summaryrefslogtreecommitdiff
path: root/libs/assimp/code/AssetLib/Step/STEPFile.h
diff options
context:
space:
mode:
authorsanine <sanine.not@pm.me>2022-04-16 11:55:09 -0500
committersanine <sanine.not@pm.me>2022-04-16 11:55:09 -0500
commitdb81b925d776103326128bf629cbdda576a223e7 (patch)
tree58bea8155c686733310009f6bed7363f91fbeb9d /libs/assimp/code/AssetLib/Step/STEPFile.h
parent55860037b14fb3893ba21cf2654c83d349cc1082 (diff)
move 3rd-party librarys into libs/ and add built-in honeysuckle
Diffstat (limited to 'libs/assimp/code/AssetLib/Step/STEPFile.h')
-rw-r--r--libs/assimp/code/AssetLib/Step/STEPFile.h971
1 files changed, 971 insertions, 0 deletions
diff --git a/libs/assimp/code/AssetLib/Step/STEPFile.h b/libs/assimp/code/AssetLib/Step/STEPFile.h
new file mode 100644
index 0000000..e09faad
--- /dev/null
+++ b/libs/assimp/code/AssetLib/Step/STEPFile.h
@@ -0,0 +1,971 @@
+/*
+Open Asset Import Library (assimp)
+----------------------------------------------------------------------
+
+Copyright (c) 2006-2020, assimp team
+
+
+All rights reserved.
+
+Redistribution and use of this software in source and binary forms,
+with or without modification, are permitted provided that the
+following conditions are met:
+
+* Redistributions of source code must retain the above
+ copyright notice, this list of conditions and the
+ following disclaimer.
+
+* Redistributions in binary form must reproduce the above
+ copyright notice, this list of conditions and the
+ following disclaimer in the documentation and/or other
+ materials provided with the distribution.
+
+* Neither the name of the assimp team, nor the names of its
+ contributors may be used to endorse or promote products
+ derived from this software without specific prior
+ written permission of the assimp team.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+----------------------------------------------------------------------
+*/
+
+#ifndef INCLUDED_AI_STEPFILE_H
+#define INCLUDED_AI_STEPFILE_H
+
+#include <bitset>
+#include <map>
+#include <memory>
+#include <set>
+#include <typeinfo>
+#include <vector>
+
+#include "AssetLib/FBX/FBXDocument.h" //ObjectMap::value_type
+
+#include <assimp/DefaultLogger.hpp>
+
+#ifdef _MSC_VER
+# pragma warning(push)
+# pragma warning(disable : 4127 4456 4245 4512 )
+#endif // _MSC_VER
+
+//
+#if _MSC_VER > 1500 || (defined __GNUC___)
+# define ASSIMP_STEP_USE_UNORDERED_MULTIMAP
+#else
+# define step_unordered_map map
+# define step_unordered_multimap multimap
+#endif
+
+#ifdef ASSIMP_STEP_USE_UNORDERED_MULTIMAP
+# include <unordered_map>
+# if defined(_MSC_VER) && _MSC_VER <= 1600
+# define step_unordered_map tr1::unordered_map
+# define step_unordered_multimap tr1::unordered_multimap
+# else
+# define step_unordered_map unordered_map
+# define step_unordered_multimap unordered_multimap
+# endif
+#endif
+
+#include <assimp/LineSplitter.h>
+
+// uncomment this to have the loader evaluate all entities upon loading.
+// this is intended as stress test - by default, entities are evaluated
+// lazily and therefore not unless needed.
+
+//#define ASSIMP_IFC_TEST
+
+namespace Assimp {
+
+// ********************************************************************************
+// before things get complicated, this is the basic outline:
+
+namespace STEP {
+
+namespace EXPRESS {
+
+// base data types known by EXPRESS schemata - any custom data types will derive one of those
+class DataType;
+class UNSET; /*: public DataType */
+class ISDERIVED; /*: public DataType */
+// class REAL; /*: public DataType */
+class ENUM; /*: public DataType */
+// class STRING; /*: public DataType */
+// class INTEGER; /*: public DataType */
+class ENTITY; /*: public DataType */
+class LIST; /*: public DataType */
+// class SELECT; /*: public DataType */
+
+// a conversion schema is not exactly an EXPRESS schema, rather it
+// is a list of pointers to conversion functions to build up the
+// object tree from an input file.
