aboutsummaryrefslogtreecommitdiffstats
path: root/e-util
diff options
context:
space:
mode:
authorDan Winship <danw@src.gnome.org>2003-01-07 02:55:31 +0800
committerDan Winship <danw@src.gnome.org>2003-01-07 02:55:31 +0800
commit24ee471cb154a84a004efc7ec92dde1c1474426c (patch)
tree0930170c07fac4ee55c2a77f7223e9b3e1ea581f /e-util
parent987ac08737af5022f528cebf57cef11719bd4a56 (diff)
downloadgsoc2013-evolution-24ee471cb154a84a004efc7ec92dde1c1474426c.tar.gz
gsoc2013-evolution-24ee471cb154a84a004efc7ec92dde1c1474426c.tar.zst
gsoc2013-evolution-24ee471cb154a84a004efc7ec92dde1c1474426c.zip
Add this back: Connector needs it, and there's no reason for addressbook
* e-html-utils.c: Add this back: Connector needs it, and there's no reason for addressbook and calendar to be using camel to get this functionality. * Makefile.am (libeutil_la_SOURCES): re-add e-html-utils svn path=/trunk/; revision=19246
Diffstat (limited to 'e-util')
-rw-r--r--e-util/ChangeLog8
-rw-r--r--e-util/Makefile.am2
-rw-r--r--e-util/e-html-utils.c408
-rw-r--r--e-util/e-html-utils.h39
4 files changed, 457 insertions, 0 deletions
diff --git a/e-util/ChangeLog b/e-util/ChangeLog
index f9139f7fb6..2087a74f19 100644
--- a/e-util/ChangeLog
+++ b/e-util/ChangeLog
@@ -1,3 +1,11 @@
+2003-01-06 Dan Winship <danw@ximian.com>
+
+ * e-html-utils.c: Add this back: Connector needs it, and there's
+ no reason for addressbook and calendar to be using camel to get
+ this functionality.
+
+ * Makefile.am (libeutil_la_SOURCES): re-add e-html-utils
+
2003-01-02 Jeffrey Stedfast <fejj@ximian.com>
A bunch of fixes to make e-util build with -DG_DISABLE_DEPRECATED
diff --git a/e-util/Makefile.am b/e-util/Makefile.am
index dfeea653a7..89fea7550d 100644
--- a/e-util/Makefile.am
+++ b/e-util/Makefile.am
@@ -30,6 +30,7 @@ eutilinclude_HEADERS = \
e-gtk-utils.h \
e-gui-utils.h \
e-host-utils.h \
+ e-html-utils.h \
e-iterator.h \
e-lang-utils.h \
e-list-iterator.h \
@@ -59,6 +60,7 @@ libeutil_la_SOURCES = \
e-gtk-utils.c \
e-gui-utils.c \
e-host-utils.c \
+ e-html-utils.c \
e-iterator.c \
e-lang-utils.c \
e-list-iterator.c \
diff --git a/e-util/e-html-utils.c b/e-util/e-html-utils.c
new file mode 100644
index 0000000000..12aee18d2b
--- /dev/null
+++ b/e-util/e-html-utils.c
@@ -0,0 +1,408 @@
+/* -*- Mode: C; tab-width: 8; indent-tabs-mode: t; c-basic-offset: 8 -*- */
+/* e-html-utils.c
+ * Copyright (C) 2000 Ximian, Inc.
+ * Author: Dan Winship <danw@ximian.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this library; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 02111-1307, USA.
