mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-09 19:09:41 +00:00
1875564b65
Original commit message from CVS: * configure.ac: * gst/subparse/Makefile.am: * gst/subparse/gstsubparse.c: * gst/subparse/samiparse.c: * tests/check/elements/subparse.c: Rework last change, so that we build subparse, but just disable the sami parse functionality, if we're configured to not use xml. In the tests only the sami test is disabled now.
466 lines
15 KiB
C
466 lines
15 KiB
C
/* GStreamer SAMI subtitle parser
|
|
* Copyright (c) 2006 Young-Ho Cha <ganadist at chollian net>
|
|
*
|
|
* This library is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Library General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2 of the License, or (at your option) any later version.
|
|
*
|
|
* This library 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
|
|
* Library General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Library 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 "samiparse.h"
|
|
|
|
#include <libxml/HTMLparser.h>
|
|
#include <string.h>
|
|
|
|
#define ITALIC_TAG 'i'
|
|
#define SPAN_TAG 's'
|
|
#define RUBY_TAG 'r'
|
|
#define RT_TAG 't'
|
|
#define CLEAR_TAG '0'
|
|
|
|
typedef struct _GstSamiContext GstSamiContext;
|
|
|
|
struct _GstSamiContext
|
|
{
|
|
GString *buf; /* buffer to collect content */
|
|
GString *rubybuf; /* buffer to collect ruby content */
|
|
GString *resultbuf; /* when opening the next 'sync' tag, move
|
|
* from 'buf' to avoid to append following
|
|
* content */
|
|
GString *state; /* in many sami files there are tags that
|
|
* are not closed, so for each open tag the
|
|
* parser will append a tag flag here so
|
|
* that tags can be closed properly on
|
|
* 'sync' tags. See _context_push_state()
|
|
* and _context_pop_state(). */
|
|
htmlParserCtxtPtr htmlctxt; /* html parser context */
|
|
gboolean has_result; /* set when ready to push out result */
|
|
gboolean in_sync; /* flag to avoid appending anything except the
|
|
* content of the sync elements to buf */
|
|
guint64 time1; /* previous start attribute in sync tag */
|
|
guint64 time2; /* current start attribute in sync tag */
|
|
};
|
|
|
|
static gchar *
|
|
has_tag (GString * str, const gchar tag)
|
|
{
|
|
return strrchr (str->str, tag);
|
|
}
|
|
|
|
static void
|
|
sami_context_push_state (GstSamiContext * sctx, char state)
|
|
{
|
|
g_string_append_c (sctx->state, state);
|
|
}
|
|
|
|
static void
|
|
sami_context_pop_state (GstSamiContext * sctx, char state)
|
|
{
|
|
GString *str = g_string_new ("");
|
|
GString *context_state = sctx->state;
|
|
int i;
|
|
|
|
for (i = context_state->len - 1; i >= 0; i--) {
|
|
switch (context_state->str[i]) {
|
|
case ITALIC_TAG: /* <i> */
|
|
{
|
|
g_string_append (str, "</i>");
|
|
break;
|
|
}
|
|
case SPAN_TAG: /* <span foreground= > */
|
|
{
|
|
g_string_append (str, "</span>");
|
|
break;
|
|
}
|
|
case RUBY_TAG: /* <span size= > -- ruby */