+class ConversionSchema;
+} // namespace EXPRESS
+
+struct HeaderInfo;
+class Object;
+class LazyObject;
+class DB;
+
+typedef Object *(*ConvertObjectProc)(const DB &db, const EXPRESS::LIST &params);
+} // namespace STEP
+
+// ********************************************************************************
+
+namespace STEP {
+
+// -------------------------------------------------------------------------------
+/** Exception class used by the STEP loading & parsing code. It is typically
+ * coupled with a line number. */
+// -------------------------------------------------------------------------------
+struct SyntaxError : DeadlyImportError {
+ enum : uint64_t {
+ LINE_NOT_SPECIFIED = 0xfffffffffffffffLL
+ };
+
+ SyntaxError(const std::string &s, uint64_t line = LINE_NOT_SPECIFIED);
+};
+
+// -------------------------------------------------------------------------------
+/** Exception class used by the STEP loading & parsing code when a type
+ * error (i.e. an entity expects a string but receives a bool) occurs.
+ * It is typically coupled with both an entity id and a line number.*/
+// -------------------------------------------------------------------------------
+struct TypeError : DeadlyImportError {
+ enum : uint64_t {
+ ENTITY_NOT_SPECIFIED = 0xffffffffffffffffUL,
+ ENTITY_NOT_SPECIFIED_32 = 0x00000000ffffffff
+ };
+
+ TypeError(const std::string &s, uint64_t entity = ENTITY_NOT_SPECIFIED, uint64_t line = SyntaxError::LINE_NOT_SPECIFIED);
+};
+
+// hack to make a given member template-dependent
+template <typename T, typename T2>
+T2 &Couple(T2 &in) {
+ return in;
+}
+
+namespace EXPRESS {
+
+// -------------------------------------------------------------------------------
+//** Base class for all STEP data types */
+// -------------------------------------------------------------------------------
+class DataType {
+public:
+ typedef std::shared_ptr<const DataType> Out;
+
+public:
+ virtual ~DataType() {
+ }
+
+public:
+ template <typename T>
+ const T &To() const {
+ return dynamic_cast<const T &>(*this);
+ }
+
+ template <typename T>
+ T &To() {
+ return dynamic_cast<T &>(*this);
+ }
+
+ template <typename T>
+ const T *ToPtr() const {
+ return dynamic_cast<const T *>(this);
+ }
+
+ template <typename T>
+ T *ToPtr() {
+ return dynamic_cast<T *>(this);
+ }
+
+ // utilities to deal with SELECT entities, which currently lack automatic
+ // conversion support.
+ template <typename T>
+ const T &ResolveSelect(const DB &db) const {
+ return Couple<T>(db).MustGetObject(To<EXPRESS::ENTITY>())->template To<T>();
+ }
+
+ template <typename T>
+ const T *ResolveSelectPtr(const DB &db) const {
+ const EXPRESS::ENTITY *e = ToPtr<EXPRESS::ENTITY>();
+ return e ? Couple<T>(db).MustGetObject(*e)->template ToPtr<T>() : (const T *)0;
+ }
+
+public:
+ /** parse a variable from a string and set 'inout' to the character
+ * behind the last consumed character. An optional schema enables,
+ * if specified, automatic conversion of custom data types.
+ *
+ * @throw SyntaxError
+ */
+ static std::shared_ptr<const EXPRESS::DataType> Parse(const char *&inout,
+ uint64_t line = SyntaxError::LINE_NOT_SPECIFIED,
+ const EXPRESS::ConversionSchema *schema = NULL);
+
+public:
+};
+
+typedef DataType SELECT;
+typedef DataType LOGICAL;
+
+// -------------------------------------------------------------------------------
+/** Sentinel class to represent explicitly unset (optional) fields ($) */
+// -------------------------------------------------------------------------------
+class UNSET : public DataType {
+public:
+private:
+};
+
+// -------------------------------------------------------------------------------
+/** Sentinel class to represent explicitly derived fields (*) */
+// -------------------------------------------------------------------------------
+class ISDERIVED : public DataType {
+public:
+private:
+};
+
+// -------------------------------------------------------------------------------
+/** Shared implementation for some of the primitive data type, i.e. int, float */
+// -------------------------------------------------------------------------------
+template <typename T>
+class PrimitiveDataType : public DataType {
+public:
+ // This is the type that will cd ultimatively be used to
+ // expose this data type to the user.