+ */
+
+#include <ctype.h>
+#include <stdio.h>
+#include <string.h>
+#include <glib.h>
+
+#include "e-html-utils.h"
+
+static char *
+check_size (char **buffer, int *buffer_size, char *out, int len)
+{
+ if (out + len + 1> *buffer + *buffer_size) {
+ int index = out - *buffer;
+
+ *buffer_size = MAX (index + len + 1, *buffer_size * 2);
+ *buffer = g_realloc (*buffer, *buffer_size);
+ out = *buffer + index;
+ }
+ return out;
+}
+
+/* 1 = non-email-address chars: ()<>@,;:\"[]`'{}| */
+/* 2 = trailing url garbage: ,.!?;:>)]}`'-_| */
+/* 4 = dns chars */
+static int special_chars[] = {
+ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, /* nul - 0x0f */
+ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, /* 0x10 - 0x1f */
+ 1, 2, 1, 0, 0, 0, 0, 3, 1, 3, 0, 0, 3, 6, 6, 0, /* sp - / */
+ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 1, 0, 3, 2, /* 0 - ? */
+ 1, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, /* @ - O */
+ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 1, 1, 3, 0, 2, /* P - _ */
+ 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, /* ` - o */
+ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 1, 3, 3, 0, 3 /* p - del */
+};
+
+#define is_addr_char(c) (c < 128 && !(special_chars[c] & 1))
+#define is_trailing_garbage(c) (c > 127 || (special_chars[c] & 2))
+#define is_domain_name_char(c) (c < 128 && (special_chars[c] & 4))
+
+static char *
+url_extract (const unsigned char **text, gboolean check)
+{
+ const unsigned char *end = *text, *p;
+ char *out;
+
+ while (*end && !isspace (*end) && (*end != '"') && (*end < 0x80))
+ end++;
+
+ /* Back up if we probably went too far. */
+ while (end > *text && is_trailing_garbage (*(end - 1)))
+ end--;
+
+ if (check) {
+ /* Make sure we weren't fooled. */
+ p = memchr (*text, ':', end - *text);
+ if (!p || end - p < 4)
+ return NULL;
+ }
+
+ out = g_strndup (*text, end - *text);
+ *text = end;
+ return out;
+}
+
+static char *
+email_address_extract (const unsigned char **cur, char **out, const unsigned char *linestart)
+{
+ const unsigned char *start, *end, *dot;
+ char *addr;
+
+ /* *cur points to the '@'. Look backward for a valid local-part */
+ for (start = *cur; start - 1 >= linestart && is_addr_char (*(start - 1)); start--)
+ ;
+ if (start == *cur)
+ return NULL;
+
+ /* Now look forward for a valid domain part */
+ for (end = *cur + 1, dot = NULL; is_domain_name_char (*end); end++) {
+ if (*end == '.' && !dot)
+ dot = end;
+ }
+ if (!dot)
+ return NULL;
+
+ /* Remove trailing garbage */
+ while (is_trailing_garbage (*(end - 1)))
+ end--;
+ if (dot > end)
+ return NULL;
+
+ addr = g_strndup (start, end - start);
+ *out -= *cur - start;
+ *cur = end;
+
+ return addr;
+}
+
+static gboolean
+is_citation (const unsigned char *c, gboolean saw_citation)
+{
+ const unsigned char *p;
+
+ if (*c != '>')
+ return FALSE;
+
+ /* A line that starts with a ">" is a citation, unless it's
+ * just mbox From-mangling...
+ */
+ if (strncmp (c, ">From ", 6) != 0)
+ return TRUE;
+
+ /* If the previous line was a citation, then say this
+ * one is too.
+ */
+ if (saw_citation)
+ return TRUE;
+
+ /* Same if the next line is */
+ p = (const unsigned char *)strchr ((const char *)c, '\n');
+ if (p && *++p == '>')
+ return TRUE;
+
+ /* Otherwise, it was just an isolated ">From" line. */
+ return FALSE;
+}
+
+/**
+ * e_text_to_html_full:
+ * @input: a NUL-terminated input buffer
+ * @flags: some combination of the E_TEXT_TO_HTML_* flags defined
+ * in e-html-utils.h
+ * @color: color for citation highlighting
+ *
+ * This takes a buffer of text as input and produces a buffer of
+ * "equivalent" HTML, subject to certain transformation rules.
+ *
+ * The set of possible flags is:
+ *
+ * - E_TEXT_TO_HTML_PRE: wrap the output HTML in <PRE> and </PRE>.
+ * Should only be used if @input is the entire buffer to be
+ * converted. If e_text_to_html is being called with small pieces
+ * of data, you should wrap the entire result in <PRE> yourself.
+ *
+ * - E_TEXT_TO_HTML_CONVERT_NL: convert "\n" to "<BR>\n" on output.
+ * (should not be used with E_TEXT_TO_HTML_PRE, since that would
+ * result in double-newlines).
+ *
+ * - E_TEXT_TO_HTML_CONVERT_SPACES: convert a block of N spaces
+ * into N-1 non-breaking spaces and one normal space. A space
+ * at the start of the buffer is always converted to a
+ * non-breaking space, regardless of the following character,
+ * which probably means you don't want to use this flag on
+ * pieces of data that aren't delimited by at least line breaks.
+ *
+ * If E_TEXT_TO_HTML_CONVERT_NL and E_TEXT_TO_HTML_CONVERT_SPACES
+ * are both defined, then TABs will also be converted to spaces.
+ *
+ * - E_TEXT_TO_HTML_CONVERT_URLS: wrap <a href="..."> </a> around
+ * strings that look like URLs.
+ *
+ * - E_TEXT_TO_HTML_CONVERT_ADDRESSES: wrap <a href="mailto:..."> </a> around
+ * strings that look like mail addresses.
