Merge pull request #3353 from harfbuzz/buffer-create-similar

Add `hb_buffer_create_similar()`; use it in util

docs/harfbuzz-sections.txt

@@ -45,6 +45,7 @@ hb_memory_mode_t
 HB_SEGMENT_PROPERTIES_DEFAULT
 HB_BUFFER_REPLACEMENT_CODEPOINT_DEFAULT
 hb_buffer_create
+hb_buffer_create_similar
 hb_buffer_reference
 hb_buffer_get_empty
 hb_buffer_destroy
@@ -103,6 +104,7 @@ hb_buffer_serialize_format_to_string
 hb_buffer_serialize_list_formats
 hb_segment_properties_equal
 hb_segment_properties_hash
+hb_segment_properties_overlay
 hb_buffer_diff
 hb_buffer_set_message_func
 hb_buffer_t

src/hb-buffer.cc

@@ -86,7 +86,46 @@ hb_segment_properties_hash (const hb_segment_properties_t *p)
 	 (intptr_t) (p->language);
 }
 
+/**
+ * hb_segment_properties_overlay:
+ * @p: #hb_segment_properties_t to fill in.
+ * @src: #hb_segment_properties_t to fill in from.
+ *
+ * Fills in missing fields of @p from @src in a considered manner.
+ *
+ * First, if @p does not have direction set, direction is copied from @src.
+ *
+ * Next, if @p and @src have the same direction (which can be unset), if @p
+ * does not have script set, script is copied from @src.
+ *
+ * Finally, if @p and @src have the same direction and script (which either
+ * can be unset), if @p does not have language set, language is copied from
+ * @src.
+ *
+ * Since: REPLACEME
+ **/
+void
+hb_segment_properties_overlay (hb_segment_properties_t *p,
+			       const hb_segment_properties_t *src)
+{
+  if (unlikely (!p || !src))
+    return;
+
+  if (!p->direction)
+    p->direction = src->direction;
+
+  if (p->direction != src->direction)
+    return;
 
+  if (!p->script)
+    p->script = src->script;
+
+  if (p->script != src->script)
+    return;
+
+  if (!p->language)
+    p->language = src->language;
+}
 
 /* Here is how the buffer works internally:
  *
@@ -216,12 +255,25 @@ hb_buffer_t::get_scratch_buffer (unsigned int *size)
 
 /* HarfBuzz-Internal API */
 
+void
+hb_buffer_t::similar (const hb_buffer_t &src)
+{
+  hb_unicode_funcs_destroy (unicode);
+  unicode = hb_unicode_funcs_reference (src.unicode);
+  flags = src.flags;
+  cluster_level = src.cluster_level;
+  replacement = src.invisible;
+  invisible = src.invisible;
+  not_found = src.not_found;
+}
+
 void
 hb_buffer_t::reset ()
 {
   hb_unicode_funcs_destroy (unicode);
   unicode = hb_unicode_funcs_reference (hb_unicode_funcs_get_default ());
   flags = HB_BUFFER_FLAG_DEFAULT;
+  cluster_level = HB_BUFFER_CLUSTER_LEVEL_DEFAULT;
   replacement = HB_BUFFER_REPLACEMENT_CODEPOINT_DEFAULT;
   invisible = 0;
   not_found = 0;
@@ -232,11 +284,10 @@ hb_buffer_t::reset ()
 void
 hb_buffer_t::clear ()
 {
+  content_type = HB_BUFFER_CONTENT_TYPE_INVALID;
   hb_segment_properties_t default_props = HB_SEGMENT_PROPERTIES_DEFAULT;
   props = default_props;
-  scratch_flags = HB_BUFFER_SCRATCH_FLAG_DEFAULT;
 
-  content_type = HB_BUFFER_CONTENT_TYPE_INVALID;
   successful = true;
   have_output = false;
   have_positions = false;
@@ -244,16 +295,44 @@ hb_buffer_t::clear ()
   idx = 0;
   len = 0;
   out_len = 0;
-  out_info = info;
 
-  serial = 0;
+  out_info = info;
 
   memset (context, 0, sizeof context);
   memset (context_len, 0, sizeof context_len);
 
   deallocate_var_all ();
+  serial = 0;
+  scratch_flags = HB_BUFFER_SCRATCH_FLAG_DEFAULT;
 }
 
