From 4dce5ab709f2fcad836a562f3902f2654fc584c4 Mon Sep 17 00:00:00 2001 From: Joshua Haberman Date: Sun, 13 Feb 2011 17:27:32 -0800 Subject: Fix upbc and descriptorgen, and update descriptor. --- tools/upbc.c | 627 ++++++----------------------------------------------------- 1 file changed, 57 insertions(+), 570 deletions(-) (limited to 'tools/upbc.c') diff --git a/tools/upbc.c b/tools/upbc.c index 707fa04..a636301 100644 --- a/tools/upbc.c +++ b/tools/upbc.c @@ -13,30 +13,32 @@ #include #include #include "descriptor.h" -#include "upb_data.h" #include "upb_def.h" -#include "upb_text.h" +#include "upb_msg.h" +#include "upb_glue.h" +#include "upb_strstream.h" +#include "upb_decoder.h" /* These are in-place string transformations that do not change the length of * the string (and thus never need to re-allocate). */ // Convert to C identifier: foo.bar.Baz -> foo_bar_Baz. -static void to_cident(upb_strptr str) +static void to_cident(upb_string *str) { - upb_strlen_t len = upb_strlen(str); + upb_strlen_t len = upb_string_len(str); char *buf = upb_string_getrwbuf(str, len); - for(uint32_t i = 0; i < len; i++) + for(int32_t i = 0; i < len; i++) if(buf[i] == '.' || buf[i] == '/') buf[i] = '_'; } // Convert to C proprocessor identifier: foo.bar.Baz -> FOO_BAR_BAZ. -static void to_preproc(upb_strptr str) +static void to_preproc(upb_string *str) { to_cident(str); - upb_strlen_t len = upb_strlen(str); + upb_strlen_t len = upb_string_len(str); char *buf = upb_string_getrwbuf(str, len); - for(uint32_t i = 0; i < len; i++) + for(int32_t i = 0; i < len; i++) buf[i] = toupper(buf[i]); } @@ -47,28 +49,16 @@ static int my_memrchr(const char *data, char c, size_t len) return off; } -void *strtable_to_array(upb_strtable *t, int *size) -{ - *size = t->t.count; - void **array = malloc(*size * sizeof(void*)); - upb_strtable_entry *e; - int i = 0; - for(e = upb_strtable_begin(t); e && i < *size; e = upb_strtable_next(t, e)) - array[i++] = e; - assert(i == *size && e == NULL); - return array; -} - /* The _const.h file defines the constants (enums) defined in the .proto * file. */ static void write_const_h(upb_def *defs[], int num_entries, char *outfile_name, FILE *stream) { /* Header file prologue. */ - upb_strptr include_guard_name = upb_strdupc(outfile_name); + upb_string *include_guard_name = upb_strdupc(outfile_name); to_preproc(include_guard_name); /* A bit cheesy, but will do the job. */ - upb_strlen_t len = upb_strlen(include_guard_name); + upb_strlen_t len = upb_string_len(include_guard_name); char *buf = upb_string_getrwbuf(include_guard_name, len); buf[len-1] = 'C'; @@ -85,28 +75,29 @@ static void write_const_h(upb_def *defs[], int num_entries, char *outfile_name, for(int i = 0; i < num_entries; i++) { /* Foreach enum */ if(defs[i]->type != UPB_DEF_ENUM) continue; upb_enumdef *enumdef = upb_downcast_enumdef(defs[i]); - upb_strptr enum_name = upb_strdup(UPB_UPCAST(enumdef)->fqname); - upb_strptr enum_val_prefix = upb_strdup(enum_name); + upb_string *enum_name = upb_strdup(UPB_UPCAST(enumdef)->fqname); + upb_string *enum_val_prefix = upb_strdup(enum_name); to_cident(enum_name); const char *data = upb_string_getrobuf(enum_val_prefix); - upb_strlen_t len = upb_strlen(enum_val_prefix); + upb_strlen_t len = upb_string_len(enum_val_prefix); upb_strlen_t lastsep = my_memrchr(data, UPB_SYMBOL_SEPARATOR, len); - upb_string_resize(enum_val_prefix, lastsep + 1); + upb_string_getrwbuf(enum_val_prefix, lastsep + 1); to_preproc(enum_val_prefix); fprintf(stream, "typedef