From f9afc3e55bbc289df41606d493377318c6645817 Mon Sep 17 00:00:00 2001 From: Josh Haberman Date: Wed, 13 Jan 2016 18:04:48 -0800 Subject: Changed JSON parser/printer to correctly camelCase names. --- upb/json/parser.c | 206 +++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 157 insertions(+), 49 deletions(-) (limited to 'upb/json/parser.c') diff --git a/upb/json/parser.c b/upb/json/parser.c index f9771f9..8a50b3a 100644 --- a/upb/json/parser.c +++ b/upb/json/parser.c @@ -40,6 +40,9 @@ typedef struct { const upb_msgdef *m; const upb_fielddef *f; + /* The table mapping json name to fielddef for this message. */ + upb_strtable *name_table; + /* We are in a repeated-field context, ready to emit mapentries as * submessages. This flag alters the start-of-object (open-brace) behavior to * begin a sequence of mapentry messages rather than a single submessage. */ @@ -60,7 +63,7 @@ typedef struct { struct upb_json_parser { upb_env *env; - upb_byteshandler input_handler_; + const upb_json_parsermethod *method; upb_bytessink input_; /* Stack to track the JSON scopes we are in. */ @@ -95,6 +98,19 @@ struct upb_json_parser { uint32_t digit; }; +struct upb_json_parsermethod { + upb_refcounted base; + + upb_byteshandler input_handler_; + + /* Mainly for the purposes of refcounting, so all the fielddefs we point + * to stay alive. */ + const upb_msgdef *msg; + + /* Keys are upb_msgdef*, values are upb_strtable (json_name -> fielddef) */ + upb_inttable name_tables; +}; + #define PARSER_CHECK_RETURN(x) if (!(x)) return false /* Used to signal that a capture has been suspended. */ @@ -123,6 +139,13 @@ static bool check_stack(upb_json_parser *p) { return true; } +static void set_name_table(upb_json_parser *p, upb_jsonparser_frame *frame) { + upb_value v; + bool ok = upb_inttable_lookupptr(&p->method->name_tables, frame->m, &v); + UPB_ASSERT_VAR(ok, ok); + frame->name_table = upb_value_getptr(v); +} + /* There are GCC/Clang built-ins for overflow checking which we could start * using if there was any performance benefit to it. */ @@ -719,6 +742,7 @@ static bool start_stringval(upb_json_parser *p) { upb_sink_startstr(&p->top->sink, sel, 0, &inner->sink); inner->m = p->top->m; inner->f = p->top->f; + inner->name_table = NULL; inner->is_map = false; inner->is_mapentry = false; p->top = inner; @@ -905,6 +929,7 @@ static bool handle_mapentry(upb_json_parser *p) { sel = getsel_for_handlertype(p, UPB_HANDLER_STARTSUBMSG); upb_sink_startsubmsg(&p->top->sink, sel, &inner->sink); inner->m = mapentrymsg; + inner->name_table = NULL; inner->mapfield = mapfield; inner->is_map = false; @@ -941,20 +966,20 @@ static bool end_membername(upb_json_parser *p) { } else { size_t len; const char *buf = accumulate_getptr(p, &len); - const upb_fielddef *f = upb_msgdef_ntof(p->top->m, buf, len); + upb_value v; - if (!f) { + if (upb_strtable_lookup2(p->top->name_table, buf, len, &v)) { + p->top->f = upb_value_getconstptr(v); + multipart_end(p); + + return true; + } else { /* TODO(haberman): Ignore unknown fields if requested/configured to do * so. */ upb_status_seterrf(&p->status, "No such field: %.