+void
+hb_buffer_t::enter ()
+{
+  deallocate_var_all ();
+  serial = 0;
+  scratch_flags = HB_BUFFER_SCRATCH_FLAG_DEFAULT;
+  if (likely (!hb_unsigned_mul_overflows (len, HB_BUFFER_MAX_LEN_FACTOR)))
+  {
+    max_len = hb_max (len * HB_BUFFER_MAX_LEN_FACTOR,
+		      (unsigned) HB_BUFFER_MAX_LEN_MIN);
+  }
+  if (likely (!hb_unsigned_mul_overflows (len, HB_BUFFER_MAX_OPS_FACTOR)))
+  {
+    max_ops = hb_max (len * HB_BUFFER_MAX_OPS_FACTOR,
+		      (unsigned) HB_BUFFER_MAX_OPS_MIN);
+  }
+}
+void
+hb_buffer_t::leave ()
+{
+  max_len = HB_BUFFER_MAX_LEN_DEFAULT;
+  max_ops = HB_BUFFER_MAX_OPS_DEFAULT;
+  deallocate_var_all ();
+  serial = 0;
+}
+
+
 void
 hb_buffer_t::add (hb_codepoint_t  codepoint,
 		  unsigned int    cluster)
@@ -565,12 +644,11 @@ DEFINE_NULL_INSTANCE (hb_buffer_t) =
   HB_BUFFER_REPLACEMENT_CODEPOINT_DEFAULT,
   0, /* invisible */
   0, /* not_found */
-  HB_BUFFER_SCRATCH_FLAG_DEFAULT,
-  HB_BUFFER_MAX_LEN_DEFAULT,
-  HB_BUFFER_MAX_OPS_DEFAULT,
+
 
   HB_BUFFER_CONTENT_TYPE_INVALID,
   HB_SEGMENT_PROPERTIES_DEFAULT,
+
   false, /* successful */
   false, /* have_output */
   true  /* have_positions */
@@ -609,6 +687,47 @@ hb_buffer_create ()
   return buffer;
 }
 
+/**
+ * hb_buffer_create_similar:
+ * @src: An #hb_buffer_t
+ *
+ * Resets the buffer to its initial status, as if it was just newly created
+ * with hb_buffer_create().
+ *
+ * Return value: (transfer full):
+ * A newly allocated #hb_buffer_t, similar to hb_buffer_create(). The only
+ * difference is that the buffer is configured similarly to @src.
+ *
+ * Since: REPLACEME
+ **/
+hb_buffer_t *
+hb_buffer_create_similar (const hb_buffer_t *src)
+{
+  hb_buffer_t *buffer = hb_buffer_create ();
+
+  buffer->similar (*src);
+
+  return buffer;
+}
+
+/**
+ * hb_buffer_reset:
+ * @buffer: An #hb_buffer_t
+ *
+ * Resets the buffer to its initial status, as if it was just newly created
+ * with hb_buffer_create().
+ *
+ * Since: 0.9.2
+ **/
+void
+hb_buffer_reset (hb_buffer_t *buffer)
+{
+  if (unlikely (hb_object_is_immutable (buffer)))
+    return;
+
+  buffer->reset ();
+}
+
 /**
  * hb_buffer_get_empty:
  *
@@ -1156,24 +1275,6 @@ hb_buffer_get_not_found_glyph (hb_buffer_t    *buffer)
 }
 
 
-/**
- * hb_buffer_reset:
- * @buffer: An #hb_buffer_t
- *
- * Resets the buffer to its initial status, as if it was just newly created
- * with hb_buffer_create().
- *
- * Since: 0.9.2
- **/
-void
-hb_buffer_reset (hb_buffer_t *buffer)
-{
-  if (unlikely (hb_object_is_immutable (buffer)))
-    return;
-
-  buffer->reset ();
-}
-
 /**
  * hb_buffer_clear_contents:
  * @buffer: An #hb_buffer_t
@@ -1749,6 +1850,8 @@ hb_buffer_append (hb_buffer_t *buffer,
   if (!buffer->have_positions && source->have_positions)
     buffer->clear_positions ();
 
+  hb_segment_properties_overlay (&buffer->props, &source->props);
+
   memcpy (buffer->info + orig_len, source->info + start, (end - start) * sizeof (buffer->info[0]));
   if (buffer->have_positions)
     memcpy (buffer->pos + orig_len, source->pos + start, (end - start) * sizeof (buffer->pos[0]));

src/hb-buffer.h

@@ -170,6 +170,9 @@ hb_segment_properties_equal (const hb_segment_properties_t *a,
 HB_EXTERN unsigned int
 hb_segment_properties_hash (const hb_segment_properties_t *p);
 
