Only remaining undefined symbol is upb_msg_parsestr.

Makefile

@@ -47,7 +47,7 @@ clean:
 
 # The core library (src/libupb.a)
 SRC=src/upb.c src/upb_parse.c src/upb_table.c src/upb_def.c src/upb_data.c \
-    descriptor/descriptor.c
+    descriptor/descriptor.c src/upb_text.c
 # Override the optimization level for upb_def.o, because it is not in the
 # critical path but gets very large when -O3 is used.
 src/upb_def.o: src/upb_def.c

src/upb_data.c

@@ -8,10 +8,6 @@
 #include "upb_data.h"
 #include "upb_def.h"
 
-INLINE void data_init(upb_data *d, int flags) {
-  d->v = flags;
-}
-
 static uint32_t round_up_to_pow2(uint32_t v)
 {
   /* cf. http://graphics.stanford.edu/~seander/bithacks.html#RoundUpPowerOf2 */
@@ -25,6 +21,28 @@ static uint32_t round_up_to_pow2(uint32_t v)
   return v;
 }
 
+/* upb_data *******************************************************************/
+
+static void data_elem_unref(void *d, struct upb_fielddef *f) {
+  if(f->type == UPB_TYPE(MESSAGE) || f->type == UPB_TYPE(GROUP)) {
+    upb_msg_unref((upb_msg*)d, upb_downcast_msgdef(f->def));
+  } else if(f->type == UPB_TYPE(STRING) || f->type == UPB_TYPE(BYTES)) {
+    upb_string_unref((upb_string*)d);
+  }
+}
+
+static void data_unref(void *d, struct upb_fielddef *f) {
+  if(upb_isarray(f)) {
+    upb_array_unref((upb_array*)d, f);
+  } else {
+    data_elem_unref(d, f);
+  }
+}
+
+INLINE void data_init(upb_data *d, int flags) {
+  d->v = flags;
+}
+
 static void check_not_frozen(upb_data *d) {
   // On one hand I am reluctant to put abort() calls in a low-level library
   // that are enabled in a production build.  On the other hand, this is a bug
@@ -50,6 +68,9 @@ static void string_set_bytesize(upb_string *s, upb_strlen_t newsize) {
   }
 }
 
+
+/* upb_string *******************************************************************/
+
 upb_string *upb_string_new() {
   upb_string *s = malloc(sizeof(upb_refcounted_string));
   data_init(&s->common.base, UPB_DATA_HEAPALLOCATED | UPB_DATA_REFCOUNTED);
@@ -76,6 +97,43 @@ void upb_string_resize(upb_string *s, upb_strlen_t byte_len) {
   s->common.byte_len = byte_len;
 }
 
+upb_string *upb_strreadfile(const char *filename) {
+  FILE *f = fopen(filename, "rb");
+  if(!f) return false;
+  if(fseek(f, 0, SEEK_END) != 0) goto error;
+  long size = ftell(f);
+  if(size < 0) goto error;
+  if(fseek(f, 0, SEEK_SET) != 0) goto error;
+  upb_string *s = upb_string_new();
+  char *buf = upb_string_getrwbuf(s, size);
+  if(fread(buf, size, 1, f) != 1) goto error;
+  fclose(f);
+  return s;
+
+error:
+  fclose(f);
+  return NULL;
+}
+
+
+/* upb_array ******************************************************************/
+
+// ONLY handles refcounted arrays for the moment.
+void _upb_array_free(upb_array *a, struct upb_fielddef *f)
+{
+  if(upb_elem_ismm(f)) {
+    for(upb_arraylen_t i = 0; i < a->refcounted.size; i++) {
+      union upb_value_ptr p = _upb_array_getptr(a, f, i);
+      data_elem_unref(p._void, f);
+    }
+  }
+  if(a->refcounted.size != 0) free(a->common.elements._void);
+  free(a);
+}
+
+
+/* upb_msg ********************************************************************/
+
 upb_msg *upb_msg_new(struct upb_msgdef *md) {
   upb_msg *msg = malloc(md->size);
   memset(msg, 0, md->size);
@@ -83,28 +141,16 @@ upb_msg *upb_msg_new(struct upb_msgdef *md) {
   return msg;
 }
 