|
|
{
|
|
break;
|
|
}
|
|
case RT_TAG: /* ruby */
|
|
{
|
|
/* FIXME: support for furigana/ruby once implemented in pango */
|
|
g_string_append (sctx->rubybuf, "</span>");
|
|
if (has_tag (context_state, ITALIC_TAG)) {
|
|
g_string_append (sctx->rubybuf, "</i>");
|
|
}
|
|
|
|
break;
|
|
}
|
|
default:
|
|
break;
|
|
}
|
|
if (context_state->str[i] == state) {
|
|
g_string_append (sctx->buf, str->str);
|
|
g_string_free (str, TRUE);
|
|
g_string_truncate (context_state, i);
|
|
return;
|
|
}
|
|
}
|
|
if (state == CLEAR_TAG) {
|
|
g_string_append (sctx->buf, str->str);
|
|
g_string_truncate (context_state, 0);
|
|
}
|
|
g_string_free (str, TRUE);
|
|
}
|
|
|
|
static void
|
|
handle_start_sync (GstSamiContext * sctx, const xmlChar ** atts)
|
|
{
|
|
int i;
|
|
|
|
sami_context_pop_state (sctx, CLEAR_TAG);
|
|
if (atts != NULL) {
|
|
for (i = 0; (atts[i] != NULL); i += 2) {
|
|
const xmlChar *key, *value;
|
|
|
|
key = atts[i];
|
|
value = atts[i + 1];
|
|
|
|
if (!value)
|
|
continue;
|
|
if (!xmlStrncmp ((const xmlChar *) "start", key, 5)) {
|
|
/* Only set a new start time if we don't have text pending */
|
|
if (sctx->resultbuf->len == 0)
|
|
sctx->time1 = sctx->time2;
|
|
|
|
sctx->time2 = atoi ((const char *) value) * GST_MSECOND;
|
|
g_string_append (sctx->resultbuf, sctx->buf->str);
|
|
sctx->has_result = (sctx->resultbuf->len != 0) ? TRUE : FALSE;
|
|
g_string_truncate (sctx->buf, 0);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static void
|
|
handle_start_font (GstSamiContext * sctx, const xmlChar ** atts)
|
|
{
|
|
int i;
|
|
|
|
sami_context_pop_state (sctx, SPAN_TAG);
|
|
if (atts != NULL) {
|
|
g_string_append (sctx->buf, "<span");
|
|
for (i = 0; (atts[i] != NULL); i += 2) {
|
|
const xmlChar *key, *value;
|
|
|
|
key = atts[i];
|
|
value = atts[i + 1];
|
|
|
|
if (!value)
|
|
continue;
|
|
if (!xmlStrncmp ((const xmlChar *) "color", key, 5)) {
|
|
/*
|
|
* There are invalid color value in many
|
|
* sami files.
|
|
* It will fix hex color value that start without '#'
|
|
*/
|
|
gchar *sharp = "";
|
|
int len = xmlStrlen (value);
|
|
|
|
if (!(*value == '#' && len == 7)) {
|
|
gchar *r;
|
|
|
|
/* check if it looks like hex */
|
|
if (strtol ((const char *) value, &r, 16) >= 0 &&
|
|
((xmlChar *) r == (value + 6) && len == 6)) {
|
|
sharp = "#";
|
|
}
|
|
}
|
|
/* some colours can be found in many sami files, but X RGB database
|
|
* doesn't contain a colour by this name, so map explicitly */
|
|
if (!xmlStrncasecmp (value, (const xmlChar *) "aqua", len)) {
|
|
value = (const xmlChar *) "#00ffff";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "crimson", len)) {
|
|
value = (const xmlChar *) "#dc143c";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "fuchsia", len)) {
|
|
value = (const xmlChar *) "#ff00ff";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "indigo", len)) {
|
|
value = (const xmlChar *) "#4b0082";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "lime", len)) {