enum " UPB_STRFMT " {\n", UPB_STRARG(enum_name)); - upb_enum_iter iter; bool first = true; /* Foreach enum value. */ - for(upb_enum_begin(&iter, enumdef); !upb_enum_done(&iter); upb_enum_next(&iter)) { - upb_strptr value_name = upb_strdup(iter.name); + for (upb_enum_iter iter = upb_enum_begin(enumdef); + !upb_enum_done(iter); + iter = upb_enum_next(enumdef, iter)) { + upb_string *value_name = upb_strdup(upb_enum_iter_name(iter)); to_preproc(value_name); /* " GOOGLE_PROTOBUF_FIELDDESCRIPTORPROTO_TYPE_UINT32 = 13," */ if (!first) fputs(",\n", stream); first = false; fprintf(stream, " " UPB_STRFMT UPB_STRFMT " = %" PRIu32, - UPB_STRARG(enum_val_prefix), UPB_STRARG(value_name), iter.val); + UPB_STRARG(enum_val_prefix), UPB_STRARG(value_name), upb_enum_iter_number(iter)); upb_string_unref(value_name); } fprintf(stream, "\n} " UPB_STRFMT ";\n\n", UPB_STRARG(enum_name)); @@ -119,12 +110,13 @@ static void write_const_h(upb_def *defs[], int num_entries, char *outfile_name, for(int i = 0; i < num_entries; i++) { /* Foreach enum */ upb_msgdef *m = upb_dyncast_msgdef(defs[i]); if(!m) continue; - upb_strptr msg_name = upb_strdup(UPB_UPCAST(m)->fqname); - upb_strptr msg_val_prefix = upb_strdup(msg_name); + upb_string *msg_name = upb_strdup(UPB_UPCAST(m)->fqname); + upb_string *msg_val_prefix = upb_strdup(msg_name); to_preproc(msg_val_prefix); - for(upb_field_count_t j = 0; j < m->num_fields; j++) { - upb_fielddef *f = &m->fields[j]; - upb_strptr preproc_field_name = upb_strdup(f->name); + upb_msg_iter i; + for(i = upb_msg_begin(m); !upb_msg_done(i); i = upb_msg_next(m, i)) { + upb_fielddef *f = upb_msg_iter_field(i); + upb_string *preproc_field_name = upb_strdup(f->name); to_preproc(preproc_field_name); fprintf(stream, "#define " UPB_STRFMT "_" UPB_STRFMT "_FIELDNUM %d\n", UPB_STRARG(msg_val_prefix), UPB_STRARG(preproc_field_name), f->number); @@ -143,483 +135,12 @@ static void write_const_h(upb_def *defs[], int num_entries, char *outfile_name, upb_string_unref(include_guard_name); } -/* The .h file defines structs for the types defined in the .proto file. It - * also defines constants for the enum values. - * - * Assumes that d has been validated. */ -static void write_h(upb_def *defs[], int num_defs, char *outfile_name, - char *descriptor_cident, FILE *stream) -{ - /* Header file prologue. */ - upb_strptr include_guard_name = upb_strdupc(outfile_name); - to_preproc(include_guard_name); - fputs("/* This file was generated by upbc (the upb compiler). " - "Do not edit. */\n\n", stream), - fprintf(stream, "#ifndef " UPB_STRFMT "\n", UPB_STRARG(include_guard_name)); - fprintf(stream, "#define " UPB_STRFMT "\n\n", UPB_STRARG(include_guard_name)); - fputs("#include \n\n", stream); - fputs("#ifdef __cplusplus\n", stream); - fputs("extern \"C\" {\n", stream); - fputs("#endif\n\n", stream); - - if(descriptor_cident) { - fputs("struct google_protobuf_FileDescriptorSet;\n", stream); - fprintf(stream, "extern struct google_protobuf_FileDescriptorSet *%s;\n\n", - descriptor_cident); - } - - /* Forward declarations. */ - fputs("/* Forward declarations of all message types.