-#if 0
-void upb_msg_destroy(struct upb_msg *msg) {
-  for(upb_field_count_t i = 0; i < msg->def->num_fields; i++) {
-    struct upb_fielddef *f = &msg->def->fields[i];
-    if(!upb_msg_isset(msg, f) || !upb_field_ismm(f)) continue;
-    upb_mm_destroy(upb_msg_getptr(msg, f), upb_field_ptrtype(f));
+// ONLY handles refcounted messages for the moment.
+void _upb_msg_free(upb_msg *msg, struct upb_msgdef *md)
+{
+  for(int i = 0; i < md->num_fields; i++) {
+    struct upb_fielddef *f = &md->fields[i];
+    union upb_value_ptr p = _upb_msg_getptr(msg, f);
+    if(!upb_field_ismm(f) || !p._void) continue;
+    data_unref(p._void, f);
   }
-  upb_def_unref(UPB_UPCAST(msg->def));
+  upb_def_unref(UPB_UPCAST(md));
   free(msg);
 }
 
-void upb_array_destroy(struct upb_array *arr)
-{
-  if(upb_elem_ismm(arr->fielddef)) {
-    upb_arraylen_t i;
-    /* Unref elements. */
-    for(i = 0; i < arr->size; i++) {
-      union upb_value_ptr p = upb_array_getelementptr(arr, i);
-      upb_mm_destroy(p, upb_elem_ptrtype(arr->fielddef));
-    }
-  }
-  if(arr->size != 0) free(arr->elements._void);
-  free(arr);
-}
-#endif

src/upb_data.h

@@ -378,12 +378,29 @@ typedef struct type ## _array { \
 // empty.  Caller owns one ref on it.
 upb_array *upb_array_new(void);
 
-INLINE union upb_value upb_array_get(upb_array *a, struct upb_fielddef *f,
-                                     int elem) {
+// INTERNAL-ONLY:
+// Frees the given message and releases references on members.
+void _upb_array_free(upb_array *a, struct upb_fielddef *f);
+
+// INTERNAL-ONLY:
+// Returns a pointer to the given elem.
+INLINE union upb_value_ptr _upb_array_getptr(upb_array *a,
+                                             struct upb_fielddef *f, int elem) {
   assert(elem < upb_array_len(a));
   size_t type_size = upb_type_info[f->type].size;
   union upb_value_ptr p = {._void = &a->common.elements.uint8[elem * type_size]};
-  return upb_value_read(p, f->type);
+  return p;
+}
+
+INLINE union upb_value upb_array_get(upb_array *a, struct upb_fielddef *f,
+                                     int elem) {
+  return upb_value_read(_upb_array_getptr(a, f, elem), f->type);
+}
+
+// The caller releases a ref on the given array, which it must previously have
+// owned a ref on.
+INLINE void upb_array_unref(upb_array *a, struct upb_fielddef *f) {
+  if(_upb_data_unref(&a->common.base)) _upb_array_free(a, f);
 }
 
 #if 0
@@ -392,10 +409,6 @@ INLINE union upb_value upb_array_get(upb_array *a, struct upb_fielddef *f,
 // were incompatible with src's.
 INLINE upb_array *upb_array_getref(upb_array *src, int ref_flags);
 
-// The caller releases a ref on the given array, which it must previously have
-// owned a ref on.
-INLINE void upb_array_unref(upb_array *a, struct upb_fielddef *f);
-
 // Sets the given element in the array to val.  The current length of the array
 // must be greater than elem.  If the field type is dynamic, the array will
 // take a ref on val and release a ref on what was previously in the array.
@@ -429,20 +442,33 @@ struct upb_msg {
 // Creates a new msg of the given type.
 upb_msg *upb_msg_new(struct upb_msgdef *md);
 
+// INTERNAL-ONLY:
+// Frees the given message and releases references on members.
+void _upb_msg_free(upb_msg *msg, struct upb_msgdef *md);
+
+// INTERNAL-ONLY:
+// Returns a pointer to the given field.
+INLINE union upb_value_ptr _upb_msg_getptr(upb_msg *msg, struct upb_fielddef *f) {
+  union upb_value_ptr p = {._void = &msg->data[f->byte_offset]};
+  return p;
+}
+
+// Releases a references on msg.
+INLINE void upb_msg_unref(upb_msg *msg, struct upb_msgdef *md) {
+  if(_upb_data_unref(&msg->base)) _upb_msg_free(msg, md);
+}
+
 // Tests whether the given field is explicitly set, or whether it will return
 // a default.
 INLINE bool upb_msg_has(upb_msg *msg, struct upb_fielddef *f) {
   return msg->data[f->field_index/8] % (1 << (f->field_index % 8));
 }
 