+ *
+ * - E_TEXT_TO_HTML_MARK_CITATION: wrap <font color="..."> </font> around
+ * citations (lines beginning with "> ", etc).
+ *
+ * - E_TEXT_TO_HTML_ESCAPE_8BIT: flatten everything to US-ASCII
+ *
+ * - E_TEXT_TO_HTML_CITE: quote the text with "> " at the start of each
+ * line.
+ **/
+char *
+e_text_to_html_full (const char *input, unsigned int flags, guint32 color)
+{
+ const unsigned char *cur, *next, *linestart;
+ char *buffer = NULL;
+ char *out = NULL;
+ int buffer_size = 0, col;
+ gboolean colored = FALSE, saw_citation = FALSE;
+
+ /* Allocate a translation buffer. */
+ buffer_size = strlen (input) * 2 + 5;
+ buffer = g_malloc (buffer_size);
+
+ out = buffer;
+ if (flags & E_TEXT_TO_HTML_PRE)
+ out += sprintf (out, "<PRE>");
+
+ col = 0;
+
+ for (cur = linestart = input; cur && *cur; cur = next) {
+ gunichar u;
+
+ if (flags & E_TEXT_TO_HTML_MARK_CITATION && col == 0) {
+ saw_citation = is_citation (cur, saw_citation);
+ if (saw_citation) {
+ if (!colored) {
+ gchar font [25];
+
+ g_snprintf (font, 25, "<FONT COLOR=\"#%06x\">", color);
+
+ out = check_size (&buffer, &buffer_size, out, 25);
+ out += sprintf (out, "%s", font);
+ colored = TRUE;
+ }
+ } else if (colored) {
+ gchar *no_font = "</FONT>";
+
+ out = check_size (&buffer, &buffer_size, out, 9);
+ out += sprintf (out, "%s", no_font);
+ colored = FALSE;
+ }
+
+ /* Display mbox-mangled ">From" as "From" */
+ if (*cur == '>' && !saw_citation)
+ cur++;
+ } else if (flags & E_TEXT_TO_HTML_CITE && col == 0) {
+ out = check_size (&buffer, &buffer_size, out, 5);
+ out += sprintf (out, "&gt; ");
+ }
+
+ u = g_utf8_get_char (cur);
+ if (g_unichar_isalpha (u) &&
+ (flags & E_TEXT_TO_HTML_CONVERT_URLS)) {
+ char *tmpurl = NULL, *refurl = NULL, *dispurl = NULL;
+
+ if (!strncasecmp (cur, "http://", 7) ||
+ !strncasecmp (cur, "https://", 8) ||
+ !strncasecmp (cur, "ftp://", 6) ||
+ !strncasecmp (cur, "nntp://", 7) ||
+ !strncasecmp (cur, "mailto:", 7) ||
+ !strncasecmp (cur, "news:", 5) ||
+ !strncasecmp (cur, "file:", 5)) {
+ tmpurl = url_extract (&cur, TRUE);
+ if (tmpurl) {
+ refurl = e_text_to_html (tmpurl, 0);
+ dispurl = g_strdup (refurl);
+ }
+ } else if (!strncasecmp (cur, "www.", 4) &&
+ (*(cur + 4) < 0x80) &&
+ g_unichar_isalnum (*(cur + 4))) {
+ tmpurl = url_extract (&cur, FALSE);
+ dispurl = e_text_to_html (tmpurl, 0);
+ refurl = g_strdup_printf ("http://%s",
+ dispurl);
+ }
+
+ if (tmpurl) {
+ out = check_size (&buffer, &buffer_size, out,
+ strlen (refurl) +
+ strlen (dispurl) + 15);
+ out += sprintf (out,
+ "<a href=\"%s\">%s</a>",
+ refurl, dispurl);
+ col += strlen (tmpurl);
+ g_free (tmpurl);
+ g_free (refurl);
+ g_free (dispurl);
+ }
+
+ if (!*cur)
+ break;
+ u = g_utf8_get_char (cur);
+ }
+
+ if (u == '@' && (flags & E_TEXT_TO_HTML_CONVERT_ADDRESSES)) {
+ char *addr, *dispaddr, *outaddr;
+
+ addr = email_address_extract (&cur, &out, linestart);
+ if (addr) {
+ dispaddr = e_text_to_html (addr, 0);
+ outaddr = g_strdup_printf ("<a href=\"mailto:%s\">%s</a>",
+ addr, dispaddr);
+ out = check_size (&buffer, &buffer_size, out, strlen (outaddr));
+ out += sprintf (out, "%s", outaddr);
+ col += strlen (addr);
+ g_free (addr);
+ g_free (dispaddr);
+ g_free (outaddr);
+
+ if (!*cur)
+ break;
+ u = g_utf8_get_char (cur);
+ }
+ }
+
+ if (!g_unichar_validate (u)) {
+ /* Sigh. Someone sent undeclared 8-bit data.
