luc14n0/glib
1
0
Fork 0
mirror of https://gitlab.gnome.org/GNOME/glib.git synced 2025-03-15 20:25:12 +01:00
Code Issues Packages Projects Releases Wiki Activity

Optimizations; don't scan the entire text in find_current_text_end(),

Browse Source 
2004-11-28  Matthias Clasen  <mclasen@redhat.com>

	* glib/gmarkup.c: Optimizations; don't scan the entire text
	in find_current_text_end(), split unescape_text() into multiple
	functions.  (#159001, Havoc Pennington)
This commit is contained in:
Matthias Clasen 2004-11-28 05:40:10 +00:00 committed by Matthias Clasen
parent c77ae3942c
commit 86c78552ec
6 changed files with 410 additions and 297 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
ChangeLog
View File

@ -1,3 +1,9 @@
2004-11-28 Matthias Clasen <mclasen@redhat.com>
* glib/gmarkup.c: Optimizations; don't scan the entire text
in find_current_text_end(), split unescape_text() into multiple
functions. (#159001, Havoc Pennington)
2004-11-27 Matthias Clasen <mclasen@redhat.com>
* glib/gspawn.c (g_spawn_async_with_pipes): Update the @flags

6
ChangeLog.pre-2-10
View File

@ -1,3 +1,9 @@
2004-11-28 Matthias Clasen <mclasen@redhat.com>
* glib/gmarkup.c: Optimizations; don't scan the entire text
in find_current_text_end(), split unescape_text() into multiple
functions. (#159001, Havoc Pennington)
2004-11-27 Matthias Clasen <mclasen@redhat.com>
* glib/gspawn.c (g_spawn_async_with_pipes): Update the @flags

6
ChangeLog.pre-2-12
View File

@ -1,3 +1,9 @@
2004-11-28 Matthias Clasen <mclasen@redhat.com>
* glib/gmarkup.c: Optimizations; don't scan the entire text
in find_current_text_end(), split unescape_text() into multiple
functions. (#159001, Havoc Pennington)
2004-11-27 Matthias Clasen <mclasen@redhat.com>
* glib/gspawn.c (g_spawn_async_with_pipes): Update the @flags

6
ChangeLog.pre-2-6
View File

@ -1,3 +1,9 @@
2004-11-28 Matthias Clasen <mclasen@redhat.com>
* glib/gmarkup.c: Optimizations; don't scan the entire text
in find_current_text_end(), split unescape_text() into multiple
functions. (#159001, Havoc Pennington)
2004-11-27 Matthias Clasen <mclasen@redhat.com>
* glib/gspawn.c (g_spawn_async_with_pipes): Update the @flags

6
ChangeLog.pre-2-8
View File

@ -1,3 +1,9 @@
2004-11-28 Matthias Clasen <mclasen@redhat.com>
* glib/gmarkup.c: Optimizations; don't scan the entire text
in find_current_text_end(), split unescape_text() into multiple
functions. (#159001, Havoc Pennington)
2004-11-27 Matthias Clasen <mclasen@redhat.com>
* glib/gspawn.c (g_spawn_async_with_pipes): Update the @flags