-void upb_msg_unref(upb_msg *msg, struct upb_msgdef *md);
-
 // Returns the current value if set, or the default value if not set, of the
 // specified field.  The caller does *not* own a ref.
 INLINE union upb_value upb_msg_get(upb_msg *msg, struct upb_fielddef *f) {
   if(upb_msg_has(msg, f)) {
-    union upb_value_ptr p = {._void = &msg->data[f->byte_offset]};
-    return upb_value_read(p, f->type);
+    return upb_value_read(_upb_msg_getptr(msg, f), f->type);
   } else {
     return f->default_value;
   }

src/upb_text.c

@@ -7,9 +7,7 @@
 #include <inttypes.h>
 #include "descriptor.h"
 #include "upb_text.h"
-#include "upb_string.h"
-#include "upb_msg.h"
-#include "upb_array.h"
+#include "upb_data.h"
 
 void upb_text_printval(upb_field_type_t type, union upb_value val, FILE *file)
 {
@@ -51,7 +49,7 @@ static void print_indent(struct upb_text_printer *p, FILE *stream)
 }
 
 void upb_text_printfield(struct upb_text_printer *p,
-                         struct upb_string *name,
+                         upb_string *name,
                          upb_field_type_t valtype, union upb_value val,
                          FILE *stream)
 {
@@ -65,7 +63,7 @@ void upb_text_printfield(struct upb_text_printer *p,
 }
 
 void upb_text_push(struct upb_text_printer *p,
-                   struct upb_string *submsg_type,
+                   upb_string *submsg_type,
                    FILE *stream)
 {
   print_indent(p, stream);
@@ -82,49 +80,48 @@ void upb_text_pop(struct upb_text_printer *p,
   fprintf(stream, "}\n");
 }
 
-static void printval(struct upb_text_printer *printer, union upb_value_ptr p,
+static void printval(struct upb_text_printer *printer, union upb_value v,
                      struct upb_fielddef *f,
                      FILE *stream);
 
-static void printmsg(struct upb_text_printer *printer, struct upb_msg *msg,
-                     FILE *stream)
+static void printmsg(struct upb_text_printer *printer,
+                     upb_msg *msg, struct upb_msgdef *md, FILE *stream)
 {
-  struct upb_msgdef *m = msg->def;
-  for(upb_field_count_t i = 0; i < m->num_fields; i++) {
-    struct upb_fielddef *f = &m->fields[i];
-    if(!upb_msg_isset(msg, f)) continue;
-    union upb_value_ptr p = upb_msg_getptr(msg, f);
+  for(upb_field_count_t i = 0; i < md->num_fields; i++) {
+    struct upb_fielddef *f = &md->fields[i];
+    if(!upb_msg_has(msg, f)) continue;
+    union upb_value v = upb_msg_get(msg, f);
     if(upb_isarray(f)) {
-      struct upb_array *arr = *p.arr;
-      for(uint32_t j = 0; j < arr->len; j++) {
-        union upb_value_ptr elem_p = upb_array_getelementptr(arr, j);
-        printval(printer, elem_p, f, stream);
+      upb_array *arr = v.arr;
+      for(uint32_t j = 0; j < upb_array_len(arr); j++) {
+        union upb_value elem = upb_array_get(arr, f, j);
+        printval(printer, elem, f, stream);
       }
     } else {
-      printval(printer, p, f, stream);
+      printval(printer, v, f, stream);
     }
   }
 }
 
-static void printval(struct upb_text_printer *printer, union upb_value_ptr p,
+static void printval(struct upb_text_printer *printer, union upb_value v,
                      struct upb_fielddef *f,
                      FILE *stream)
 {
   if(upb_issubmsg(f)) {
     upb_text_push(printer, f->name, stream);
-    printmsg(printer, *p.msg, stream);
+    printmsg(printer, v.msg, upb_downcast_msgdef(f->def), stream);
     upb_text_pop(printer, stream);
   } else {
-    upb_text_printfield(printer, f->name, f->type,
-                        upb_value_read(p, f->type), stream);
+    upb_text_printfield(printer, f->name, f->type, v, stream);
   }
 }
 
 
-void upb_msg_print(struct upb_msg *msg, bool single_line, FILE *stream)
+void upb_msg_print(upb_msg *msg, struct upb_msgdef *md, bool single_line,
+                   FILE *stream)
 {
   struct upb_text_printer printer;
   upb_text_printer_init(&printer, single_line);
-  printmsg(&printer, msg, stream);
+  printmsg(&printer, msg, md, stream);
 }