+HB_EXTERN void
+hb_segment_properties_overlay (hb_segment_properties_t *p,
+			       const hb_segment_properties_t *src);
 
 
 /**
@@ -184,6 +187,13 @@ typedef struct hb_buffer_t hb_buffer_t;
 HB_EXTERN hb_buffer_t *
 hb_buffer_create (void);
 
+HB_EXTERN hb_buffer_t *
+hb_buffer_create_similar (const hb_buffer_t *src);
+
+HB_EXTERN void
+hb_buffer_reset (hb_buffer_t *buffer);
+
+
 HB_EXTERN hb_buffer_t *
 hb_buffer_get_empty (void);
 
@@ -391,8 +401,9 @@ HB_EXTERN hb_codepoint_t
 hb_buffer_get_not_found_glyph (hb_buffer_t    *buffer);
 
 
-HB_EXTERN void
-hb_buffer_reset (hb_buffer_t *buffer);
+/*
+ * Content API.
+ */
 
 HB_EXTERN void
 hb_buffer_clear_contents (hb_buffer_t *buffer);

src/hb-buffer.hh

@@ -87,18 +87,21 @@ struct hb_buffer_t
 {
   hb_object_header_t header;
 
-  /* Information about how the text in the buffer should be treated */
+  /*
+   * Information about how the text in the buffer should be treated.
+   */
+
   hb_unicode_funcs_t *unicode; /* Unicode functions */
   hb_buffer_flags_t flags; /* BOT / EOT / etc. */
   hb_buffer_cluster_level_t cluster_level;
   hb_codepoint_t replacement; /* U+FFFD or something else. */
   hb_codepoint_t invisible; /* 0 or something else. */
   hb_codepoint_t not_found; /* 0 or something else. */
-  hb_buffer_scratch_flags_t scratch_flags; /* Have space-fallback, etc. */
-  unsigned int max_len; /* Maximum allowed len. */
-  int max_ops; /* Maximum allowed operations. */
 
-  /* Buffer contents */
+  /*
+   * Buffer contents
+   */
+
   hb_buffer_content_type_t content_type;
   hb_segment_properties_t props; /* Script, language, direction */
 
@@ -115,8 +118,6 @@ struct hb_buffer_t
   hb_glyph_info_t     *out_info;
   hb_glyph_position_t *pos;
 
-  unsigned int serial;
-
   /* Text before / after the main buffer contents.
    * Always in Unicode, and ordered outward.
    * Index 0 is for "pre-context", 1 for "post-context". */
@@ -124,7 +125,25 @@ struct hb_buffer_t
   hb_codepoint_t context[2][CONTEXT_LENGTH];
   unsigned int context_len[2];
 
-  /* Debugging API */
+
+  /*
+   * Managed by enter / leave
+   */
+
+#ifndef HB_NDEBUG
+  uint8_t allocated_var_bits;
+#endif
+  uint8_t serial;
+  hb_buffer_scratch_flags_t scratch_flags; /* Have space-fallback, etc. */
+  unsigned int max_len; /* Maximum allowed len. */
+  int max_ops; /* Maximum allowed operations. */
+  /* The bits here reflect current allocations of the bytes in glyph_info_t's var1 and var2. */
+
+
+  /*
+   * Messaging callback
+   */
+
 #ifndef HB_NO_BUFFER_MESSAGE
   hb_buffer_message_func_t message_func;
   void *message_data;
@@ -134,11 +153,6 @@ struct hb_buffer_t
   static constexpr unsigned message_depth = 0u;
 #endif
 
-  /* Internal debugging. */
-  /* The bits here reflect current allocations of the bytes in glyph_info_t's var1 and var2. */
-#ifndef HB_NDEBUG
-  uint8_t allocated_var_bits;
-#endif
 
 
   /* Methods */
@@ -190,12 +204,17 @@ struct hb_buffer_t
   hb_glyph_info_t &prev ()      { return out_info[out_len ? out_len - 1 : 0]; }
   hb_glyph_info_t prev () const { return out_info[out_len ? out_len - 1 : 0]; }
 
+  HB_INTERNAL void similar (const hb_buffer_t &src);
   HB_INTERNAL void reset ();
   HB_INTERNAL void clear ();
 
