streaming_write_entry(): support files with holes
[gitweb.git] / pretty.c
index d493cade26890d3e16ea072ced8e0f95679f5670..dff5c8d1831ca4019d1a502a393cb95f20ad18f7 100644 (file)
--- a/pretty.c
+++ b/pretty.c
 #include "reflog-walk.h"
 
 static char *user_format;
+static struct cmt_fmt_map {
+       const char *name;
+       enum cmit_fmt format;
+       int is_tformat;
+       int is_alias;
+       const char *user_format;
+} *commit_formats;
+static size_t builtin_formats_len;
+static size_t commit_formats_len;
+static size_t commit_formats_alloc;
+static struct cmt_fmt_map *find_commit_format(const char *sought);
 
 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
 {
@@ -21,22 +32,118 @@ static void save_user_format(struct rev_info *rev, const char *cp, int is_tforma
        rev->commit_format = CMIT_FMT_USERFORMAT;
 }
 
-void get_commit_format(const char *arg, struct rev_info *rev)
+static int git_pretty_formats_config(const char *var, const char *value, void *cb)
 {
+       struct cmt_fmt_map *commit_format = NULL;
+       const char *name;
+       const char *fmt;
        int i;
-       static struct cmt_fmt_map {
-               const char *n;
-               size_t cmp_len;
-               enum cmit_fmt v;
-       } cmt_fmts[] = {
-               { "raw",        1,      CMIT_FMT_RAW },
-               { "medium",     1,      CMIT_FMT_MEDIUM },
-               { "short",      1,      CMIT_FMT_SHORT },
-               { "email",      1,      CMIT_FMT_EMAIL },
-               { "full",       5,      CMIT_FMT_FULL },
-               { "fuller",     5,      CMIT_FMT_FULLER },
-               { "oneline",    1,      CMIT_FMT_ONELINE },
+
+       if (prefixcmp(var, "pretty."))
+               return 0;
+
+       name = var + strlen("pretty.");
+       for (i = 0; i < builtin_formats_len; i++) {
+               if (!strcmp(commit_formats[i].name, name))
+                       return 0;
+       }
+
+       for (i = builtin_formats_len; i < commit_formats_len; i++) {
+               if (!strcmp(commit_formats[i].name, name)) {
+                       commit_format = &commit_formats[i];
+                       break;
+               }
+       }
+
+       if (!commit_format) {
+               ALLOC_GROW(commit_formats, commit_formats_len+1,
+                          commit_formats_alloc);
+               commit_format = &commit_formats[commit_formats_len];
+               memset(commit_format, 0, sizeof(*commit_format));
+               commit_formats_len++;
+       }
+
+       commit_format->name = xstrdup(name);
+       commit_format->format = CMIT_FMT_USERFORMAT;
+       git_config_string(&fmt, var, value);
+       if (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
+               commit_format->is_tformat = fmt[0] == 't';
+               fmt = strchr(fmt, ':') + 1;
+       } else if (strchr(fmt, '%'))
+               commit_format->is_tformat = 1;
+       else
+               commit_format->is_alias = 1;
+       commit_format->user_format = fmt;
+
+       return 0;
+}
+
+static void setup_commit_formats(void)
+{
+       struct cmt_fmt_map builtin_formats[] = {
+               { "raw",        CMIT_FMT_RAW,           0 },
+               { "medium",     CMIT_FMT_MEDIUM,        0 },
+               { "short",      CMIT_FMT_SHORT,         0 },
+               { "email",      CMIT_FMT_EMAIL,         0 },
+               { "fuller",     CMIT_FMT_FULLER,        0 },
+               { "full",       CMIT_FMT_FULL,          0 },
+               { "oneline",    CMIT_FMT_ONELINE,       1 }
        };
+       commit_formats_len = ARRAY_SIZE(builtin_formats);
+       builtin_formats_len = commit_formats_len;
+       ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
+       memcpy(commit_formats, builtin_formats,
+              sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
+
+       git_config(git_pretty_formats_config, NULL);
+}
+
+static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
+                                                       const char *original,
+                                                       int num_redirections)
+{
+       struct cmt_fmt_map *found = NULL;
+       size_t found_match_len = 0;
+       int i;
+
+       if (num_redirections >= commit_formats_len)
+               die("invalid --pretty format: "
+                   "'%s' references an alias which points to itself",
+                   original);
+
+       for (i = 0; i < commit_formats_len; i++) {
+               size_t match_len;
+
+               if (prefixcmp(commit_formats[i].name, sought))
+                       continue;
+
+               match_len = strlen(commit_formats[i].name);
+               if (found == NULL || found_match_len > match_len) {
+                       found = &commit_formats[i];
+                       found_match_len = match_len;
+               }
+       }
+
+       if (found && found->is_alias) {
+               found = find_commit_format_recursive(found->user_format,
+                                                    original,
+                                                    num_redirections+1);
+       }
+
+       return found;
+}
+
+static struct cmt_fmt_map *find_commit_format(const char *sought)
+{
+       if (!commit_formats)
+               setup_commit_formats();
+
+       return find_commit_format_recursive(sought, sought, 0);
+}
+
+void get_commit_format(const char *arg, struct rev_info *rev)
+{
+       struct cmt_fmt_map *commit_format;
 