311
glib/gmarkup.c
View File

@ -235,12 +235,24 @@ set_error (GMarkupParseContext *context,
g_propagate_error (error, tmp_error);
}
/* To make these faster, we first use the ascii-only tests, then check
* for the usual non-alnum name-end chars, and only then call the
* expensive unicode stuff. Nobody uses non-ascii in XML tag/attribute
* names, so this is a reasonable hack that virtually always avoids
* the guniprop call.
*/
#define IS_COMMON_NAME_END_CHAR(c) \
((c) == '=' || (c) == '/' || (c) == '>' || (c) == ' ')
static gboolean
is_name_start_char (gunichar c)
{
if (g_unichar_isalpha (c) ||
if (g_ascii_isalpha (c) ||
(!IS_COMMON_NAME_END_CHAR (c) &&
(g_unichar_isalpha (c) ||
c == '_' ||
c == ':')
c == ':')))
return TRUE;
else
return FALSE;
@ -249,11 +261,13 @@ is_name_start_char (gunichar c)
static gboolean
is_name_char (gunichar c)
{
if (g_unichar_isalnum (c) ||
if (g_ascii_isalnum (c) ||
(!IS_COMMON_NAME_END_CHAR (c) &&
(g_unichar_isalnum (c) ||
c == '.' ||
c == '-' ||
c == '_' ||
c == ':')
c == ':')))
return TRUE;
else
return FALSE;
@ -326,54 +340,51 @@ typedef enum
USTATE_AFTER_CHARREF_HASH
} UnescapeState;
static gboolean
unescape_text (GMarkupParseContext *context,
const gchar *text,
const gchar *text_end,
gchar **unescaped,
typedef struct
{
GMarkupParseContext *context;
GString *str;
UnescapeState state;
const gchar *text;
const gchar *text_end;
const gchar *entity_start;
} UnescapeContext;
static const gchar*
unescape_text_state_inside_text (UnescapeContext *ucontext,
const gchar *p,
GError **error)
{
#define MAX_ENT_LEN 5
GString *str;
const gchar *p;
UnescapeState state;
const gchar *start;
gboolean normalize_attribute;
str = g_string_new (NULL);
if (context->state == STATE_INSIDE_ATTRIBUTE_VALUE_SQ ||
context->state == STATE_INSIDE_ATTRIBUTE_VALUE_DQ)
if (ucontext->context->state == STATE_INSIDE_ATTRIBUTE_VALUE_SQ ||
ucontext->context->state == STATE_INSIDE_ATTRIBUTE_VALUE_DQ)
normalize_attribute = TRUE;
else
normalize_attribute = FALSE;
state = USTATE_INSIDE_TEXT;
p = text;
start = p;
while (p != text_end && context->state != STATE_ERROR)
{
g_assert (p < text_end);
switch (state)
while (p != ucontext->text_end)
{
case USTATE_INSIDE_TEXT:
if (*p == '&')
{
while (p != text_end && *p != '&')
break;
}
else if (normalize_attribute && (*p == '\t' || *p == '\n'))
{
if ((*p == '\t' || *p == '\n') && normalize_attribute)
{
g_string_append_len (str, start, p - start);
g_string_append_c (str, ' ');
g_string_append_len (ucontext->str, start, p - start);
g_string_append_c (ucontext->str, ' ');
p = g_utf8_next_char (p);
start = p;
}
else if (*p == '\r')
{
g_string_append_len (str, start, p - start);
g_string_append_c (str, normalize_attribute ? ' ' : '\n');
g_string_append_len (ucontext->str, start, p - start);
g_string_append_c (ucontext->str, normalize_attribute ? ' ' : '\n');
p = g_utf8_next_char (p);
if (*p == '\n')
if (p != ucontext->text_end && *p == '\n')
p = g_utf8_next_char (p);
start = p;
}
@ -382,35 +393,38 @@ unescape_text (GMarkupParseContext *context,
}
if (p != start)
{
g_string_append_len (str, start, p - start);
g_string_append_len (ucontext->str, start, p - start);
start = NULL;
}
if (p != text_end && *p == '&')
if (p != ucontext->text_end && *p == '&')
{
p = g_utf8_next_char (p);
state = USTATE_AFTER_AMPERSAND;
ucontext->state = USTATE_AFTER_AMPERSAND;
}
}
break;
case USTATE_AFTER_AMPERSAND:
return p;
}
__attribute__ ((noinline))
static const gchar*
unescape_text_state_after_ampersand (UnescapeContext *ucontext,