*s\n", (int)len, buf); upb_env_reporterror(p->env, &p->status); return false; } - - p->top->f = f; - multipart_end(p); - - return true; } } @@ -996,6 +1021,7 @@ static bool start_subobject(upb_json_parser *p) { sel = getsel_for_handlertype(p, UPB_HANDLER_STARTSEQ); upb_sink_startseq(&p->top->sink, sel, &inner->sink); inner->m = upb_fielddef_msgsubdef(p->top->f); + inner->name_table = NULL; inner->mapfield = p->top->f; inner->f = NULL; inner->is_map = true; @@ -1016,6 +1042,7 @@ static bool start_subobject(upb_json_parser *p) { sel = getsel_for_handlertype(p, UPB_HANDLER_STARTSUBMSG); upb_sink_startsubmsg(&p->top->sink, sel, &inner->sink); inner->m = upb_fielddef_msgsubdef(p->top->f); + set_name_table(p, inner); inner->f = NULL; inner->is_map = false; inner->is_mapentry = false; @@ -1065,6 +1092,7 @@ static bool start_array(upb_json_parser *p) { sel = getsel_for_handlertype(p, UPB_HANDLER_STARTSEQ); upb_sink_startseq(&p->top->sink, sel, &inner->sink); inner->m = p->top->m; + inner->name_table = NULL; inner->f = p->top->f; inner->is_map = false; inner->is_mapentry = false; @@ -1122,11 +1150,11 @@ static void end_object(upb_json_parser *p) { * final state once, when the closing '"' is seen. */ -#line 1218 "upb/json/parser.rl" +#line 1246 "upb/json/parser.rl" -#line 1130 "upb/json/parser.c" +#line 1158 "upb/json/parser.c" static const char _json_actions[] = { 0, 1, 0, 1, 2, 1, 3, 1, 5, 1, 6, 1, 7, 1, 8, 1, @@ -1275,7 +1303,7 @@ static const int json_en_value_machine = 27; static const int json_en_main = 1; -#line 1221 "upb/json/parser.rl" +#line 1249 "upb/json/parser.rl" size_t parse(void *closure, const void *hd, const char *buf, size_t size, const upb_bufhandle *handle) { @@ -1297,7 +1325,7 @@ size_t parse(void *closure, const void *hd, const char *buf, size_t size, capture_resume(parser, buf); -#line 1301 "upb/json/parser.c" +#line 1329 "upb/json/parser.c" { int _klen; unsigned int _trans; @@ -1372,118 +1400,118 @@ _match: switch ( *_acts++ ) { case 0: -#line 1133 "upb/json/parser.rl" +#line 1161 "upb/json/parser.rl" { p--; {cs = stack[--top]; goto _again;} } break; case 1: -#line 1134 "upb/json/parser.rl" +#line 1162 "upb/json/parser.rl" { p--; {stack[top++] = cs; cs = 10; goto _again;} } break; case 2: -#line 1138 "upb/json/parser.rl" +#line 1166 "upb/json/parser.rl" { start_text(parser, p); } break; case 3: -#line 1139 "upb/json/parser.rl" +#line 1167 "upb/json/parser.rl" { CHECK_RETURN_TOP(end_text(parser, p)); } break; case 4: -#line 1145 "upb/json/parser.rl" +#line 1173 "upb/json/parser.rl" { start_hex(parser); } break; case 5: -#line 1146 "upb/json/parser.rl" +#line 1174 "upb/json/parser.rl" { hexdigit(parser, p); } break; case 6: -#line 1147 "upb/json/parser.rl" +#line 1175 "upb/json/parser.rl" { CHECK_RETURN_TOP(end_hex(parser)); } break; case 7: -#line 1153 "upb/json/parser.rl" +#line 1181 "upb/json/parser.rl" { CHECK_RETURN_TOP(escape(parser, p)); } break; case 8: -#line 1159 "upb/json/parser.rl" +#line 1187 "upb/json/parser.rl" { p--; {cs = stack[--top]; goto _again;} } break; case 9: -#line 1162 "upb/json/parser.rl" +#line 1190 "upb/json/parser.rl" { {stack[top++] = cs; cs = 19; goto _again;} } break; case 10: -#line 1164 "upb/json/parser.rl" +#line 1192 "upb/json/parser.rl" { p--; {stack[top++] = cs; cs = 27; goto _again;} } break; case 11: -#line 1169 "upb/json/parser.rl" +#line 