\n", stream); - fputs(" * So they can refer to each other in ", stream); - fputs("possibly-recursive ways. */\n\n", stream); - - for(int i = 0; i < num_defs; i++) { /* Foreach message */ - upb_msgdef *m = upb_dyncast_msgdef(defs[i]); - if(!m) continue; - upb_strptr msg_name = upb_strdup(UPB_UPCAST(m)->fqname); - to_cident(msg_name); - fprintf(stream, "struct " UPB_STRFMT ";\n", UPB_STRARG(msg_name)); - fprintf(stream, "typedef struct " UPB_STRFMT "\n " UPB_STRFMT ";\n", - UPB_STRARG(msg_name), UPB_STRARG(msg_name)); - fprintf(stream, "UPB_DEFINE_MSG_ARRAY(" UPB_STRFMT ")\n\n", - UPB_STRARG(msg_name)); - upb_string_unref(msg_name); - } - - /* Message Declarations. */ - fputs("/* The message definitions themselves. */\n\n", stream); - for(int i = 0; i < num_defs; i++) { /* Foreach message */ - upb_msgdef *m = upb_dyncast_msgdef(defs[i]); - if(!m) continue; - upb_strptr msg_name = upb_strdup(UPB_UPCAST(m)->fqname); - to_cident(msg_name); - fprintf(stream, "struct " UPB_STRFMT " {\n", UPB_STRARG(msg_name)); - fputs(" upb_data base;\n", stream); - fputs(" union {\n", stream); - fprintf(stream, " uint8_t bytes[%" PRIu32 "];\n", m->set_flags_bytes); - fputs(" struct {\n", stream); - for(upb_field_count_t j = 0; j < m->num_fields; j++) { - static char* labels[] = {"", "optional", "required", "repeated"}; - upb_fielddef *f = &m->fields[j]; - fprintf(stream, " bool " UPB_STRFMT ":1; /* = %" PRIu32 ", %s. */\n", - UPB_STRARG(f->name), f->number, labels[f->label]); - } - fputs(" } has;\n", stream); - fputs(" } set_flags;\n", stream); - for(upb_field_count_t j = 0; j < m->num_fields; j++) { - upb_fielddef *f = &m->fields[j]; - if(upb_issubmsg(f)) { - upb_strptr type_name = upb_strdup(f->def->fqname); - to_cident(type_name); - if(f->label == UPB_LABEL(REPEATED)) { - fprintf(stream, " UPB_MSG_ARRAYPTR(" UPB_STRFMT ") " UPB_STRFMT ";\n", - UPB_STRARG(type_name), UPB_STRARG(f->name)); - } else { - fprintf(stream, " " UPB_STRFMT "* " UPB_STRFMT ";\n", - UPB_STRARG(type_name), UPB_STRARG(f->name)); - } - upb_string_unref(type_name); - } else if(f->label == UPB_LABEL(REPEATED)) { - fprintf(stream, " upb_arrayptr " UPB_STRFMT ";\n", UPB_STRARG(f->name)); - } else { - static char* c_types[] = { - "", "double", "float", "int64_t", "uint64_t", "int32_t", "uint64_t", - "uint32_t", "bool", "upb_strptr", "", "", - "upb_strptr", "uint32_t", "int32_t", "int32_t", "int64_t", - "int32_t", "int64_t" - }; - fprintf(stream, " %s " UPB_STRFMT ";\n", - c_types[f->type], UPB_STRARG(f->name)); - } - } - fputs("};\n\n", stream); - upb_string_unref(msg_name); - } - - /* Epilogue. */ - fputs("#ifdef __cplusplus\n", stream); - fputs("} /* extern \"C\" */\n", stream); - fputs("#endif\n\n", stream); - fprintf(stream, "#endif /* " UPB_STRFMT " */\n", UPB_STRARG(include_guard_name)); - upb_string_unref(include_guard_name); -} - -/* Format of table entries that we use when analyzing data structures for - * write_messages_c. */ -typedef struct { - upb_strtable_entry e; - int offset; - int num; -} strtable_entry; - -typedef struct { - upb_strtable_entry e; - upb_fielddef *field; - upb_strptr cident; /* Type name converted with to_cident(). */ - /* A list of all values of this type, in an established order. */ - upb_value *values; - int values_size, values_len; - struct array { - int offset; - int len; - upb_arrayptr ptr; /* So we can find it later. */ - } *arrays; - int arrays_size, arrays_len; -} typetable_entry; - -typedef struct { - upb_inttable_entry e; - void *msg; - int num; /* Unique offset into the list of all msgs of this type. */ -} msgtable_entry; - -int compare_entries(const void *_e1, const void *_e2) -{ - strtable_entry *const*e1 = _e1, *const*e2 = _e2; - return upb_strcmp((*e1)->e.key, (*e2)->e.key); -} - -/* Mutually recursive functions to recurse over a set of possibly nested - * messages and extract all the strings. - * - * TODO: make these use a generic msg visitor. */ - -static void add_strings_from_msg(upb_msg *msg, upb_msgdef *md, upb_strtable *t); - -static void add_strings_from_value(upb_value p, upb_fielddef *f, - upb_strtable *t) -{ - if(upb_isstringtype(f->type)) { - strtable_entry e = {.e = {.key = p.str}}; - if(upb_strtable_lookup(t, e.e.key) == NULL) - upb_strtable_insert(t, &e.e); - } else if(upb_issubmsg(f)) { - add_strings_from_msg(p.msg, upb_downcast_msgdef(f->def), t); - } -} - -static void add_strings_from_msg(upb_msg *msg, upb_msgdef *md, upb_strtable *t) -{ - for(upb_field_count_t i = 0; i < md->num_fields; i++) { - upb_fielddef *f = &md->fields[i]; - if(!upb_msg_has(msg, f)) continue; - upb_value p = upb_msg_get(msg, f); - if(upb_isarray(f)) { - upb_arrayptr arr = p.arr; - for(uint32_t j = 0; j < upb_array_len(arr); j++) - add_strings_from_value(upb_array_get(arr, f, j), f, t); - } else { - add_strings_from_value(p, f, t); - } - } -} - -/* Mutually recursive functions to recurse over a set of possibly nested - * messages and extract all the messages (keyed by type). - * - * TODO: make these use a generic msg visitor. */ - -typetable_entry *get_or_insert_typeentry(upb_strtable *t, upb_fielddef *f) -{ - upb_strptr type_name = upb_issubmsg(f) ? - upb_strdup(f->def->fqname) : upb_strdupc(upb_types[f->type].ctype); - typetable_entry *type_e = upb_strtable_lookup(t, type_name); - if(type_e == NULL) { - upb_strptr cident = upb_strdup(type_name); - to_cident(cident); - typetable_entry new_type_e = { - .e = {.key = type_name}, .field = f, .cident = cident, - .values = NULL, .values_size = 0, .values_len = 0, - .arrays = NULL, .arrays_size = 0, .arrays_len = 0 - }; - assert(upb_strtable_lookup(t, type_name) == NULL); - assert(upb_strtable_lookup(t, new_type_e.e.key) == NULL); - upb_strtable_insert(t, &new_type_e.e); - type_e = upb_strtable_lookup(t, type_name); - assert(type_e); - } - upb_string_unref(type_name); - return type_e; -} - -static void add_value(upb_value v, upb_fielddef *f, upb_strtable *t) -{ - typetable_entry *type_e = get_or_insert_typeentry(t, f); - if(type_e->values_len == type_e->values_size) { - type_e->values_size = UPB_MAX(type_e->values_size * 2, 4); - type_e->values = realloc(type_e->values, sizeof(*type_e->values) * type_e->values_size); - } - type_e->values[type_e->values_len++] = v; -} - -static void add_submsgs(upb_msg *msg, upb_msgdef *md, upb_strtable *t) -{ - for(upb_field_count_t i = 0; i < md->num_fields; i++) { - upb_fielddef *f = &md->fields[i]; - if(!upb_msg_has(msg, f)) continue; - upb_value v = upb_msg_get(msg, f); - if(upb_isarray(f)) { - if(upb_isstring(f)) continue; /* Handled by a different code-path. */ - upb_arrayptr arr = v.arr; - - /* Add to our list of arrays for this type. */ - typetable_entry *arr_type_e = get_or_insert_typeentry(t, f); - if(arr_type_e->arrays_len == arr_type_e->arrays_size) { - arr_type_e->arrays_size = UPB_MAX(arr_type_e->arrays_size * 2, 4); - arr_type_e->arrays = realloc(arr_type_e->arrays, - sizeof(*arr_type_e->arrays)*arr_type_e->arrays_size); - } - arr_type_e->arrays[arr_type_e->arrays_len].offset = arr_type_e->values_len; - arr_type_e->arrays[arr_type_e->arrays_len].len = upb_array_len(arr); - arr_type_e->arrays[arr_type_e->arrays_len].ptr = v.arr; - arr_type_e->arrays_len++; - - /* Add the individual values in the array. */ - for(uint32_t j = 0; j < upb_array_len(arr); j++) - add_value(upb_array_get(arr, f, j), f, t); - - /* Add