        rev->use_terminator = 0;
        if (!arg || !*arg) {
@@ -47,21 +154,22 @@ void get_commit_format(const char *arg, struct rev_info *rev)
                save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
                return;
        }
-       for (i = 0; i < ARRAY_SIZE(cmt_fmts); i++) {
-               if (!strncmp(arg, cmt_fmts[i].n, cmt_fmts[i].cmp_len) &&
-                   !strncmp(arg, cmt_fmts[i].n, strlen(arg))) {
-                       if (cmt_fmts[i].v == CMIT_FMT_ONELINE)
-                               rev->use_terminator = 1;
-                       rev->commit_format = cmt_fmts[i].v;
-                       return;
-               }
-       }
+
        if (strchr(arg, '%')) {
                save_user_format(rev, arg, 1);
                return;
        }
 
-       die("invalid --pretty format: %s", arg);
+       commit_format = find_commit_format(arg);
+       if (!commit_format)
+               die("invalid --pretty format: %s", arg);
+
+       rev->commit_format = commit_format->format;
+       rev->use_terminator = commit_format->is_tformat;
+       if (commit_format->format == CMIT_FMT_USERFORMAT) {
+               save_user_format(rev, commit_format->user_format,
+                                commit_format->is_tformat);
+       }
 }
 
 /*
@@ -100,6 +208,58 @@ int has_non_ascii(const char *s)
        return 0;
 }
 
+static int is_rfc822_special(char ch)
+{
+       switch (ch) {
+       case '(':
+       case ')':
+       case '<':
+       case '>':
+       case '[':
+       case ']':
+       case ':':
+       case ';':
+       case '@':
+       case ',':
+       case '.':
+       case '"':
+       case '\\':
+               return 1;
+       default:
+               return 0;
+       }
+}
+
+static int has_rfc822_specials(const char *s, int len)
+{
+       int i;
+       for (i = 0; i < len; i++)
+               if (is_rfc822_special(s[i]))
+                       return 1;
+       return 0;
+}
+
+static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
+{
+       int i;
+
+       /* just a guess, we may have to also backslash-quote */
+       strbuf_grow(out, len + 2);
+
+       strbuf_addch(out, '"');
+       for (i = 0; i < len; i++) {
+               switch (s[i]) {
+               case '"':
+               case '\\':
+                       strbuf_addch(out, '\\');
+                       /* fall through */
+               default:
+                       strbuf_addch(out, s[i]);
+               }
+       }
+       strbuf_addch(out, '"');
+}
+
 static int is_rfc2047_special(char ch)
 {
        return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
@@ -108,36 +268,53 @@ static int is_rfc2047_special(char ch)
 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
                       const char *encoding)
 {
-       int i, last;
+       static const int max_length = 78; /* per rfc2822 */
+       int i;
+       int line_len;
+
+       /* How many bytes are already used on the current line? */
+       for (i = sb->len - 1; i >= 0; i--)
+               if (sb->buf[i] == '\n')
+                       break;
+       line_len = sb->len - (i+1);
 