+ typedef T Out;
+
+ PrimitiveDataType() {}
+ PrimitiveDataType(const T &val) :
+ val(val) {}
+
+ PrimitiveDataType(const PrimitiveDataType &o) {
+ (*this) = o;
+ }
+
+ operator const T &() const {
+ return val;
+ }
+
+ PrimitiveDataType &operator=(const PrimitiveDataType &o) {
+ val = o.val;
+ return *this;
+ }
+
+protected:
+ T val;
+};
+
+typedef PrimitiveDataType<int64_t> INTEGER;
+typedef PrimitiveDataType<double> REAL;
+typedef PrimitiveDataType<double> NUMBER;
+typedef PrimitiveDataType<std::string> STRING;
+
+// -------------------------------------------------------------------------------
+/** Generic base class for all enumerated types */
+// -------------------------------------------------------------------------------
+class ENUMERATION : public STRING {
+public:
+ ENUMERATION(const std::string &val) :
+ STRING(val) {}
+
+private:
+};
+
+typedef ENUMERATION BOOLEAN;
+
+// -------------------------------------------------------------------------------
+/** This is just a reference to an entity/object somewhere else */
+// -------------------------------------------------------------------------------
+class ENTITY : public PrimitiveDataType<uint64_t> {
+public:
+ ENTITY(uint64_t val) :
+ PrimitiveDataType<uint64_t>(val) {
+ ai_assert(val != 0);
+ }
+
+ ENTITY() :
+ PrimitiveDataType<uint64_t>(TypeError::ENTITY_NOT_SPECIFIED) {
+ // empty
+ }
+
+private:
+};
+
+// -------------------------------------------------------------------------------
+/** Wrap any STEP aggregate: LIST, SET, ... */
+// -------------------------------------------------------------------------------
+class LIST : public DataType {
+public:
+ // access a particular list index, throw std::range_error for wrong indices
+ std::shared_ptr<const DataType> operator[](size_t index) const {
+ return members[index];
+ }
+
+ size_t GetSize() const {
+ return members.size();
+ }
+
+public:
+ /** @see DaraType::Parse */
+ static std::shared_ptr<const EXPRESS::LIST> Parse(const char *&inout,
+ uint64_t line = SyntaxError::LINE_NOT_SPECIFIED,
+ const EXPRESS::ConversionSchema *schema = NULL);
+
+private:
+ typedef std::vector<std::shared_ptr<const DataType>> MemberList;
+ MemberList members;
+};
+
+class BINARY : public PrimitiveDataType<uint32_t> {
+public:
+ BINARY(uint32_t val) :
+ PrimitiveDataType<uint32_t>(val) {
+ // empty
+ }
+
+ BINARY() :
+ PrimitiveDataType<uint32_t>(TypeError::ENTITY_NOT_SPECIFIED_32) {
+ // empty
+ }
+};
+
+// -------------------------------------------------------------------------------
+/* Not exactly a full EXPRESS schema but rather a list of conversion functions
+ * to extract valid C++ objects out of a STEP file. Those conversion functions
+ * may, however, perform further schema validations. */
+// -------------------------------------------------------------------------------
+class ConversionSchema {
+public:
+ struct SchemaEntry {
+ SchemaEntry(const char *name, ConvertObjectProc func) :
+ mName(name), mFunc(func) {
+ // empty
+ }
+
+ const char *mName;
+ ConvertObjectProc mFunc;
+ };
+
+ typedef std::map<std::string, ConvertObjectProc> ConverterMap;
+
+ template <size_t N>
+ explicit ConversionSchema(const SchemaEntry (&schemas)[N]) {
+ *this = schemas;
+ }
+
+ ConversionSchema() {
+ }
+
+ ConvertObjectProc GetConverterProc(const std::string &name) const {
+ ConverterMap::const_iterator it = converters.find(name);
+ return it == converters.end() ? nullptr : (*it).second;
+ }
+
+ bool IsKnownToken(const std::string &name) const {
+ return converters.find(name) != converters.end();
+ }
+
+ const char *GetStaticStringForToken(const std::string &token) const {
+ ConverterMap::const_iterator it = converters.find(token);
+ return it == converters.end() ? nullptr : (*it).first.c_str();
+ }
+
+ template <size_t N>
+ const ConversionSchema &operator=(const SchemaEntry (&schemas)[N]) {
+ for (size_t i = 0; i < N; ++i) {
+ const SchemaEntry &schema = schemas[i];
+ converters[schema.mName] = schema.mFunc;
+ }
+ return *this;
+ }
+
+private:
+ ConverterMap converters;
+};
+} // namespace EXPRESS
+
+// ------------------------------------------------------------------------------
+/** Bundle all the relevant info from a STEP header, parts of which may later
+ * be plainly dumped to the logfile, whereas others may help the caller pick an