+  /* Called around shape() */
+  HB_INTERNAL void enter ();
+  HB_INTERNAL void leave ();
+
   unsigned int backtrack_len () const { return have_output ? out_len : idx; }
   unsigned int lookahead_len () const { return len - idx; }
-  unsigned int next_serial () { return serial++; }
+  uint8_t next_serial () { return ++serial ? serial : ++serial; }
 
   HB_INTERNAL void add (hb_codepoint_t  codepoint,
 			unsigned int    cluster);

src/hb-ot-layout.hh

@@ -482,10 +482,9 @@ _hb_glyph_info_get_lig_num_comps (const hb_glyph_info_t *info)
 }
 
 static inline uint8_t
-_hb_allocate_lig_id (hb_buffer_t *buffer) {
+_hb_allocate_lig_id (hb_buffer_t *buffer)
+{
   uint8_t lig_id = buffer->next_serial () & 0x07;
-  if (unlikely (!lig_id))
-    lig_id = _hb_allocate_lig_id (buffer); /* in case of overflow */
   return lig_id;
 }
 

src/hb-ot-shape.cc

@@ -1145,18 +1145,7 @@ hb_propagate_flags (hb_buffer_t *buffer)
 static void
 hb_ot_shape_internal (hb_ot_shape_context_t *c)
 {
-  c->buffer->deallocate_var_all ();
-  c->buffer->scratch_flags = HB_BUFFER_SCRATCH_FLAG_DEFAULT;
-  if (likely (!hb_unsigned_mul_overflows (c->buffer->len, HB_BUFFER_MAX_LEN_FACTOR)))
-  {
-    c->buffer->max_len = hb_max (c->buffer->len * HB_BUFFER_MAX_LEN_FACTOR,
-				 (unsigned) HB_BUFFER_MAX_LEN_MIN);
-  }
-  if (likely (!hb_unsigned_mul_overflows (c->buffer->len, HB_BUFFER_MAX_OPS_FACTOR)))
-  {
-    c->buffer->max_ops = hb_max (c->buffer->len * HB_BUFFER_MAX_OPS_FACTOR,
-				 (unsigned) HB_BUFFER_MAX_OPS_MIN);
-  }
+  c->buffer->enter ();
 
   /* Save the original direction, we use it later. */
   c->target_direction = c->buffer->props.direction;
@@ -1188,9 +1177,7 @@ hb_ot_shape_internal (hb_ot_shape_context_t *c)
 
   c->buffer->props.direction = c->target_direction;
 
-  c->buffer->max_len = HB_BUFFER_MAX_LEN_DEFAULT;
-  c->buffer->max_ops = HB_BUFFER_MAX_OPS_DEFAULT;
-  c->buffer->deallocate_var_all ();
+  c->buffer->leave ();
 }
 
 

util/shape-options.hh

@@ -60,15 +60,6 @@ struct shape_options_t
     hb_buffer_guess_segment_properties (buffer);
   }
 
-  static void copy_buffer_properties (hb_buffer_t *dst, hb_buffer_t *src)
-  {
-    hb_segment_properties_t props;
-    hb_buffer_get_segment_properties (src, &props);
-    hb_buffer_set_segment_properties (dst, &props);
-    hb_buffer_set_flags (dst, hb_buffer_get_flags (src));
-    hb_buffer_set_cluster_level (dst, hb_buffer_get_cluster_level (src));
-  }
-
   void populate_buffer (hb_buffer_t *buffer, const char *text, int text_len,
 			const char *text_before, const char *text_after)
   {
@@ -183,9 +174,8 @@ struct shape_options_t
 
     /* Check that breaking up shaping at safe-to-break is indeed safe. */
 
-    hb_buffer_t *fragment = hb_buffer_create ();
-    hb_buffer_t *reconstruction = hb_buffer_create ();
-    copy_buffer_properties (reconstruction, buffer);
+    hb_buffer_t *fragment = hb_buffer_create_similar (buffer);
+    hb_buffer_t *reconstruction = hb_buffer_create_similar (buffer);
 
     unsigned int num_glyphs;
     hb_glyph_info_t *info = hb_buffer_get_glyph_infos (buffer, &num_glyphs);
@@ -234,7 +224,6 @@ struct shape_options_t
 	printf("start %d end %d text start %d end %d\n", start, end, text_start, text_end);
 
       hb_buffer_clear_contents (fragment);
-      copy_buffer_properties (fragment, buffer);
 
       hb_buffer_flags_t flags = hb_buffer_get_flags (fragment);
       if (0 < text_start)