|
|
value = (const xmlChar *) "#00ff00";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "olive", len)) {
|
|
value = (const xmlChar *) "#808000";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "silver", len)) {
|
|
value = (const xmlChar *) "#c0c0c0";
|
|
} else if (!xmlStrncasecmp (value, (const xmlChar *) "teal", len)) {
|
|
value = (const xmlChar *) "#008080";
|
|
}
|
|
g_string_append_printf (sctx->buf, " foreground=\"%s%s\"", sharp,
|
|
value);
|
|
} else if (!xmlStrncasecmp ((const xmlChar *) "face", key, 4)) {
|
|
g_string_append_printf (sctx->buf, " font_family=\"%s\"", value);
|
|
}
|
|
}
|
|
g_string_append_c (sctx->buf, '>');
|
|
sami_context_push_state (sctx, SPAN_TAG);
|
|
}
|
|
}
|
|
|
|
static void
|
|
start_sami_element (void *ctx, const xmlChar * name, const xmlChar ** atts)
|
|
{
|
|
GstSamiContext *sctx = (GstSamiContext *) ctx;
|
|
|
|
if (!xmlStrncmp ((const xmlChar *) "sync", name, 4)) {
|
|
handle_start_sync (sctx, atts);
|
|
sctx->in_sync = TRUE;
|
|
} else if (!xmlStrncmp ((const xmlChar *) "font", name, 4)) {
|
|
handle_start_font (sctx, atts);
|
|
} else if (!xmlStrncmp ((const xmlChar *) "ruby", name, 4)) {
|
|
sami_context_push_state (sctx, RUBY_TAG);
|
|
} else if (!xmlStrncmp ((const xmlChar *) "br", name, 2)) {
|
|
g_string_append_c (sctx->buf, '\n');
|
|
/* FIXME: support for furigana/ruby once implemented in pango */
|
|
} else if (!xmlStrncmp ((const xmlChar *) "rt", name, 2)) {
|
|
if (has_tag (sctx->state, ITALIC_TAG)) {
|
|
g_string_append (sctx->rubybuf, "<i>");
|
|
}
|
|
g_string_append (sctx->rubybuf, "<span size='xx-small' rise='-100'>");
|
|
sami_context_push_state (sctx, RT_TAG);
|
|
} else if (!xmlStrncmp ((const xmlChar *) "p", name, 1)) {
|
|
} else if (!xmlStrncmp ((const xmlChar *) "i", name, 1)) {
|
|
g_string_append (sctx->buf, "<i>");
|
|
sami_context_push_state (sctx, ITALIC_TAG);
|
|
}
|
|
}
|
|
|
|
static void
|
|
end_sami_element (void *ctx, const xmlChar * name)
|
|
{
|
|
GstSamiContext *sctx = (GstSamiContext *) ctx;
|
|
|
|
if (!xmlStrncmp ((const xmlChar *) "sync", name, 4)) {
|
|
sctx->in_sync = FALSE;
|
|
} else if (!xmlStrncmp ((const xmlChar *) "body", name, 4)) {
|
|
/* We will usually have one buffer left when the body is closed
|
|
* as we need the next sync to actually send it */
|
|
if (sctx->buf->len != 0) {
|
|
/* Only set a new start time if we don't have text pending */
|
|
if (sctx->resultbuf->len == 0)
|
|
sctx->time1 = sctx->time2;
|
|
|
|
sctx->time2 = GST_CLOCK_TIME_NONE;
|
|
g_string_append (sctx->resultbuf, sctx->buf->str);
|
|
sctx->has_result = (sctx->resultbuf->len != 0) ? TRUE : FALSE;
|
|
g_string_truncate (sctx->buf, 0);
|
|
}
|
|
} else if (!xmlStrncmp ((const xmlChar *) "font", name, 4)) {
|
|
sami_context_pop_state (sctx, SPAN_TAG);
|
|
} else if (!xmlStrncmp ((const xmlChar *) "ruby", name, 4)) {
|
|
sami_context_pop_state (sctx, RUBY_TAG);