const gchar *p,
GError **error)
{
ucontext->entity_start = NULL;
if (*p == '#')
{
p = g_utf8_next_char (p);
start = p;
state = USTATE_AFTER_CHARREF_HASH;
ucontext->entity_start = p;
ucontext->state = USTATE_AFTER_CHARREF_HASH;
}
else if (!is_name_start_char (g_utf8_get_char (p)))
{
if (*p == ';')
{
set_unescape_error (context, error,
p, text_end,
set_unescape_error (ucontext->context, error,
p, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Empty entity '&;' seen; valid "
"entities are: &amp; &quot; &lt; &gt; &apos;"));
@ -419,8 +433,8 @@ unescape_text (GMarkupParseContext *context,
{
gchar buf[7];
set_unescape_error (context, error,
p, text_end,
set_unescape_error (ucontext->context, error,
p, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Character '%s' is not valid at "
"the start of an entity name; "
@ -433,21 +447,26 @@ unescape_text (GMarkupParseContext *context,
}
else
{
start = p;
state = USTATE_INSIDE_ENTITY_NAME;
ucontext->entity_start = p;
ucontext->state = USTATE_INSIDE_ENTITY_NAME;
}
return p;
}
break;
case USTATE_INSIDE_ENTITY_NAME:
__attribute__ ((noinline))
static const gchar*
unescape_text_state_inside_entity_name (UnescapeContext *ucontext,
const gchar *p,
GError **error)
{
#define MAX_ENT_LEN 5
gchar buf[MAX_ENT_LEN+1] = {
'\0', '\0', '\0', '\0', '\0', '\0'
};
gchar *dest;
while (p != text_end)
while (p != ucontext->text_end)
{
if (*p == ';')
break;
@ -455,8 +474,8 @@ unescape_text (GMarkupParseContext *context,
{
gchar ubuf[7];
set_unescape_error (context, error,
p, text_end,
set_unescape_error (ucontext->context, error,
p, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Character '%s' is not valid "
"inside an entity name"),
@ -467,13 +486,13 @@ unescape_text (GMarkupParseContext *context,
p = g_utf8_next_char (p);
}
if (context->state != STATE_ERROR)
if (ucontext->context->state != STATE_ERROR)
{
if (p != text_end)
if (p != ucontext->text_end)
{
const gchar *src;
src = start;
src = ucontext->entity_start;
dest = buf;
while (src != p)
{
@ -484,23 +503,22 @@ unescape_text (GMarkupParseContext *context,
/* move to after semicolon */
p = g_utf8_next_char (p);
start = p;
state = USTATE_INSIDE_TEXT;
ucontext->state = USTATE_INSIDE_TEXT;
if (strcmp (buf, "lt") == 0)
g_string_append_c (str, '<');
g_string_append_c (ucontext->str, '<');
else if (strcmp (buf, "gt") == 0)
g_string_append_c (str, '>');
g_string_append_c (ucontext->str, '>');
else if (strcmp (buf, "amp") == 0)
g_string_append_c (str, '&');
g_string_append_c (ucontext->str, '&');
else if (strcmp (buf, "quot") == 0)
g_string_append_c (str, '"');
g_string_append_c (ucontext->str, '"');
else if (strcmp (buf, "apos") == 0)
g_string_append_c (str, '\'');
g_string_append_c (ucontext->str, '\'');
else
{
set_unescape_error (context, error,
p, text_end,
set_unescape_error (ucontext->context, error,
p, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Entity name '%s' is not known"),
buf);
@ -508,9 +526,9 @@ unescape_text (GMarkupParseContext *context,
}
else
{
set_unescape_error (context, error,
set_unescape_error (ucontext->context, error,
/* give line number of the & */
start, text_end,
ucontext->entity_start, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Entity did not end with a semicolon; "
"most likely you used an ampersand "
@ -518,12 +536,22 @@ unescape_text (GMarkupParseContext *context,
"an entity - escape ampersand as &amp;"));
}
}
}
break;
#undef MAX_ENT_LEN
case USTATE_AFTER_CHARREF_HASH:
return p;
}
__attribute__ ((noinline))
static const gchar*
unescape_text_state_after_charref_hash (UnescapeContext *ucontext,
const gchar *p,
GError **error)
{
gboolean is_hex = FALSE;
const char *start;
start = ucontext->entity_start;
if (*p == 'x')
{
is_hex = TRUE;
@ -531,10 +559,10 @@ unescape_text (GMarkupParseContext *context,
start = p;
}
while (p != text_end && *p != ';')
while (p != ucontext->text_end && *p != ';')
p = g_utf8_next_char (p);
if (p != text_end)
if (p != ucontext->text_end)
{
g_assert (*p == ';');
@ -555,8 +583,8 @@ unescape_text (GMarkupParseContext *context,
if (end != digit_end || errno != 0)
{
set_unescape_error (context, error,
start, text_end,
set_unescape_error (ucontext->context, error,
start, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Failed to parse '%s', which "
"should have been a digit "
@ -576,12 +604,12 @@ unescape_text (GMarkupParseContext *context,
(l >= 0x10000 && l <= 0x10FFFF))
{
gchar buf[7];
g_string_append (str, char_str (l, buf));
g_string_append (ucontext->str, char_str (l, buf));
}
else
{
set_unescape_error (context, error,
start, text_end,
set_unescape_error (ucontext->context, error,
start, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Character reference '%s' does not encode a permitted character"),
digit);
@ -592,13 +620,12 @@ unescape_text (GMarkupParseContext *context,
/* Move to next state */
p = g_utf8_next_char (p); /* past semicolon */
start = p;
state = USTATE_INSIDE_TEXT;
ucontext->state = USTATE_INSIDE_TEXT;
}
else
{
set_unescape_error (context, error,
start, text_end,
set_unescape_error (ucontext->context, error,
start, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Empty character reference; "
"should include a digit such as "
@ -607,8 +634,8 @@ unescape_text (GMarkupParseContext *context,
}
else
{
set_unescape_error (context, error,
start, text_end,
set_unescape_error (ucontext->context, error,
start, ucontext->text_end,
G_MARKUP_ERROR_PARSE,
_("Character reference did not end with a "
"semicolon; "
@ -616,6 +643,66 @@ unescape_text (GMarkupParseContext *context,
"character without intending to start "
"an entity - escape ampersand as &amp;"));
}
return p;
}
static gboolean
unescape_text (GMarkupParseContext *context,
const gchar *text,
const gchar *text_end,
GString **unescaped,
GError **error)
{
UnescapeContext ucontext;
const gchar *p;
ucontext.context = context;
ucontext.text = text;
ucontext.text_end = text_end;
ucontext.entity_start = NULL;
ucontext.str = g_string_sized_new (text_end - text);
ucontext.state = USTATE_INSIDE_TEXT;
p = text;
while (p != text_end && context->state != STATE_ERROR)
{
g_assert (p < text_end);
switch (ucontext.state)
{
case USTATE_INSIDE_TEXT:
{
p = unescape_text_state_inside_text (&ucontext,
p,
error);
}
break;
case USTATE_AFTER_AMPERSAND:
{
p = unescape_text_state_after_ampersand (&ucontext,
p,
error);
}
break;
case USTATE_INSIDE_ENTITY_NAME:
{
p = unescape_text_state_inside_entity_name (&ucontext,
p,
error);
}
break;
case USTATE_AFTER_CHARREF_HASH:
{
p = unescape_text_state_after_charref_hash (&ucontext,
p,
error);
}
break;
@ -627,7 +714,7 @@ unescape_text (GMarkupParseContext *context,
if (context->state != STATE_ERROR)
{
switch (state)
switch (ucontext.state)
{
case USTATE_INSIDE_TEXT:
break;
@ -649,31 +736,28 @@ unescape_text (GMarkupParseContext *context,
if (context->state == STATE_ERROR)
{
g_string_free (str, TRUE);
g_string_free (ucontext.str, TRUE);
*unescaped = NULL;
return FALSE;