        for (i = 0; i < len; i++) {
                int ch = line[i];
-               if (non_ascii(ch))
+               if (non_ascii(ch) || ch == '\n')
                        goto needquote;
                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
                        goto needquote;
        }
-       strbuf_add(sb, line, len);
+       strbuf_add_wrapped_bytes(sb, line, len, 0, 1, max_length - line_len);
        return;
 
 needquote:
        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
        strbuf_addf(sb, "=?%s?q?", encoding);
-       for (i = last = 0; i < len; i++) {
+       line_len += strlen(encoding) + 5; /* 5 for =??q? */
+       for (i = 0; i < len; i++) {
                unsigned ch = line[i] & 0xFF;
+
+               if (line_len >= max_length - 2) {
+                       strbuf_addf(sb, "?=\n =?%s?q?", encoding);
+                       line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
+               }
+
                /*
                 * We encode ' ' using '=20' even though rfc2047
                 * allows using '_' for readability.  Unfortunately,
                 * many programs do not understand this and just
                 * leave the underscore in place.
                 */
-               if (is_rfc2047_special(ch) || ch == ' ') {
-                       strbuf_add(sb, line + last, i - last);
+               if (is_rfc2047_special(ch) || ch == ' ' || ch == '\n') {
                        strbuf_addf(sb, "=%02X", ch);
-                       last = i + 1;
+                       line_len += 3;
+               }
+               else {
+                       strbuf_addch(sb, ch);
+                       line_len++;
                }
        }
-       strbuf_add(sb, line + last, len - last);
        strbuf_addstr(sb, "?=");
 }
 
@@ -162,13 +339,29 @@ void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
        if (fmt == CMIT_FMT_EMAIL) {
                char *name_tail = strchr(line, '<');
                int display_name_length;
+               int final_line;
                if (!name_tail)
                        return;
                while (line < name_tail && isspace(name_tail[-1]))
                        name_tail--;
                display_name_length = name_tail - line;
                strbuf_addstr(sb, "From: ");
-               add_rfc2047(sb, line, display_name_length, encoding);
+               if (!has_rfc822_specials(line, display_name_length)) {
+                       add_rfc2047(sb, line, display_name_length, encoding);
+               } else {
+                       struct strbuf quoted = STRBUF_INIT;
+                       add_rfc822_quoted(&quoted, line, display_name_length);
+                       add_rfc2047(sb, quoted.buf, quoted.len, encoding);
+                       strbuf_release(&quoted);
+               }
+               for (final_line = 0; final_line < sb->len; final_line++)
+                       if (sb->buf[sb->len - final_line - 1] == '\n')
+                               break;
+               if (namelen - display_name_length + final_line > 78) {
+                       strbuf_addch(sb, '\n');
+                       if (!isspace(name_tail[0]))
+                               strbuf_addch(sb, ' ');
+               }
                strbuf_add(sb, name_tail, namelen - display_name_length);
                strbuf_addch(sb, '\n');
        } else {
@@ -295,8 +488,8 @@ static char *replace_encoding_header(char *buf, const char *encoding)
        return strbuf_detach(&tmp, NULL);
 }
 
-static char *logmsg_reencode(const struct commit *commit,
-                            const char *output_encoding)
+char *logmsg_reencode(const struct commit *commit,
+                     const char *output_encoding)
 {
        static const char *utf8 = "UTF-8";
        const char *use_encoding;
@@ -447,6 +640,7 @@ struct format_commit_context {
        const struct pretty_print_context *pretty_ctx;
        unsigned commit_header_parsed:1;
        unsigned commit_message_parsed:1;
+       char *message;
        size_t width, indent1, indent2;
 