submsgs. We must do this separately so that the msgs in this - * array are contiguous (and don't have submsgs of the same type - * interleaved). */ - for(uint32_t j = 0; j < upb_array_len(arr); j++) - add_submsgs(upb_array_get(arr, f, j).msg, upb_downcast_msgdef(f->def), t); - } else { - if(!upb_issubmsg(f)) continue; - add_value(v, f, t); - add_submsgs(v.msg, upb_downcast_msgdef(f->def), t); - } - } -} - -/* write_messages_c emits a .c file that contains the data of a protobuf, - * serialized as C structures. */ -static void write_message_c(upb_msg *msg, upb_msgdef *md, - char *cident, char *hfile_name, - int argc, char *argv[], char *infile_name, - FILE *stream) -{ - fputs( - "/*\n" - " * This file is a data dump of a protocol buffer into a C structure.\n" - " * It was created by the upb compiler (upbc) with the following\n" - " * command-line:\n" - " *\n", stream); - fputs(" * ", stream); - for(int i = 0; i < argc; i++) { - fputs(argv[i], stream); - if(i < argc-1) fputs(" ", stream); - } - fputs("\n *\n", stream); - fprintf(stream, " * This file is a dump of '%s'.\n", infile_name); - fputs( - " * It contains exactly the same data, but in a C structure form\n" - " * instead of a serialized protobuf. This file contains no code,\n" - " * only data.\n" - " *\n" - " * This file was auto-generated. Do not edit. */\n\n", stream); - - fprintf(stream, "#include \"%s\"\n\n", hfile_name); - - /* Gather all strings into a giant string. Use a hash to prevent adding the - * same string more than once. */ - upb_strtable strings; - upb_strtable_init(&strings, 16, sizeof(strtable_entry)); - add_strings_from_msg(msg, md, &strings); - - int size; - strtable_entry **str_entries = strtable_to_array(&strings, &size); - /* Sort for nice size and reproduceability. */ - qsort(str_entries, size, sizeof(void*), compare_entries); - - /* Emit strings. */ - fputs("static char strdata[] =\n \"", stream); - int col = 2; - int offset = 0; - for(int i = 0; i < size; i++) { - upb_strptr s = str_entries[i]->e.key; - str_entries[i]->offset = offset; - str_entries[i]->num = i; - const char *buf = upb_string_getrobuf(s); - for(uint32_t j = 0; j < upb_strlen(s); j++) { - if(++col == 80) { - fputs("\"\n \"", stream); - col = 3; - } - fputc(buf[j], stream); - } - offset += upb_strlen(s); - } - fputs("\";\n\n", stream); - - fputs("static upb_static_string strings[] = {\n", stream); - for(int i = 0; i < size; i++) { - strtable_entry *e = str_entries[i]; - fprintf(stream, " UPB_STATIC_STRING_INIT_LEN(&strdata[%d], %d),\n", e->offset, upb_strlen(e->e.key)); - } - fputs("};\n\n", stream); - free(str_entries); - - /* Gather a list of types for which we are emitting data, and give each msg - * a unique number within its type. */ - upb_strtable types; - upb_strtable_init(&types, 16, sizeof(typetable_entry)); - upb_value val = {.msg = msg}; - /* A fake field to get the recursion going. */ - upb_fielddef fake_field = { - .type = UPB_TYPE(MESSAGE), - .def = UPB_UPCAST(md), - }; - add_value(val, &fake_field, &types); - add_submsgs(msg, md, &types); - - /* Emit foward declarations for all msgs of all types, and define arrays. */ - fprintf(stream, "/* Forward declarations of messages, and array decls. */\n"); - typetable_entry *e = upb_strtable_begin(&types); - for(; e; e = upb_strtable_next(&types, &e->e)) { - fprintf(stream, "static " UPB_STRFMT " " UPB_STRFMT "_values[%d];\n\n", - UPB_STRARG(e->cident), UPB_STRARG(e->cident), e->values_len); - if(e->arrays_len > 0) { - fprintf(stream, "static " UPB_STRFMT " *" UPB_STRFMT "_array_elems[] = {\n", - UPB_STRARG(e->cident), UPB_STRARG(e->cident)); - for(int i = 0; i < e->arrays_len; i++) { - struct array *arr = &e->arrays[i]; - for(int j = 0; j < arr->len; j++) - fprintf(stream, " &" UPB_STRFMT "_values[%d],\n", UPB_STRARG(e->cident), arr->offset + j); - } - fprintf(stream, "};\n"); - - int cum_offset = 0; - fprintf(stream, "static upb_static_array " UPB_STRFMT "_arrays[%d] = {\n", - UPB_STRARG(e->cident), e->arrays_len); - for(int i = 0; i < e->arrays_len; i++) { - struct array *arr = &e->arrays[i]; - fprintf(stream, " UPB_STATIC_ARRAY_INIT(&" UPB_STRFMT "_array_elems[%d], %d),\n", - UPB_STRARG(e->cident), cum_offset, arr->len); - cum_offset += arr->len; - } - fprintf(stream, "};\n"); - } - } - - /* Emit definitions. */ - for(e = upb_strtable_begin(&types); e; e = upb_strtable_next(&types, &e->e)) { - fprintf(stream, "static " UPB_STRFMT " " UPB_STRFMT "_values[%d] = {\n\n", - UPB_STRARG(e->cident), UPB_STRARG(e->cident), e->values_len); - for(int i = 0; i < e->values_len; i++) { - upb_value val = e->values[i]; - if(upb_issubmsg(e->field)) { - upb_msgdef *m = upb_downcast_msgdef(e->field->def); - void *msgdata = val.msg; - fputs(" {.base = {UPB_DATA_FROZEN},\n", stream); - /* Print set flags. */ - fputs(" .set_flags = {.has = {\n", stream); - for(upb_field_count_t j = 0; j < m->num_fields; j++) { - upb_fielddef *f = &m->fields[j]; - fprintf(stream, " ." UPB_STRFMT " = ", UPB_STRARG(f->name)); - if(upb_msg_has(msgdata, f)) - fprintf(stream, "true"); - else - fprintf(stream, "false"); - fputs(",\n", stream); - } - fputs(" }},\n", stream); - /* Print msg data. */ - for(upb_field_count_t j = 0; j < m->num_fields; j++) { - upb_fielddef *f = &m->fields[j]; - upb_value val = upb_msg_get(msgdata, f); - fprintf(stream, " ." UPB_STRFMT " = ", UPB_STRARG(f->name)); - if(!upb_msg_has(msgdata, f)) { - if(upb_isarray(f) && upb_issubmsg(f)) { - // This is gross and needs cleanup. - fputs("{UPB_ARRAY_NULL_INITIALIZER}, /* Not set. */", stream); - } else if(upb_isarray(f)) { - fputs("UPB_ARRAY_NULL_INITIALIZER, /* Not set. */", stream); - } else if(upb_isstring(f)) { - fputs("UPB_STRING_NULL_INITIALIZER, /* Not set. */", stream); - } else { - fputs("0, /* Not set. */", stream); - } - } else if(upb_isstring(f)) { - if(upb_isarray(f)) { - fputs("Ack, string arrays are not supported yet!\n", stderr); - exit(1); - } else { - strtable_entry *str_e = upb_strtable_lookup(&strings, val.str); - assert(str_e); - fprintf(stream, "UPB_STATIC_STRING_PTR_INIT(strings[%d]), /* \"" UPB_STRFMT "\" */", - str_e->num, UPB_STRARG(val.str)); - } - } else if(upb_isarray(f)) { - /* Find this submessage in the list of msgs for that type. */ - typetable_entry *type_e = get_or_insert_typeentry(&types, f); - assert(type_e); - int arr_num = -1; - for(int k = 0; k < type_e->arrays_len; k++) { - if(upb_array_ptreql(type_e->arrays[k].ptr, val.arr)) { - arr_num = k; - break; - } - } - assert(arr_num != -1); - fprintf(stream, "UPB_STATIC_ARRAY_PTR_TYPED_INIT(" UPB_STRFMT "_arrays[%d]),", UPB_STRARG(type_e->cident), arr_num); - } else if(upb_issubmsg(f)) { - /* Find this submessage in the list of msgs for that type. */ - typetable_entry *type_e = get_or_insert_typeentry(&types, f); - assert(type_e); - int msg_num = -1; - for(int k = 0; k < type_e->values_len; k++) { - if(type_e->values[k].msg == val.msg) { - msg_num = k; - break; - } - } - assert(msg_num != -1); - fprintf(stream, "&" UPB_STRFMT "_values[%d],", UPB_STRARG(type_e->cident), msg_num); - } else { - upb_text_printval(f->type, val, stream); - fputs(",", stream); - } - fputs("\n", stream); - } - fputs(" },\n", stream); - } else if(upb_isstring(e->field)) { - - } else { - /* Non string, non-message data. */ - upb_text_printval(e->field->type, val, stream); - } - } - fputs("};\n", stream); - } - - typetable_entry *toplevel_type = get_or_insert_typeentry(&types, &fake_field); - assert(toplevel_type); - fputs("/* The externally-visible definition. */\n", stream); - /* It is always at offset zero, because we add it first. */ - fprintf(stream, UPB_STRFMT " *%s = &" UPB_STRFMT "_values[0];\n", - UPB_STRARG(toplevel_type->cident), cident, - UPB_STRARG(toplevel_type->cident)); - - /* Free tables. */ - for(e = upb_strtable_begin(&types); e; e = upb_strtable_next(&types, &e->e)) { - upb_string_unref(e->cident); - free(e->values); - free(e->arrays); - } - upb_strtable_free(&types); - upb_strtable_free(&strings); -} - const char usage[] = "upbc -- upb compiler.\n" "upb v0.1 http://blog.reverberate.org/upb/\n" "\n" "Usage: upbc [options] descriptor-file\n" "\n" - " -i C-IDENFITER Output the descriptor as a C data structure with the\n" - " given identifier (otherwise only a header will be\n" - " generated\n" - "\n" " -o OUTFILE-BASE Write to OUTFILE-BASE.h and OUTFILE-BASE.c instead\n" " of using the input file as a basename.\n" ; @@ -641,22 +162,10 @@ void error(char *err, ...) exit(1); } -void sort_fields_in_descriptor(google_protobuf_DescriptorProto *d) -{ - // XXX: modifying the array in place is totally not allowed. - if(d->set_flags.has.field) { - upb_fielddef_sortfds(_upb_array_getptr_raw(d->field.ptr, 0, 0)._void, - google_protobuf_FieldDescriptorProto_array_len(d->field)); - } - if(d->set_flags.has.nested_type) - for(uint32_t i = 0; i < google_protobuf_DescriptorProto_array_len(d->nested_type); i++) - sort_fields_in_descriptor(google_protobuf_DescriptorProto_array_get(d->nested_type, i)); -} - int main(int argc, char *argv[]) { /* Parse arguments. */ - char *outfile_base = NULL, *input_file = NULL, *cident = NULL; + char *outfile_base = NULL, *input_file = NULL; for(int i = 1; i < argc; i++) { if(strcmp(argv[i], "-o") == 0) { if(++i == argc) @@ -664,12 +173,6 @@ int main(int argc, char *argv[]) else if(outfile_base) usage_err("-o was specified multiple times."); outfile_base = argv[i]; - } else if(strcmp(argv[i], "-i") == 0) { - if(++i == argc) - usage_err("-i must be followed by a C-IDENTIFIER."); - else if(cident) - usage_err("-i was specified multiple times."); - cident = argv[i]; } else { if(input_file) usage_err("You can only specify one input file."); @@ -680,70 +183,54 @@ int main(int argc, char *argv[]) if(!outfile_base) outfile_base = input_file; // Read and parse input file. - upb_strptr descriptor = upb_strreadfile(input_file); - if(upb_string_isnull(descriptor)) + upb_string *descriptor = upb_strreadfile(input_file); + if(!descriptor) error("Couldn't read input file."); + + // TODO: make upb_parsedesc use a separate symtab, so we can use it here when + // importing descriptor.proto. upb_symtab *s = upb_symtab_new(); - upb_msg *fds_msg = upb_msg_new(s->fds_msgdef); + upb_symtab_add_descriptorproto(s); + upb_symtab *s2 = upb_symtab_new(); upb_status status = UPB_STATUS_INIT; - upb_msg_decodestr(fds_msg, s->fds_msgdef, descriptor, &status); - if(!upb_ok(&status)) - error("Failed