}
else
{
*unescaped = g_string_free (str, FALSE);
*unescaped = ucontext.str;
return TRUE;
}
#undef MAX_ENT_LEN
}
static gboolean
static inline gboolean
advance_char (GMarkupParseContext *context)
{
g_return_val_if_fail (context->iter != context->current_text_end, FALSE);
context->iter = g_utf8_next_char (context->iter);
context->char_number += 1;
if (context->iter == context->current_text_end)
{
return FALSE;
if (*context->iter == '\n')
}
else if (*context->iter == '\n')
{
context->line_number += 1;
context->char_number = 1;
@ -682,7 +766,7 @@ advance_char (GMarkupParseContext *context)
return TRUE;
}
static gboolean
static inline gboolean
xml_isspace (char c)
{
return c == ' ' || c == '\t' || c == '\n' || c == '\r';
@ -716,7 +800,7 @@ add_to_partial (GMarkupParseContext *context,
const gchar *text_end)
{
if (context->partial_chunk == NULL)
context->partial_chunk = g_string_new (NULL);
context->partial_chunk = g_string_sized_new (text_end - text_start);
if (text_start != text_end)
g_string_append_len (context->partial_chunk, text_start,
@ -750,23 +834,18 @@ current_attribute (GMarkupParseContext *context)
static void
find_current_text_end (GMarkupParseContext *context)
{
/* This function must be safe (non-segfaulting) on invalid UTF8 */
/* This function must be safe (non-segfaulting) on invalid UTF8.
* It assumes the string starts with a character start
*/
const gchar *end = context->current_text + context->current_text_len;
const gchar *p;
const gchar *next;
g_assert (context->current_text_len > 0);
p = context->current_text;
next = g_utf8_find_next_char (p, end);
p = g_utf8_find_prev_char (context->current_text, end);
while (next && *next)
{
if (p == next)
next++;
p = next;
next = g_utf8_find_next_char (p, end);
}
g_assert (p != NULL); /* since current_text was a char start */
/* p is now the start of the last character or character portion. */
g_assert (p != end);
@ -1352,6 +1431,8 @@ g_markup_parse_context_parse (GMarkupParseContext *context,
* with the partial chunk if any; set it for the current
* attribute.
*/
GString *unescaped;
add_to_partial (context, context->start, context->iter);
g_assert (context->cur_attr >= 0);
@ -1360,10 +1441,11 @@ g_markup_parse_context_parse (GMarkupParseContext *context,
context->partial_chunk->str,
context->partial_chunk->str +
context->partial_chunk->len,
&context->attr_values[context->cur_attr],
&unescaped,
error))
{
/* success, advance past quote and set state. */
context->attr_values[context->cur_attr] = g_string_free (unescaped, FALSE);
advance_char (context);
context->state = STATE_BETWEEN_ATTRIBUTES;
context->start = NULL;
@ -1390,7 +1472,7 @@ g_markup_parse_context_parse (GMarkupParseContext *context,
if (context->iter != context->current_text_end)
{
gchar *unescaped = NULL;
GString *unescaped = NULL;
/* The text has ended at the open angle. Call the text
* callback.
@ -1407,12 +1489,12 @@ g_markup_parse_context_parse (GMarkupParseContext *context,
if (context->parser->text)
(*context->parser->text) (context,
unescaped,
strlen (unescaped),
unescaped->str,
unescaped->len,
context->user_data,
&tmp_error);
g_free (unescaped);
g_string_free (unescaped, TRUE);
if (tmp_error == NULL)
{
@ -1869,7 +1951,8 @@ g_markup_escape_text (const gchar *text,
if (length < 0)
length = strlen (text);
str = g_string_new (NULL);
/* prealloc at least as long as original text */
str = g_string_sized_new (length);
append_escaped_text (str, text, length);
return g_string_free (str, FALSE);