1197 "upb/json/parser.rl" { start_member(parser); } break; case 12: -#line 1170 "upb/json/parser.rl" +#line 1198 "upb/json/parser.rl" { CHECK_RETURN_TOP(end_membername(parser)); } break; case 13: -#line 1173 "upb/json/parser.rl" +#line 1201 "upb/json/parser.rl" { end_member(parser); } break; case 14: -#line 1179 "upb/json/parser.rl" +#line 1207 "upb/json/parser.rl" { start_object(parser); } break; case 15: -#line 1182 "upb/json/parser.rl" +#line 1210 "upb/json/parser.rl" { end_object(parser); } break; case 16: -#line 1188 "upb/json/parser.rl" +#line 1216 "upb/json/parser.rl" { CHECK_RETURN_TOP(start_array(parser)); } break; case 17: -#line 1192 "upb/json/parser.rl" +#line 1220 "upb/json/parser.rl" { end_array(parser); } break; case 18: -#line 1197 "upb/json/parser.rl" +#line 1225 "upb/json/parser.rl" { start_number(parser, p); } break; case 19: -#line 1198 "upb/json/parser.rl" +#line 1226 "upb/json/parser.rl" { CHECK_RETURN_TOP(end_number(parser, p)); } break; case 20: -#line 1200 "upb/json/parser.rl" +#line 1228 "upb/json/parser.rl" { CHECK_RETURN_TOP(start_stringval(parser)); } break; case 21: -#line 1201 "upb/json/parser.rl" +#line 1229 "upb/json/parser.rl" { CHECK_RETURN_TOP(end_stringval(parser)); } break; case 22: -#line 1203 "upb/json/parser.rl" +#line 1231 "upb/json/parser.rl" { CHECK_RETURN_TOP(parser_putbool(parser, true)); } break; case 23: -#line 1205 "upb/json/parser.rl" +#line 1233 "upb/json/parser.rl" { CHECK_RETURN_TOP(parser_putbool(parser, false)); } break; case 24: -#line 1207 "upb/json/parser.rl" +#line 1235 "upb/json/parser.rl" { /* null value */ } break; case 25: -#line 1209 "upb/json/parser.rl" +#line 1237 "upb/json/parser.rl" { CHECK_RETURN_TOP(start_subobject(parser)); } break; case 26: -#line 1210 "upb/json/parser.rl" +#line 1238 "upb/json/parser.rl" { end_subobject(parser); } break; case 27: -#line 1215 "upb/json/parser.rl" +#line 1243 "upb/json/parser.rl" { p--; {cs = stack[--top]; goto _again;} } break; -#line 1487 "upb/json/parser.c" +#line 1515 "upb/json/parser.c" } } @@ -1496,7 +1524,7 @@ _again: _out: {} } -#line 1242 "upb/json/parser.rl" +#line 1270 "upb/json/parser.rl" if (p != pe) { upb_status_seterrf(&parser->status, "Parse error at %s\n", p); @@ -1537,13 +1565,13 @@ static void json_parser_reset(upb_json_parser *p) { /* Emit Ragel initialization of the parser. */ -#line 1541 "upb/json/parser.c" +#line 1569 "upb/json/parser.c" { cs = json_start; top = 0; } -#line 1282 "upb/json/parser.rl" +#line 1310 "upb/json/parser.rl" p->current_state = cs; p->parser_top = top; accumulate_clear(p); @@ -1553,10 +1581,65 @@ static void json_parser_reset(upb_json_parser *p) { upb_status_clear(&p->status); } +static void visit_json_parsermethod(const upb_refcounted *r, + upb_refcounted_visit *visit, + void *closure) { + const upb_json_parsermethod *method = (upb_json_parsermethod*)r; + visit(r, upb_msgdef_upcast2(method->msg), closure); +} + +static void free_json_parsermethod(upb_refcounted *r) { + upb_json_parsermethod *method = (upb_json_parsermethod*)r; + + upb_inttable_iter i; + upb_inttable_begin(&i, &method->name_tables); + for(; !upb_inttable_done(&i); upb_inttable_next(&i)) { + upb_value val = upb_inttable_iter_value(&i); + upb_strtable *t = upb_value_getptr(val); + upb_strtable_uninit(t); + free(t); + } + + upb_inttable_uninit(&method->name_tables); + + free(r); +} + +static void add_jsonname_table(upb_json_parsermethod *m, const upb_msgdef* md) { + upb_msg_field_iter i; + upb_strtable *t; + + if (upb_inttable_lookupptr(&m->name_tables, md, NULL)) { + return; + } + + /* TODO(haberman): handle malloc failure. */ + t = malloc(sizeof(*t)); + upb_strtable_init(t, UPB_CTYPE_CONSTPTR); + upb_inttable_insertptr(&m->name_tables, md, upb_value_ptr(t)); + + for(upb_msg_field_begin(&i, md); + !upb_msg_field_done(&i); + upb_msg_field_next(&i)) { + const upb_fielddef *f = upb_msg_iter_field(&i); + /* It would be nice to stack-allocate this, but protobufs do not limit the + * length of fields to any reasonable limit. */ + char *buf = malloc(strlen(upb_fielddef_name(f)) + 1); + upb_fielddef_getjsonname(f, buf); + upb_strtable_insert(t, buf, upb_value_constptr(f)); + free(buf); + + if (upb_fielddef_issubmsg(f)) { + add_jsonname_table(m, upb_fielddef_msgsubdef(f)); + } + } +} /* Public API *****************************************************************/ -upb_json_parser *upb_json_parser_create(upb_env *env, upb_sink *output) { +upb_json_parser *upb_json_parser_create(upb_env *env, + const upb_json_parsermethod *method, + upb_sink *output) { #ifndef NDEBUG const size_t size_before = upb_env_bytesallocated(env); #endif @@ -1564,17 +1647,16 @@ upb_json_parser *upb_json_parser_create(upb_env *env, upb_sink *output) { if (!p) return false; p->env = env; + p->method = method; p->limit = p->stack + UPB_JSON_MAX_DEPTH; p->accumulate_buf = NULL; p->accumulate_buf_size = 0; - upb_byteshandler_init(&p->input_handler_); - upb_byteshandler_setstring(&p->input_handler_, parse, NULL); - upb_byteshandler_setendstr(&p->input_handler_, end, NULL); - upb_bytessink_reset(&p->input_, &p->input_handler_, p); + upb_bytessink_reset(&p->input_, &method->input_handler_, p); json_parser_reset(p); upb_sink_reset(&p->top->sink, output->handlers, output->closure); p->top->m = upb_handlers_msgdef(output->handlers); + set_name_table(p, p->top); /* If this fails, uncomment and increase the value in parser.h. */ /* fprintf(stderr, "%zd\n", upb_env_bytesallocated(env) - size_before); */ @@ -1585,3 +1667,29 @@ upb_json_parser *upb_json_parser_create(upb_env *env, upb_sink *output) { upb_bytessink *upb_json_parser_input(upb_json_parser *p) { return &p->input_; } + +upb_json_parsermethod *upb_json_parsermethod_new(const upb_msgdef* md, + const void* owner) { + static const struct upb_refcounted_vtbl vtbl = {visit_json_parsermethod, + free_json_parsermethod}; + upb_json_parsermethod *ret = malloc(sizeof(*ret)); + upb_refcounted_init(upb_json_parsermethod_upcast_mutable(ret), &vtbl, owner); + + ret->msg = md; + upb_ref2(md, ret); + + upb_byteshandler_init(&ret->input_handler_); + upb_byteshandler_setstring(&ret->input_handler_, parse, ret); + upb_byteshandler_setendstr(&ret->input_handler_, end, ret); + + upb_inttable_init(&ret->name_tables, UPB_CTYPE_PTR); + + add_jsonname_table(ret, md); + + return ret; +} + +const upb_byteshandler *upb_json_parsermethod_inputhandler( + const upb_json_parsermethod *m) { + return &m->input_handler_; +} -- cgit v1.2.3