        /* These offsets are relative to the start of the commit message. */
@@ -483,7 +677,7 @@ static int add_again(struct strbuf *sb, struct chunk *chunk)
 
 static void parse_commit_header(struct format_commit_context *context)
 {
-       const char *msg = context->commit->buffer;
+       const char *msg = context->message;
        int i;
 
        for (i = 0; msg[i]; i++) {
@@ -569,8 +763,8 @@ const char *format_subject(struct strbuf *sb, const char *msg,
 
 static void parse_commit_message(struct format_commit_context *c)
 {
-       const char *msg = c->commit->buffer + c->message_off;
-       const char *start = c->commit->buffer;
+       const char *msg = c->message + c->message_off;
+       const char *start = c->message;
 
        msg = skip_empty_lines(msg);
        c->subject_off = msg - start;
@@ -633,7 +827,7 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 {
        struct format_commit_context *c = context;
        const struct commit *commit = c->commit;
-       const char *msg = commit->buffer;
+       const char *msg = c->message;
        struct commit_list *p;
        int h1, h2;
 
@@ -716,7 +910,7 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                if (add_again(sb, &c->abbrev_commit_hash))
                        return 1;
                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
-                                                    DEFAULT_ABBREV));
+                                                    c->pretty_ctx->abbrev));
                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
                return 1;
        case 'T':               /* tree hash */
@@ -726,7 +920,7 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                if (add_again(sb, &c->abbrev_tree_hash))
                        return 1;
                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
-                                                    DEFAULT_ABBREV));
+                                                    c->pretty_ctx->abbrev));
                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
                return 1;
        case 'P':               /* parent hashes */
@@ -743,17 +937,14 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                        if (p != commit->parents)
                                strbuf_addch(sb, ' ');
                        strbuf_addstr(sb, find_unique_abbrev(
-                                       p->item->object.sha1, DEFAULT_ABBREV));
+                                       p->item->object.sha1,
+                                       c->pretty_ctx->abbrev));
                }
                c->abbrev_parent_hashes.len = sb->len -
                                              c->abbrev_parent_hashes.off;
                return 1;
        case 'm':               /* left/right/bottom */
-               strbuf_addch(sb, (commit->object.flags & BOUNDARY)
-                                ? '-'
-                                : (commit->object.flags & SYMMETRIC_LEFT)
-                                ? '<'
-                                : '>');
+               strbuf_addstr(sb, get_revision_mark(NULL, commit));
                return 1;
        case 'd':
                format_decoration(sb, commit);
@@ -775,9 +966,12 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
                }
                return 0;       /* unknown %g placeholder */
        case 'N':
-               get_commit_notes(commit, sb, git_log_output_encoding ?
-                            git_log_output_encoding : git_commit_encoding, 0);
-               return 1;
+               if (c->pretty_ctx->show_notes) {
+                       format_display_notes(commit->object.sha1, sb,
+                                   get_log_output_encoding(), 0);
+                       return 1;
+               }
+               return 0;
        }
 
        /* For the rest we have to parse the commit header. */
@@ -796,6 +990,10 @@ static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
        case 'e':       /* encoding */
                strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
                return 1;
+       case 'B':       /* raw body */
+               /* message_off is always left at the initial newline */
+               strbuf_addstr(sb, msg + c->message_off + 1);
+               return 1;
        }
 
