// This file Copyright © 2008-2022 Mnemosyne LLC. // It may be used under GPLv2 (SPDX: GPL-2.0-only), GPLv3 (SPDX: GPL-3.0-only), // or any future license endorsed by Mnemosyne LLC. // License text can be found in the licenses/ folder. #include #include #include /* EILSEQ, EINVAL */ #include /* fabs() */ #include #include #include #include #include #include #define UTF_CPP_CPLUSPLUS 201703L #include #include #include #define LIBTRANSMISSION_VARIANT_MODULE #include "transmission.h" #include "error.h" #include "jsonsl.h" #include "log.h" #include "quark.h" #include "tr-assert.h" #include "tr-buffer.h" #include "utils.h" #include "variant-common.h" #include "variant.h" using namespace std::literals; using Buffer = libtransmission::Buffer; /* arbitrary value... this is much deeper than our code goes */ static auto constexpr MaxDepth = size_t{ 64 }; struct json_wrapper_data { bool has_content; size_t size; std::string_view key; std::string keybuf; std::string strbuf; tr_error* error; std::deque stack; tr_variant* top; int parse_opts; /* A very common pattern is for a container's children to be similar, * e.g. they may all be objects with the same set of keys. So when * a container is popped off the stack, remember its size to use as * a preallocation heuristic for the next container at that depth. */ std::array preallocGuess; }; static tr_variant* get_node(struct jsonsl_st* jsn) { auto* data = static_cast(jsn->data); auto* parent = std::empty(data->stack) ? nullptr : data->stack.back(); tr_variant* node = nullptr; if (parent == nullptr) { node = data->top; } else if (tr_variantIsList(parent)) { node = tr_variantListAdd(parent); } else if (tr_variantIsDict(parent) && !std::empty(data->key)) { node = tr_variantDictAdd(parent, tr_quark_new(data->key)); data->key = ""sv; } return node; } static void error_handler(jsonsl_t jsn, jsonsl_error_t error, jsonsl_state_st* /*state*/, jsonsl_char_t const* buf) { auto* data = static_cast(jsn->data); tr_error_set( &data->error, EILSEQ, fmt::format( _("Couldn't parse JSON at position {position} '{text}': {error} ({error_code})"), fmt::arg("position", jsn->pos), fmt::arg("text", std::string_view{ buf, std::min(size_t{ 16U }, data->size - jsn->pos) }), fmt::arg("error", jsonsl_strerror(error)), fmt::arg("error_code", error))); } static int error_callback(jsonsl_t jsn, jsonsl_error_t error, struct jsonsl_state_st* state, jsonsl_char_t* at) { error_handler(jsn, error, state, at); return 0; /* bail */ } static void action_callback_PUSH( jsonsl_t jsn, jsonsl_action_t /*action*/, struct jsonsl_state_st* state, jsonsl_char_t const* /*buf*/) { auto* const data = static_cast(jsn->data); if ((state->type == JSONSL_T_LIST) || (state->type == JSONSL_T_OBJECT)) { data->has_content = true; tr_variant* node = get_node(jsn); data->stack.push_back(node); size_t const depth = std::size(data->stack); size_t const n = depth < MaxDepth ? data->preallocGuess[depth] : 0; if (state->type == JSONSL_T_LIST) { tr_variantInitList(node, n); } else { tr_variantInitDict(node, n); } } } /* like sscanf(in+2, "%4x", &val) but less slow */ static bool decode_hex_string(char const* in, unsigned int* setme) { TR_ASSERT(in != nullptr); unsigned int val = 0; char const* const end = in + 6; TR_ASSERT(in[0] == '\\'); TR_ASSERT(in[1] == 'u'); in += 2; do { val <<= 4; if ('0' <= *in && *in <= '9') { val += *in - '0'; } else if ('a' <= *in && *in <= 'f') { val += *in - 'a' + 10U; } else if ('A' <= *in && *in <= 'F') { val += *in - 'A' + 10U; } else { return false; } } while (++in != end); *setme = val; return true; } static std::string_view extract_escaped_string(char const* in, size_t in_len, std::string& buf) { char const* const in_end = in + in_len; buf.clear(); while (in < in_end) { bool unescaped = false; if (*in == '\\' && in_end - in >= 2) { switch (in[1]) { case 'b': buf.push_back('\b'); in += 2; unescaped = true; break; case 'f': buf.push_back('\f'); in += 2; unescaped = true; break; case 'n': buf.push_back('\n'); in += 2; unescaped = true; break; case 'r': buf.push_back('\r'); in += 2; unescaped = true; break; case 't': buf.push_back('\t'); in += 2; unescaped = true; break; case '/': buf.push_back('/'); in += 2; unescaped = true; break; case '"': buf.push_back('"'); in += 2; unescaped = true; break; case '\\': buf.push_back('\\'); in += 2; unescaped = true; break; case 'u': if (in_end - in >= 6) { unsigned int val = 0; if (decode_hex_string(in, &val)) { try { auto buf8 = std::array{}; auto const it = utf8::append(val, std::data(buf8)); buf.append(std::data(buf8), it - std::data(buf8)); } catch (utf8::exception const&) // invalid codepoint { buf.push_back('?'); } unescaped = true; in += 6; break; } } } } if (!unescaped) { buf.push_back(*in); ++in; } } return buf; } static std::pair extract_string(jsonsl_t jsn, struct jsonsl_state_st* state, std::string& buf) { // figure out where the string is char const* in_begin = jsn->base + state->pos_begin; if (*in_begin == '"') { in_begin++; } char const* const in_end = jsn->base + state->pos_cur; size_t const in_len = in_end - in_begin; if (memchr(in_begin, '\\', in_len) == nullptr) { /* it's not escaped */ return std::make_pair(std::string_view{ in_begin, in_len }, true); } return std::make_pair(extract_escaped_string(in_begin, in_len, buf), false); } static void action_callback_POP( jsonsl_t jsn, jsonsl_action_t /*action*/, struct jsonsl_state_st* state, jsonsl_char_t const* /*buf*/) { auto* data = static_cast(jsn->data); if (state->type == JSONSL_T_STRING) { auto const [str, inplace] = extract_string(jsn, state, data->strbuf); if (inplace && ((data->parse_opts & TR_VARIANT_PARSE_INPLACE) != 0)) { tr_variantInitStrView(get_node(jsn), str); } else { tr_variantInitStr(get_node(jsn), str); } data->has_content = true; } else if (state->type == JSONSL_T_HKEY) { data->has_content = true; auto const [key, inplace] = extract_string(jsn, state, data->keybuf); data->key = key; } else if (state->type == JSONSL_T_LIST || state->type == JSONSL_T_OBJECT) { auto const depth = std::size(data->stack); auto const* const v = data->stack.back(); data->stack.pop_back(); if (depth < MaxDepth) { data->preallocGuess[depth] = v->val.l.count; } } else if (state->type == JSONSL_T_SPECIAL) { if ((state->special_flags & JSONSL_SPECIALf_NUMNOINT) != 0) { auto sv = std::string_view{ jsn->base + state->pos_begin, jsn->pos - state->pos_begin }; tr_variantInitReal(get_node(jsn), tr_parseNum(sv).value_or(0.0)); } else if ((state->special_flags & JSONSL_SPECIALf_NUMERIC) != 0) { char const* begin = jsn->base + state->pos_begin; data->has_content = true; tr_variantInitInt(get_node(jsn), std::strtoll(begin, nullptr, 10)); } else if ((state->special_flags & JSONSL_SPECIALf_BOOLEAN) != 0) { bool const b = (state->special_flags & JSONSL_SPECIALf_TRUE) != 0; data->has_content = true; tr_variantInitBool(get_node(jsn), b); } else if ((state->special_flags & JSONSL_SPECIALf_NULL) != 0) { data->has_content = true; tr_variantInitQuark(get_node(jsn), TR_KEY_NONE); } } } bool tr_variantParseJson(tr_variant& setme, int parse_opts, std::string_view json, char const** setme_end, tr_error** error) { TR_ASSERT((parse_opts & TR_VARIANT_PARSE_JSON) != 0); auto data = json_wrapper_data{}; jsonsl_t jsn = jsonsl_new(MaxDepth); jsn->action_callback_PUSH = action_callback_PUSH; jsn->action_callback_POP = action_callback_POP; jsn->error_callback = error_callback; jsn->data = &data; jsonsl_enable_all_callbacks(jsn); data.error = nullptr; data.size = std::size(json); data.has_content = false; data.key = ""sv; data.parse_opts = parse_opts; data.preallocGuess = {}; data.stack = {}; data.top = &setme; /* parse it */ jsonsl_feed(jsn, static_cast(std::data(json)), std::size(json)); /* EINVAL if there was no content */ if (data.error == nullptr && !data.has_content) { tr_error_set(&data.error, EINVAL, "No content"); } /* maybe set the end ptr */ if (setme_end != nullptr) { *setme_end = std::data(json) + jsn->pos; } /* cleanup */ auto const success = data.error == nullptr; if (data.error != nullptr) { tr_error_propagate(error, &data.error); } jsonsl_destroy(jsn); return success; } /**** ***** ****/ struct ParentState { int variantType; int childIndex; int childCount; }; struct JsonWalk { explicit JsonWalk(bool do_indent) : doIndent{ do_indent } { } std::deque parents; Buffer out; bool doIndent; }; static void jsonIndent(struct JsonWalk* data) { static auto buf = std::array{}; if (buf.front() == '\0') { memset(std::data(buf), ' ', std::size(buf)); buf[0] = '\n'; } if (data->doIndent) { data->out.add(std::data(buf), std::size(data->parents) * 4 + 1); } } static void