|
|
} else if (!xmlStrncmp ((const xmlChar *) "i", name, 1)) {
|
|
sami_context_pop_state (sctx, ITALIC_TAG);
|
|
}
|
|
}
|
|
|
|
static void
|
|
characters_sami (void *ctx, const xmlChar * ch, int len)
|
|
{
|
|
GstSamiContext *sctx = (GstSamiContext *) ctx;
|
|
gchar *escaped;
|
|
gchar *tmp;
|
|
gint i;
|
|
|
|
/* Skip everything except content of the sync elements */
|
|
if (!sctx->in_sync)
|
|
return;
|
|
|
|
escaped = g_markup_escape_text ((const gchar *) ch, len);
|
|
g_strstrip (escaped);
|
|
|
|
/* Remove double spaces forom the string as those are
|
|
* usually added by newlines and indention */
|
|
tmp = escaped;
|
|
for (i = 0; i <= strlen (escaped); i++) {
|
|
escaped[i] = *tmp;
|
|
if (*tmp != ' ') {
|
|
tmp++;
|
|
continue;
|
|
}
|
|
while (*tmp == ' ')
|
|
tmp++;
|
|
}
|
|
|
|
if (has_tag (sctx->state, RT_TAG)) {
|
|
g_string_append_c (sctx->rubybuf, ' ');
|
|
g_string_append (sctx->rubybuf, escaped);
|
|
g_string_append_c (sctx->rubybuf, ' ');
|
|
} else {
|
|
g_string_append (sctx->buf, escaped);
|
|
}
|
|
g_free (escaped);
|
|
}
|
|
|
|
static xmlSAXHandler samiSAXHandlerStruct = {
|
|
NULL, /* internalSubset */
|
|
NULL, /* isStandalone */
|
|
NULL, /* hasInternalSubset */
|
|
NULL, /* hasExternalSubset */
|
|
NULL, /* resolveEntity */
|
|
NULL, /* getEntity */
|
|
NULL, /* entityDecl */
|
|
NULL, /* notationDecl */
|
|
NULL, /* attributeDecl */
|
|
NULL, /* elementDecl */
|
|
NULL, /* unparsedEntityDecl */
|
|
NULL, /* setDocumentLocator */
|
|
NULL, /* startDocument */
|
|
NULL, /* endDocument */
|
|
start_sami_element, /* startElement */
|
|
end_sami_element, /* endElement */
|
|
NULL, /* reference */
|
|
characters_sami, /* characters */
|
|
NULL, /* ignorableWhitespace */
|
|
NULL, /* processingInstruction */
|
|
NULL, /* comment */
|
|
NULL, /* xmlParserWarning */
|
|
NULL, /* xmlParserError */
|
|
NULL, /* xmlParserError */
|
|
NULL, /* getParameterEntity */
|
|
NULL, /* cdataBlock */
|
|
NULL, /* externalSubset */
|
|
1, /* initialized */
|
|
NULL, /* private */
|
|
NULL, /* startElementNsSAX2Func */
|
|
NULL, /* endElementNsSAX2Func */
|
|
NULL /* xmlStructuredErrorFunc */
|
|
};
|
|
static xmlSAXHandlerPtr samiSAXHandler = &samiSAXHandlerStruct;
|
|
|
|
void
|
|
sami_context_init (ParserState * state)
|
|
{
|
|
GstSamiContext *context;
|
|
|
|
g_assert (state->user_data == NULL);
|
|
state->user_data = (gpointer) g_new0 (GstSamiContext, 1);
|
|
context = (GstSamiContext *) state->user_data;
|
|
|
|
context->htmlctxt = htmlCreatePushParserCtxt (samiSAXHandler, context,
|
|
"", 0, NULL, XML_CHAR_ENCODING_UTF8);
|
|
context->buf = g_string_new ("");
|
|
context->rubybuf = g_string_new ("");
|
|
context->resultbuf = g_string_new ("");
|
|
context->state = g_string_new ("");
|
|
}
|
|
|
|
void
|
|
sami_context_deinit (ParserState * state)
|
|
{
|
|
GstSamiContext *context = (GstSamiContext *) state->user_data;
|
|
|
|
if (context) {
|
|