to parse input file descriptor: %s", status.msg); - google_protobuf_FileDescriptorSet *fds = (void*)fds_msg; - upb_symtab_add_desc(s, descriptor, &status); - if(!upb_ok(&status)) - error("Failed to add descriptor: %s", status.msg); + upb_stringsrc strsrc; + upb_stringsrc_init(&strsrc); + upb_stringsrc_reset(&strsrc, descriptor); + + upb_decoder d; + upb_msgdef *fds_msgdef = upb_symtab_fds_def(s); + upb_decoder_init(&d, fds_msgdef); + upb_decoder_reset(&d, upb_stringsrc_bytesrc(&strsrc)); - // We need to sort the fields of all the descriptors. This is currently - // somewhat special-cased to when we are emitting a descriptor for - // FileDescriptorProto, which is used internally for bootstrapping. - // - // The fundamental issue is that we will be parsing descriptors into memory - // using a reflection-based code-path, but upb then reads the descriptors - // from memory using the C structs emitted by upbc. This means that the - // msgdef we will use internally to parse the descriptors must use the same - // field order as the .h files we are about to generate. But the msgdefs we - // will use to generate those .h files have already been sorted according to - // this scheme. - // - // If/when we ever make upbc more general, we'll have to revisit this. - for(uint32_t i = 0; i < google_protobuf_FileDescriptorProto_array_len(fds->file); i++) { - google_protobuf_FileDescriptorProto *fd = google_protobuf_FileDescriptorProto_array_get(fds->file, i); - if(!fd->set_flags.has.message_type) continue; - for(uint32_t j = 0; j < google_protobuf_DescriptorProto_array_len(fd->message_type); j++) - sort_fields_in_descriptor(google_protobuf_DescriptorProto_array_get(fd->message_type, j)); + upb_symtab_addfds(s2, upb_decoder_src(&d), &status); + upb_stringsrc_uninit(&strsrc); + upb_decoder_uninit(&d); + upb_def_unref(UPB_UPCAST(fds_msgdef)); + + if(!upb_ok(&status)) { + upb_printerr(&status); + error("Failed to parse input file descriptor\n"); } + upb_status_uninit(&status); + /* Emit output files. */ const int maxsize = 256; - char h_filename[maxsize], h_const_filename[maxsize], c_filename[maxsize]; - if(snprintf(h_filename, maxsize, "%s.h", outfile_base) >= maxsize || - snprintf(c_filename, maxsize, "%s.c", outfile_base) >= maxsize || - snprintf(h_const_filename, maxsize, "%s_const.h", outfile_base) >= maxsize) + char h_const_filename[maxsize]; + if(snprintf(h_const_filename, maxsize, "%s_const.h", outfile_base) >= maxsize) error("File base too long.\n"); - FILE *h_file = fopen(h_filename, "w"); - if(!h_file) error("Failed to open .h output file"); FILE *h_const_file = fopen(h_const_filename, "w"); - if(!h_const_file) error("Failed to open _const.h output file"); + if(!h_const_file) error("Failed to open _const.h output file\n"); int symcount; upb_def **defs = upb_symtab_getdefs(s, &symcount, UPB_DEF_ANY); - write_h(defs, symcount, h_filename, cident, h_file); - write_const_h(defs, symcount, h_filename, h_const_file); + write_const_h(defs, symcount, h_const_filename, h_const_file); for (int i = 0; i < symcount; i++) upb_def_unref(defs[i]); free(defs); - if(cident) { - FILE *c_file = fopen(c_filename, "w"); - if(!c_file) error("Failed to open .h output file"); - write_message_c(fds_msg, s->fds_msgdef, cident, h_filename, argc, argv, input_file, c_file); - fclose(c_file); - } - upb_msg_unref(fds_msg, s->fds_msgdef); upb_string_unref(descriptor); upb_symtab_unref(s); - fclose(h_file); fclose(h_const_file); return 0; -- cgit v1.2.3