jsonChildFunc(struct JsonWalk* data) { if (!std::empty(data->parents)) { auto& pstate = data->parents.back(); switch (pstate.variantType) { case TR_VARIANT_TYPE_DICT: { int const i = pstate.childIndex; ++pstate.childIndex; if (i % 2 == 0) { data->out.add(data->doIndent ? ": "sv : ":"sv); } else { bool const is_last = pstate.childIndex == pstate.childCount; if (!is_last) { data->out.push_back(','); jsonIndent(data); } } break; } case TR_VARIANT_TYPE_LIST: ++pstate.childIndex; if (bool const is_last = pstate.childIndex == pstate.childCount; !is_last) { data->out.push_back(','); jsonIndent(data); } break; default: break; } } } static void jsonPushParent(struct JsonWalk* data, tr_variant const* v) { int const n_children = tr_variantIsDict(v) ? v->val.l.count * 2 : v->val.l.count; data->parents.push_back({ v->type, 0, n_children }); } static void jsonPopParent(struct JsonWalk* data) { data->parents.pop_back(); } static void jsonIntFunc(tr_variant const* val, void* vdata) { auto buf = std::array{}; auto const* const out = fmt::format_to(std::data(buf), FMT_COMPILE("{:d}"), val->val.i); auto* const data = static_cast(vdata); data->out.add(std::data(buf), static_cast(out - std::data(buf))); jsonChildFunc(data); } static void jsonBoolFunc(tr_variant const* val, void* vdata) { auto* data = static_cast(vdata); if (val->val.b) { data->out.add("true"sv); } else { data->out.add("false"sv); } jsonChildFunc(data); } static void jsonRealFunc(tr_variant const* val, void* vdata) { auto* data = static_cast(vdata); if (fabs(val->val.d - (int)val->val.d) < 0.00001) { auto buf = std::array{}; auto const* const out = fmt::format_to(std::data(buf), FMT_COMPILE("{:.0f}"), val->val.d); data->out.add(std::data(buf), static_cast(out - std::data(buf))); } else { auto buf = std::array{}; auto const* const out = fmt::format_to(std::data(buf), FMT_COMPILE("{:.4f}"), val->val.d); data->out.add(std::data(buf), static_cast(out - std::data(buf))); } jsonChildFunc(data); } static void jsonStringFunc(tr_variant const* val, void* vdata) { auto* data = static_cast(vdata); auto sv = std::string_view{}; (void)!tr_variantGetStrView(val, &sv); auto& out = data->out; out.reserve(std::size(data->out) + std::size(sv) * 6 + 2); out.push_back('"'); for (; !std::empty(sv); sv.remove_prefix(1)) { switch (sv.front()) { case '\b': out.add(R"(\b)"sv); break; case '\f': out.add(R"(\f)"sv); break; case '\n': out.add(R"(\n)"sv); break; case '\r': out.add(R"(\r)"sv); break; case '\t': out.add(R"(\t)"sv); break; case '"': out.add(R"(\")"sv); break; case '\\': out.add(R"(\\)"sv); break; default: if (isprint((unsigned char)sv.front()) != 0) { out.push_back(sv.front()); } else { try { auto arr = std::array{}; auto const* const begin8 = std::data(sv); auto const* const end8 = begin8 + std::size(sv); auto const* walk8 = begin8; auto const uch32 = utf8::next(walk8, end8); auto const result = fmt::format_to_n(std::data(arr), std::size(arr), FMT_COMPILE("\\u{:04x}"), uch32); out.add(std::data(arr), result.size); sv.remove_prefix(walk8 - begin8 - 1); } catch (utf8::exception const&) { out.push_back('?'); } } break; } } out.push_back('"'); jsonChildFunc(data); } static void jsonDictBeginFunc(tr_variant const* val, void* vdata) { auto* data = static_cast(vdata); jsonPushParent(data, val); data->out.push_back('{'); if (val->val.l.count != 0) { jsonIndent(data); } } static void jsonListBeginFunc(tr_variant const* val, void* vdata) { size_t const n_children = tr_variantListSize(val); auto* data = static_cast(vdata); jsonPushParent(data, val); data->out.push_back('['); if (n_children != 0) { jsonIndent(data); } } static void jsonContainerEndFunc(tr_variant const* val, void* vdata) { auto* data = static_cast(vdata); jsonPopParent(data); jsonIndent(data); if (tr_variantIsDict(val)) { data->out.push_back('}'); } else /* list */ { data->out.push_back(']'); } jsonChildFunc(data); } static struct VariantWalkFuncs const walk_funcs = { jsonIntFunc, // jsonBoolFunc, // jsonRealFunc, // jsonStringFunc, // jsonDictBeginFunc, // jsonListBeginFunc, // jsonContainerEndFunc, // }; std::string tr_variantToStrJson(tr_variant const* top, bool lean) { auto data = JsonWalk{ !lean }; tr_variantWalk(top, &walk_funcs, &data, true); auto& buf = data.out; if (!std::empty(buf)) { buf.push_back('\n'); } return buf.toString(); }