+ * appropriate loading strategy.*/
+// ------------------------------------------------------------------------------
+struct HeaderInfo {
+ std::string timestamp;
+ std::string app;
+ std::string fileSchema;
+};
+
+// ------------------------------------------------------------------------------
+/** Base class for all concrete object instances */
+// ------------------------------------------------------------------------------
+class Object {
+public:
+ Object(const char *classname = "unknown") :
+ id(0), classname(classname) {
+ // empty
+ }
+
+ virtual ~Object() {
+ // empty
+ }
+
+ // utilities to simplify casting to concrete types
+ template <typename T>
+ const T &To() const {
+ return dynamic_cast<const T &>(*this);
+ }
+
+ template <typename T>
+ T &To() {
+ return dynamic_cast<T &>(*this);
+ }
+
+ template <typename T>
+ const T *ToPtr() const {
+ return dynamic_cast<const T *>(this);
+ }
+
+ template <typename T>
+ T *ToPtr() {
+ return dynamic_cast<T *>(this);
+ }
+
+ uint64_t GetID() const {
+ return id;
+ }
+
+ std::string GetClassName() const {
+ return classname;
+ }
+
+ void SetID(uint64_t newval) {
+ id = newval;
+ }
+
+private:
+ uint64_t id;
+ const char *const classname;
+};
+
+template <typename T>
+size_t GenericFill(const STEP::DB &db, const EXPRESS::LIST &params, T *in);
+// (intentionally undefined)
+
+// ------------------------------------------------------------------------------
+/** CRTP shared base class for use by concrete entity implementation classes */
+// ------------------------------------------------------------------------------
+template <typename TDerived, size_t arg_count>
+struct ObjectHelper : virtual Object {
+ ObjectHelper() :
+ aux_is_derived(0) {
+ // empty
+ }
+
+ static Object *Construct(const STEP::DB &db, const EXPRESS::LIST &params) {
+ // make sure we don't leak if Fill() throws an exception
+ std::unique_ptr<TDerived> impl(new TDerived());
+
+ // GenericFill<T> is undefined so we need to have a specialization
+ const size_t num_args = GenericFill<TDerived>(db, params, &*impl);
+ (void)num_args;
+
+ // the following check is commented because it will always trigger if
+ // parts of the entities are generated with dummy wrapper code.
+ // This is currently done to reduce the size of the loader
+ // code.
+ //if (num_args != params.GetSize() && impl->GetClassName() != "NotImplemented") {
+ // DefaultLogger::get()->debug("STEP: not all parameters consumed");
+ //}
+ return impl.release();
+ }
+
+ // note that this member always exists multiple times within the hierarchy
+ // of an individual object, so any access to it must be disambiguated.
+ std::bitset<arg_count> aux_is_derived;
+};
+
+// ------------------------------------------------------------------------------
+/** Class template used to represent OPTIONAL data members in the converted schema */
+// ------------------------------------------------------------------------------
+template <typename T>
+struct Maybe {
+ Maybe() :
+ have() {
+ // empty
+ }
+
+ explicit Maybe(const T &ptr) :
+ ptr(ptr), have(true) {
+ // empty
+ }
+
+ void flag_invalid() {
+ have = false;
+ }
+
+ void flag_valid() {
+ have = true;
+ }
+
+ bool operator!() const {
+ return !have;
+ }
+
+ operator bool() const {
+ return have;
+ }
+
+ operator const T &() const {
+ return Get();
+ }
+
+ const T &Get() const {
+ ai_assert(have);
+ return ptr;
+ }
+
+ Maybe &operator=(const T &_ptr) {
+ ptr = _ptr;
+ have = true;
+ return *this;
+ }
+
+private:
+ template <typename T2>
+ friend struct InternGenericConvert;
+
+ operator T &() {
+ return ptr;
+ }
+
+ T ptr;
+ bool have;
+};
+
+// ------------------------------------------------------------------------------
+/** A LazyObject is created when needed. Before this happens, we just keep
+ the text line that contains the object definition. */
+// -------------------------------------------------------------------------------
+class LazyObject {
+ friend class DB;
+
+public:
+ LazyObject(DB &db, uint64_t id, uint64_t line, const char *type, const char *args);
+ ~LazyObject();
+
+ Object &operator*() {
+ if (!obj) {
+ LazyInit();
+ ai_assert(obj);
+ }
+ return *obj;
+ }
+
+ const Object &operator*() const {
+ if (!obj) {
+ LazyInit();
