence_len(text)))
+                       text += skip;
+
+               c = *text;
                if (!c || isspace(c)) {
-                       if (w < width || !space) {
+                       if (w <= width || !space) {
                                const char *start = bol;
+                               if (!c && text == start)
+                                       return;
                                if (space)
                                        start = space;
                                else
-                                       print_spaces(indent);
-                               fwrite(start, text - start, 1, stdout);
+                                       strbuf_addchars(buf, ' ', indent);
+                               strbuf_add(buf, start, text - start);
                                if (!c)
-                                       return w;
-                               else if (c == '\t')
-                                       w |= 0x07;
+                                       return;
                                space = text;
+                               if (c == '\t')
+                                       w |= 0x07;
+                               else if (c == '\n') {
+                                       space++;
+                                       if (*space == '\n') {
+                                               strbuf_addch(buf, '\n');
+                                               goto new_line;
+                                       }
+                                       else if (!isalnum(*space))
+                                               goto new_line;
+                                       else
+                                               strbuf_addch(buf, ' ');
+                               }
                                w++;
                                text++;
                        }
                        else {
-                               putchar('\n');
+new_line:
+                               strbuf_addch(buf, '\n');
                                text = bol = space + isspace(*space);
                                space = NULL;
                                w = indent = indent2;
                        }
                        continue;
                }
-               if (assume_utf8)
+               if (assume_utf8) {
                        w += utf8_width(&text, NULL);
-               else {
+                       if (!text) {
+                               assume_utf8 = 0;
+                               text = start;
+                               strbuf_setlen(buf, orig_len);
+                               goto retry;
+                       }
+               } else {
                        w++;
                        text++;
                }
        }
 }
 
+void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
+                            int indent, int indent2, int width)
+{
+       char *tmp = xstrndup(data, len);
+       strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
+       free(tmp);
+}
+
+void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
+                        const char *subst)
+{
+       struct strbuf sb_dst = STRBUF_INIT;
+       char *src = sb_src->buf;
+       char *end = src + sb_src->len;
+       char *dst;
+       int w = 0, subst_len = 0;
+
+       if (subst)
+               subst_len = strlen(subst);
+       strbuf_grow(&sb_dst, sb_src->len + subst_len);
+       dst = sb_dst.buf;
+
+       while (src < end) {
+               char *old;
+               size_t n;
+
+               while ((n = display_mode_esc_sequence_len(src))) {
+                       memcpy(dst, src, n);
+                       src += n;
+                       dst += n;
+               }
+
+               old = src;
+               n = utf8_width((const char**)&src, NULL);
+               if (!src)       /* broken utf-8, do nothing */
+                       return;
+               if (n && w >= pos && w < pos + width) {
+                       if (subst) {
+                               memcpy(dst, subst, subst_len);
+                               dst += subst_len;
+                               subst = NULL;
+                       }
+                       w += n;
+                       continue;
+               }
+               memcpy(dst, old, src - old);
+               dst += src - old;
+               w += n;
+       }
+       strbuf_setlen(&sb_dst, dst - sb_dst.buf);
+       strbuf_swap(sb_src, &sb_dst);
+       strbuf_release(&sb_dst);
+}
+
 int is_encoding_utf8(const char *name)
 {
        if (!name)
@@ -349,29 +476,50 @@ int is_encoding_utf8(const char *name)
        return 0;
 }
 