        /* Now we need to parse the commit message. */
@@ -825,6 +1023,7 @@ static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
                NO_MAGIC,
                ADD_LF_BEFORE_NON_EMPTY,
                DEL_LF_BEFORE_EMPTY,
+               ADD_SP_BEFORE_NON_EMPTY
        } magic = NO_MAGIC;
 
        switch (placeholder[0]) {
@@ -834,6 +1033,9 @@ static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
        case '+':
                magic = ADD_LF_BEFORE_NON_EMPTY;
                break;
+       case ' ':
+               magic = ADD_SP_BEFORE_NON_EMPTY;
+               break;
        default:
                break;
        }
@@ -848,24 +1050,70 @@ static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
                while (sb->len && sb->buf[sb->len - 1] == '\n')
                        strbuf_setlen(sb, sb->len - 1);
-       } else if ((orig_len != sb->len) && magic == ADD_LF_BEFORE_NON_EMPTY) {
-               strbuf_insert(sb, orig_len, "\n", 1);
+       } else if (orig_len != sb->len) {
+               if (magic == ADD_LF_BEFORE_NON_EMPTY)
+                       strbuf_insert(sb, orig_len, "\n", 1);
+               else if (magic == ADD_SP_BEFORE_NON_EMPTY)
+                       strbuf_insert(sb, orig_len, " ", 1);
        }
        return consumed + 1;
 }
 
+static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
+                                  void *context)
+{
+       struct userformat_want *w = context;
+
+       if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
+               placeholder++;
+
+       switch (*placeholder) {
+       case 'N':
+               w->notes = 1;
+               break;
+       }
+       return 0;
+}
+
+void userformat_find_requirements(const char *fmt, struct userformat_want *w)
+{
+       struct strbuf dummy = STRBUF_INIT;
+
+       if (!fmt) {
+               if (!user_format)
+                       return;
+               fmt = user_format;
+       }
+       strbuf_expand(&dummy, user_format, userformat_want_item, w);
+       strbuf_release(&dummy);
+}
+
 void format_commit_message(const struct commit *commit,
                           const char *format, struct strbuf *sb,
                           const struct pretty_print_context *pretty_ctx)
 {
        struct format_commit_context context;
+       static const char utf8[] = "UTF-8";
+       const char *enc;
+       const char *output_enc = pretty_ctx->output_encoding;
 
        memset(&context, 0, sizeof(context));
        context.commit = commit;
        context.pretty_ctx = pretty_ctx;
        context.wrap_start = sb->len;
+       context.message = commit->buffer;
+       if (output_enc) {
+               enc = get_header(commit, "encoding");
+               enc = enc ? enc : utf8;
+               if (strcmp(enc, output_enc))
+                       context.message = logmsg_reencode(commit, output_enc);
+       }
+
        strbuf_expand(sb, format, format_commit_item, &context);
        rewrap_message_tail(sb, &context, 0, 0, 0);
+
+       if (context.message != commit->buffer)
+               free(context.message);
 }
 
 static void pp_header(enum cmit_fmt fmt,
@@ -939,11 +1187,10 @@ void pp_title_line(enum cmit_fmt fmt,
                   const char *encoding,
                   int need_8bit_cte)
 {
-       const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
        struct strbuf title;
 
        strbuf_init(&title, 80);
-       *msg_p = format_subject(&title, *msg_p, line_separator);
+       *msg_p = format_subject(&title, *msg_p, " ");
 
        strbuf_grow(sb, title.len + 1024);
        if (subject) {
@@ -1006,11 +1253,7 @@ char *reencode_commit_message(const struct commit *commit, const char **encoding
 {
        const char *encoding;
 
-       encoding = (git_log_output_encoding
-                   ? git_log_output_encoding
-                   : git_commit_encoding);
-       if (!encoding)
-               encoding = "UTF-8";
+       encoding = get_log_output_encoding();
        if (encoding_p)
                *encoding_p = encoding;
        return logmsg_reencode(commit, encoding);
@@ -1095,8 +1338,8 @@ void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
                strbuf_addch(sb, '\n');
 
        if (context->show_notes)
-               get_commit_notes(commit, sb, encoding,
-                                NOTES_SHOW_HEADER | NOTES_INDENT);
+               format_display_notes(commit->object.sha1, sb, encoding,
+                                    NOTES_SHOW_HEADER | NOTES_INDENT);
 
        free(reencoded);
 }