htmlParserCtxtPtr htmlctxt = context->htmlctxt;
|
|
|
|
/* destroy sax context */
|
|
htmlDocPtr doc;
|
|
|
|
htmlParseChunk (htmlctxt, "", 0, 1);
|
|
doc = htmlctxt->myDoc;
|
|
htmlFreeParserCtxt (htmlctxt);
|
|
context->htmlctxt = NULL;
|
|
if (doc)
|
|
xmlFreeDoc (doc);
|
|
g_string_free (context->buf, TRUE);
|
|
g_string_free (context->rubybuf, TRUE);
|
|
g_string_free (context->resultbuf, TRUE);
|
|
g_string_free (context->state, TRUE);
|
|
g_free (context);
|
|
state->user_data = NULL;
|
|
}
|
|
}
|
|
|
|
void
|
|
sami_context_reset (ParserState * state)
|
|
{
|
|
GstSamiContext *context = (GstSamiContext *) state->user_data;
|
|
|
|
if (context) {
|
|
g_string_truncate (context->buf, 0);
|
|
g_string_truncate (context->rubybuf, 0);
|
|
g_string_truncate (context->resultbuf, 0);
|
|
g_string_truncate (context->state, 0);
|
|
context->has_result = FALSE;
|
|
context->in_sync = FALSE;
|
|
context->time1 = 0;
|
|
context->time2 = 0;
|
|
}
|
|
}
|
|
|
|
static gchar *
|
|
fix_invalid_entities (const gchar * line)
|
|
{
|
|
const gchar *cp, *pp; /* current pointer, previous pointer */
|
|
gssize size;
|
|
GString *ret = g_string_new (NULL);
|
|
|
|
pp = line;
|
|
cp = strchr (line, '&');
|
|
while (cp) {
|
|
size = cp - pp;
|
|
ret = g_string_append_len (ret, pp, size);
|
|
cp++;
|
|
if (g_ascii_strncasecmp (cp, "nbsp;", 5)
|
|
&& (!g_ascii_strncasecmp (cp, "nbsp", 4))) {
|
|
/* translate " " to " " */
|
|
ret = g_string_append_len (ret, " ", 6);
|
|
cp += 4;
|
|
} else if (g_ascii_strncasecmp (cp, "quot;", 5)
|
|
&& g_ascii_strncasecmp (cp, "amp;", 4)
|
|
&& g_ascii_strncasecmp (cp, "apos;", 5)
|
|
&& g_ascii_strncasecmp (cp, "lt;", 3)
|
|
&& g_ascii_strncasecmp (cp, "gt;", 3)
|
|
&& g_ascii_strncasecmp (cp, "nbsp;", 5)
|
|
&& cp[0] != '#') {
|
|
/* translate "&" to "&" */
|
|
ret = g_string_append_len (ret, "&", 5);
|
|
} else {
|
|
/* do not translate */
|
|
ret = g_string_append_c (ret, '&');
|
|
}
|
|
|
|
pp = cp;
|
|
cp = strchr (pp, '&');
|
|
}
|
|
ret = g_string_append (ret, pp);
|
|
return g_string_free (ret, FALSE);
|
|
}
|
|
|
|
gchar *
|
|
parse_sami (ParserState * state, const gchar * line)
|
|
{
|
|
gchar *fixed_line;
|
|
GstSamiContext *context = (GstSamiContext *) state->user_data;
|
|
|
|
fixed_line = fix_invalid_entities (line);
|
|
htmlParseChunk (context->htmlctxt, fixed_line, strlen (fixed_line), 0);
|
|
g_free (fixed_line);
|
|
|
|
if (context->has_result) {
|
|
gchar *r;
|
|
|
|
if (context->rubybuf->len) {
|
|
context->rubybuf = g_string_append_c (context->rubybuf, '\n');
|
|
g_string_prepend (context->resultbuf, context->rubybuf->str);
|
|
context->rubybuf = g_string_truncate (context->rubybuf, 0);
|
|
}
|
|
|
|
r = g_string_free (context->resultbuf, FALSE);
|
|
context->resultbuf = g_string_new ("");
|
|
state->start_time = context->time1;
|
|
state->duration = context->time2 - context->time1;
|
|
context->has_result = FALSE;
|
|
return r;
|
|
}
|
|
return NULL;
|
|
}
|