+int same_encoding(const char *src, const char *dst)
+{
+       if (is_encoding_utf8(src) && is_encoding_utf8(dst))
+               return 1;
+       return !strcasecmp(src, dst);
+}
+
+/*
+ * Wrapper for fprintf and returns the total number of columns required
+ * for the printed string, assuming that the string is utf8.
+ */
+int utf8_fprintf(FILE *stream, const char *format, ...)
+{
+       struct strbuf buf = STRBUF_INIT;
+       va_list arg;
+       int columns;
+
+       va_start(arg, format);
+       strbuf_vaddf(&buf, format, arg);
+       va_end(arg);
+
+       columns = fputs(buf.buf, stream);
+       if (0 <= columns) /* keep the error from the I/O */
+               columns = utf8_strwidth(buf.buf);
+       strbuf_release(&buf);
+       return columns;
+}
+
 /*
  * Given a buffer and its encoding, return it re-encoded
  * with iconv.  If the conversion fails, returns NULL.
  */
 #ifndef NO_ICONV
-#ifdef OLD_ICONV
+#if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
        typedef const char * iconv_ibp;
 #else
        typedef char * iconv_ibp;
 #endif
-char *reencode_string(const char *in, const char *out_encoding, const char *in_encoding)
+char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
 {
-       iconv_t conv;
-       size_t insz, outsz, outalloc;
+       size_t outsz, outalloc;
        char *out, *outpos;
        iconv_ibp cp;
 
-       if (!in_encoding)
-               return NULL;
-       conv = iconv_open(out_encoding, in_encoding);
-       if (conv == (iconv_t) -1)
-               return NULL;
-       insz = strlen(in);
        outsz = insz;
        outalloc = outsz + 1; /* for terminating NUL */
        out = xmalloc(outalloc);
@@ -385,7 +533,6 @@ char *reencode_string(const char *in, const char *out_encoding, const char *in_e
                        size_t sofar;
                        if (errno != E2BIG) {
                                free(out);
-                               iconv_close(conv);
                                return NULL;
                        }
                        /* insz has remaining number of bytes.
@@ -401,10 +548,83 @@ char *reencode_string(const char *in, const char *out_encoding, const char *in_e
                }
                else {
                        *outpos = '\0';
+                       if (outsz_p)
+                               *outsz_p = outpos - out;
                        break;
                }
        }
+       return out;
+}
+
+char *reencode_string_len(const char *in, int insz,
+                         const char *out_encoding, const char *in_encoding,
+                         int *outsz)
+{
+       iconv_t conv;
+       char *out;
+
+       if (!in_encoding)
+               return NULL;
+
+       conv = iconv_open(out_encoding, in_encoding);
+       if (conv == (iconv_t) -1) {
+               /*
+                * Some platforms do not have the variously spelled variants of
+                * UTF-8, so let's fall back to trying the most official
+                * spelling. We do so only as a fallback in case the platform
+                * does understand the user's spelling, but not our official
+                * one.
+                */
+               if (is_encoding_utf8(in_encoding))
+                       in_encoding = "UTF-8";
+               if (is_encoding_utf8(out_encoding))
+                       out_encoding = "UTF-8";
+               conv = iconv_open(out_encoding, in_encoding);
+               if (conv == (iconv_t) -1)
+                       return NULL;
+       }
+
+       out = reencode_string_iconv(in, insz, conv, outsz);
        iconv_close(conv);
        return out;
 }
 #endif
+
+/*
+ * Returns first character length in bytes for multi-byte `text` according to
+ * `encoding`.
+ *
+ * - The `text` pointer is updated to point at the next character.
+ * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
+ *   we can consume from text, and on exit `*remainder_p` is reduced by returned
+ *   character length. Otherwise `text` is treated as limited by NUL.
+ */
+int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
+{
+       int chrlen;
+       const char *p = *text;
+       size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
+
+       if (r < 1)
+               return 0;
+
+       if (is_encoding_utf8(encoding)) {
+               pick_one_utf8_char(&p, &r);
+
+               chrlen = p ? (p - *text)
+                          : 1 /* not valid UTF-8 -> raw byte sequence */;
+       }
+       else {
+               /*
+                * TODO use iconv to decode one char and obtain its chrlen
+                * for now, let's treat encodings != UTF-8 as one-byte
+                */
+               chrlen = 1;
+       }
+
+       *text += chrlen;
+       if (remainder_p)
+               *remainder_p -= chrlen;
+
+       return chrlen;
+}