+ ai_assert(obj);
+ }
+ return *obj;
+ }
+
+ template <typename T>
+ const T &To() const {
+ return dynamic_cast<const T &>(**this);
+ }
+
+ template <typename T>
+ T &To() {
+ return dynamic_cast<T &>(**this);
+ }
+
+ template <typename T>
+ const T *ToPtr() const {
+ return dynamic_cast<const T *>(&**this);
+ }
+
+ template <typename T>
+ T *ToPtr() {
+ return dynamic_cast<T *>(&**this);
+ }
+
+ Object *operator->() {
+ return &**this;
+ }
+
+ const Object *operator->() const {
+ return &**this;
+ }
+
+ bool operator==(const std::string &atype) const {
+ return type == atype;
+ }
+
+ bool operator!=(const std::string &atype) const {
+ return type != atype;
+ }
+
+ uint64_t GetID() const {
+ return id;
+ }
+
+private:
+ void LazyInit() const;
+
+private:
+ mutable uint64_t id;
+ const char *const type;
+ DB &db;
+ mutable const char *args;
+ mutable Object *obj;
+};
+
+template <typename T>
+inline bool operator==(const std::shared_ptr<LazyObject> &lo, T whatever) {
+ return *lo == whatever; // XXX use std::forward if we have 0x
+}
+
+template <typename T>
+inline bool operator==(const std::pair<uint64_t, std::shared_ptr<LazyObject>> &lo, T whatever) {
+ return *(lo.second) == whatever; // XXX use std::forward if we have 0x
+}
+
+// ------------------------------------------------------------------------------
+/** Class template used to represent lazily evaluated object references in the converted schema */
+// ------------------------------------------------------------------------------
+template <typename T>
+struct Lazy {
+ typedef Lazy Out;
+ Lazy(const LazyObject *obj = nullptr) :
+ obj(obj) {
+ // empty
+ }
+
+ operator const T *() const {
+ return obj->ToPtr<T>();
+ }
+
+ operator const T &() const {
+ return obj->To<T>();
+ }
+
+ const T &operator*() const {
+ return obj->To<T>();
+ }
+
+ const T *operator->() const {
+ return &obj->To<T>();
+ }
+
+ const LazyObject *obj;
+};
+
+// ------------------------------------------------------------------------------
+/** Class template used to represent LIST and SET data members in the converted schema */
+// ------------------------------------------------------------------------------
+template <typename T, uint64_t min_cnt, uint64_t max_cnt = 0uL>
+struct ListOf : public std::vector<typename T::Out> {
+ typedef typename T::Out OutScalar;
+ typedef ListOf Out;
+
+ ListOf() {
+ static_assert(min_cnt <= max_cnt || !max_cnt, "min_cnt <= max_cnt || !max_cnt");
+ }
+};
+
+// ------------------------------------------------------------------------------
+template <typename TOut>
+struct PickBaseType {
+ typedef EXPRESS::PrimitiveDataType<TOut> Type;
+};
+
+template <typename TOut>
+struct PickBaseType<Lazy<TOut>> {
+ typedef EXPRESS::ENTITY Type;
+};
+
+template <>
+struct PickBaseType<std::shared_ptr<const EXPRESS::DataType>>;
+
+// ------------------------------------------------------------------------------
+template <typename T>
+struct InternGenericConvert {
+ void operator()(T &out, const std::shared_ptr<const EXPRESS::DataType> &in, const STEP::DB & /*db*/) {
+ try {
+ out = dynamic_cast<const typename PickBaseType<T>::Type &>(*in);
+ } catch (std::bad_cast &) {
+ throw TypeError("type error reading literal field");
+ }
+ }
+};
+
+template <>
+struct InternGenericConvert<std::shared_ptr<const EXPRESS::DataType>> {
+ void operator()(std::shared_ptr<const EXPRESS::DataType> &out, const std::shared_ptr<const EXPRESS::DataType> &in, const STEP::DB & /*db*/) {
+ out = in;
+ }
+};
+
+template <typename T>
+struct InternGenericConvert<Maybe<T>> {
+ void operator()(Maybe<T> &out, const std::shared_ptr<const EXPRESS::DataType> &in, const STEP::DB &db) {
+ GenericConvert((T &)out, in, db);
+ out.flag_valid();
+ }
+};
+
+#if _MSC_VER > 1920
+#pragma warning(push)
+#pragma warning(disable : 4127)
+#endif // _WIN32
+
+template <typename T, uint64_t min_cnt, uint64_t max_cnt>
+struct InternGenericConvertList {
+ void operator()(ListOf<T, min_cnt, max_cnt> &out, const std::shared_ptr<const EXPRESS::DataType> &inp_base, const STEP::DB &db) {
+
+ const EXPRESS::LIST *inp = dynamic_cast<const EXPRESS::LIST *>(inp_base.get());
+ if (!inp) {
+ throw TypeError("type error reading aggregate");
+ }
+
+ // XXX is this really how the EXPRESS notation ([?:3],[1:3]) is intended?