+ * Assume it's iso-8859-1.
+ */
+ u = *cur;
+ next = cur + 1;
+ } else
+ next = g_utf8_next_char (cur);
+
+ out = check_size (&buffer, &buffer_size, out, 10);
+
+ switch (u) {
+ case '<':
+ strcpy (out, "&lt;");
+ out += 4;
+ col++;
+ break;
+
+ case '>':
+ strcpy (out, "&gt;");
+ out += 4;
+ col++;
+ break;
+
+ case '&':
+ strcpy (out, "&amp;");
+ out += 5;
+ col++;
+ break;
+
+ case '"':
+ strcpy (out, "&quot;");
+ out += 6;
+ col++;
+ break;
+
+ case '\n':
+ if (flags & E_TEXT_TO_HTML_CONVERT_NL) {
+ strcpy (out, "<br>");
+ out += 4;
+ }
+ *out++ = *cur;
+ linestart = cur;
+ col = 0;
+ break;
+
+ case '\t':
+ if (flags & (E_TEXT_TO_HTML_CONVERT_SPACES |
+ E_TEXT_TO_HTML_CONVERT_NL)) {
+ do {
+ out = check_size (&buffer, &buffer_size,
+ out, 7);
+ strcpy (out, "&nbsp;");
+ out += 6;
+ col++;
+ } while (col % 8);
+ break;
+ }
+ /* otherwise, FALL THROUGH */
+
+ case ' ':
+ if (flags & E_TEXT_TO_HTML_CONVERT_SPACES) {
+ if (cur == (const unsigned char *)input ||
+ *(cur + 1) == ' ' || *(cur + 1) == '\t' ||
+ *(cur - 1) == '\n') {
+ strcpy (out, "&nbsp;");
+ out += 6;
+ col++;
+ break;
+ }
+ }
+ /* otherwise, FALL THROUGH */
+
+ default:
+ if ((u >= 0x20 && u < 0x80) ||
+ (u == '\r' || u == '\t')) {
+ /* Default case, just copy. */
+ *out++ = u;
+ } else {
+ if (flags & E_TEXT_TO_HTML_ESCAPE_8BIT)
+ *out++ = '?';
+ else
+ out += g_snprintf(out, 9, "&#%d;", u);
+ }
+ col++;
+ break;
+ }
+ }
+
+ out = check_size (&buffer, &buffer_size, out, 7);
+ if (flags & E_TEXT_TO_HTML_PRE)
+ strcpy (out, "</PRE>");
+ else
+ *out = '\0';
+
+ return buffer;
+}
+
+char *
+e_text_to_html (const char *input, unsigned int flags)
+{
+ return e_text_to_html_full (input, flags, 0);
+}
diff --git a/e-util/e-html-utils.h b/e-util/e-html-utils.h
new file mode 100644
index 0000000000..551996e089
--- /dev/null
+++ b/e-util/e-html-utils.h
@@ -0,0 +1,39 @@
+/* -*- Mode: C; tab-width: 8; indent-tabs-mode: t; c-basic-offset: 8 -*- */
+/* e-html-utils.c */
+/*
+ * Copyright (C) 2000 Ximian, Inc.
+ * Author: Dan Winship <danw@ximian.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this library; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __E_HTML_UTILS__
+#define __E_HTML_UTILS__
+
+#include <glib.h>
+
+#define E_TEXT_TO_HTML_PRE (1 << 0)
+#define E_TEXT_TO_HTML_CONVERT_NL (1 << 1)
+#define E_TEXT_TO_HTML_CONVERT_SPACES (1 << 2)
+#define E_TEXT_TO_HTML_CONVERT_URLS (1 << 3)
+#define E_TEXT_TO_HTML_MARK_CITATION (1 << 4)
+#define E_TEXT_TO_HTML_CONVERT_ADDRESSES (1 << 5)
+#define E_TEXT_TO_HTML_ESCAPE_8BIT (1 << 6)
+#define E_TEXT_TO_HTML_CITE (1 << 7)
+
+char *e_text_to_html_full (const char *input, unsigned int flags, guint32 color);
+char *e_text_to_html (const char *input, unsigned int flags);
+
+#endif /* __E_HTML_UTILS__ */