+ const size_t len = inp->GetSize();
+ if (0 != max_cnt && len > max_cnt) {
+ ASSIMP_LOG_WARN("too many aggregate elements");
+ } else if (len < min_cnt) {
+ ASSIMP_LOG_WARN("too few aggregate elements");
+ }
+
+ out.reserve(inp->GetSize());
+ for (size_t i = 0; i < inp->GetSize(); ++i) {
+
+ out.push_back(typename ListOf<T, min_cnt, max_cnt>::OutScalar());
+ try {
+ GenericConvert(out.back(), (*inp)[i], db);
+ } catch (const TypeError &t) {
+ throw TypeError(t.what() + std::string(" of aggregate"));
+ }
+ }
+ }
+};
+
+template <typename T>
+struct InternGenericConvert<Lazy<T>> {
+ void operator()(Lazy<T> &out, const std::shared_ptr<const EXPRESS::DataType> &in_base, const STEP::DB &db) {
+ const EXPRESS::ENTITY *in = dynamic_cast<const EXPRESS::ENTITY *>(in_base.get());
+ if (!in) {
+ throw TypeError("type error reading entity");
+ }
+ out = Couple<T>(db).GetObject(*in);
+ }
+};
+
+template <typename T1>
+inline void GenericConvert(T1 &a, const std::shared_ptr<const EXPRESS::DataType> &b, const STEP::DB &db) {
+ return InternGenericConvert<T1>()(a, b, db);
+}
+
+template <typename T1, uint64_t N1, uint64_t N2>
+inline void GenericConvert(ListOf<T1, N1, N2> &a, const std::shared_ptr<const EXPRESS::DataType> &b, const STEP::DB &db) {
+ return InternGenericConvertList<T1, N1, N2>()(a, b, db);
+}
+
+// ------------------------------------------------------------------------------
+/** Lightweight manager class that holds the map of all objects in a
+ * STEP file. DB's are exclusively maintained by the functions in
+ * STEPFileReader.h*/
+// -------------------------------------------------------------------------------
+class DB {
+ friend DB *ReadFileHeader(std::shared_ptr<IOStream> stream);
+ friend void ReadFile(DB &db, const EXPRESS::ConversionSchema &scheme,
+ const char *const *types_to_track, size_t len,
+ const char *const *inverse_indices_to_track, size_t len2);
+
+ friend class LazyObject;
+
+public:
+ // objects indexed by ID - this can grow pretty large (i.e some hundred million
+ // entries), so use raw pointers to avoid *any* overhead.
+ typedef std::map<uint64_t, const LazyObject *> ObjectMap;
+
+ // objects indexed by their declarative type, but only for those that we truly want
+ typedef std::set<const LazyObject *> ObjectSet;
+ typedef std::map<std::string, ObjectSet> ObjectMapByType;
+
+ // list of types for which to keep inverse indices for all references
+ // that the respective objects keep.
+ // the list keeps pointers to strings in static storage
+ typedef std::set<const char *> InverseWhitelist;
+
+ // references - for each object id the ids of all objects which reference it
+ // this is used to simulate STEP inverse indices for selected types.
+ typedef std::step_unordered_multimap<uint64_t, uint64_t> RefMap;
+ typedef std::pair<RefMap::const_iterator, RefMap::const_iterator> RefMapRange;
+
+private:
+ DB(const std::shared_ptr<StreamReaderLE> &reader) :
+ reader(reader), splitter(*reader, true, true), evaluated_count(), schema(nullptr) {}
+
+public:
+ ~DB() {
+ for (ObjectMap::value_type &o : objects) {
+ delete o.second;
+ }
+ }
+
+ uint64_t GetObjectCount() const {
+ return objects.size();
+ }
+
+ uint64_t GetEvaluatedObjectCount() const {
+ return evaluated_count;
+ }
+
+ const HeaderInfo &GetHeader() const {
+ return header;
+ }
+
+ const EXPRESS::ConversionSchema &GetSchema() const {
+ return *schema;
+ }
+
+ const ObjectMap &GetObjects() const {
+ return objects;
+ }
+
+ const ObjectMapByType &GetObjectsByType() const {
+ return objects_bytype;
+ }
+
+ const RefMap &GetRefs() const {
+ return refs;
+ }
+
+ bool KeepInverseIndicesForType(const char *const type) const {
+ return inv_whitelist.find(type) != inv_whitelist.end();
+ }
+
+ // get the yet unevaluated object record with a given id
+ const LazyObject *GetObject(uint64_t id) const {
+ const ObjectMap::const_iterator it = objects.find(id);
+ if (it != objects.end()) {
+ return (*it).second;
+ }
+ return nullptr;
+ }
+
+ // get an arbitrary object out of the soup with the only restriction being its type.
+ const LazyObject *GetObject(const std::string &type) const {
+ const ObjectMapByType::const_iterator it = objects_bytype.find(type);
+ if (it != objects_bytype.end() && (*it).second.size()) {
+ return *(*it).second.begin();
+ }
+ return NULL;
+ }
+
+ // same, but raise an exception if the object doesn't exist and return a reference
+ const LazyObject &MustGetObject(uint64_t id) const {
+ const LazyObject *o = GetObject(id);
+ if (!o) {
+ throw TypeError("requested entity is not present", id);
+ }
+ return *o;
+ }
+
+ const LazyObject &MustGetObject(const std::string &type) const {
+ const LazyObject *o = GetObject(type);
+ if (!o) {
+ throw TypeError("requested entity of type " + type + "is not present");
+ }
+ return *o;
+ }
+
+#ifdef ASSIMP_IFC_TEST
+
+ // evaluate *all* entities in the file. this is a power test for the loader
+ void EvaluateAll() {
+ for (ObjectMap::value_type &e : objects) {
+ **e.second;
+ }
+ ai_assert(evaluated_count == objects.size());
+ }
+
+#endif
+
+private:
+ // full access only offered to close friends - they should
+ // use the provided getters rather than messing around with
+ // the members directly.
+ LineSplitter &GetSplitter() {
+ return splitter;
+ }
+
+ void InternInsert(const LazyObject *lz) {
+ objects[lz->GetID()] = lz;
+
+ const ObjectMapByType::iterator it = objects_bytype.find(lz->type);
+ if (it != objects_bytype.end()) {
+ (*it).second.insert(lz);
+ }
+ }
+
+ void SetSchema(const EXPRESS::ConversionSchema &_schema) {
+ schema = &_schema;
+ }
+
+ void SetTypesToTrack(const char *const *types, size_t N) {
+ for (size_t i = 0; i < N; ++i) {
+ objects_bytype[types[i]] = ObjectSet();
+ }
+ }
+
+ void SetInverseIndicesToTrack(const char *const *types, size_t N) {
+ for (size_t i = 0; i < N; ++i) {
+ const char *const sz = schema->GetStaticStringForToken(types[i]);
+ ai_assert(sz);
+ inv_whitelist.insert(sz);
+ }
+ }
+
+ HeaderInfo &GetHeader() {
+ return header;
+ }
+
+ void MarkRef(uint64_t who, uint64_t by_whom) {
+ refs.insert(std::make_pair(who, by_whom));
+ }
+
+private:
+ HeaderInfo header;
+ ObjectMap objects;
+ ObjectMapByType objects_bytype;
+ RefMap refs;
+ InverseWhitelist inv_whitelist;
+ std::shared_ptr<StreamReaderLE> reader;
+ LineSplitter splitter;
+ uint64_t evaluated_count;
+ const EXPRESS::ConversionSchema *schema;
+};
+
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif // _MSC_VER
+
+} // namespace STEP
+
+} // namespace Assimp
+
+#endif // INCLUDED_AI_STEPFILE_H