N4
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View N4 as PDF for free.
More details
- Words: 1,493
- Pages: 47
Introduction To .NET Metadata Matt Pietrek
Housekeeping Who is this guy? Co-founder, Mindreef LLC (www.mindreef.com) President, Wheaty Productions Inc. MSDN’s “Under the Hood” columnist
web: http://www.wheaty.net email: [email protected] Source availability
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
Metadata: The Big Picture Metadata is information stored in .NET assemblies that makes them self-describing Metadata describes: Classes (types) provided by the assembly Method and field information for classes Version information Dependencies on other assemblies and modules Security attributes required
Metadata provides the information that the .NET runtime needs to ensure type safety and security
Demo: Many Views Of Metadata
Metadata: Evolution Of A COM Concept COM IDL files describe interfaces, methods, parameters, and types in sufficient detail to marshal calls between processes COM Type Libraries are just a binary representation of IDL information Metadata is effectively type libraries on steroids Microsoft provides tools to convert between metadata and type libraries
Metadata Usage In .NET The .NET runtime relies heavily on metadata to link together calls and data references at runtime Metadata is stored in a binary format, and subject to change Microsoft provides .NET classes and classic COM interfaces to read and write metadata
The Advantage Of Metadata A shared standard of representing information between any .NET compatible language Cool benefit: Intellisense for any .NET component
Metadata is automatically generated and consumed by .NET compilers No need to create .IDL files to expose your code to others No need to mess around with .H files and import libraries
Emitting Metadata Metadata (and associated IL) can be created using the .NET runtime Compilers aren’t the only ones that get to play!
System.Reflection.Emit classes are the basis for dynamically creating code: AssemblyBuilder MethodBuilder ILGenerator Many more!
Focus in this presentation is consuming metadata
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
The Metadata Hierarchy I Assembly (e.g., mscorlib)
Collections Single Instances
Modules
Module (i.e., a DLL)
Classes Global Methods & Variables
The Metadata Hierarchy II Class (e.g., System.Object)
Methods
Fields
Method (e.g., “Foo”)
Parameters
Properties
Parameter “int Param1”
Events
Parameter “String Param2”
Assembly Metadata Name Public key Version # (major & minor) Build # Revision # Dependencies on other assemblies Locale
Class Metadata Contents Name Flags (CorTypeAttr) Visibility (Public, private, assembly, …) Layout (auto, sequential, explicit…) Type (class, interface, value type, …)
Base class (if applicable)
Method Metadata Contents Name Parent class Flags (CorMethodAttr) Visibility Static / Virtual Platform Invoke (interop to legacy code)
Implementation Flags (CorMethodImpl) IL / OptIL / Native Managed / Unmanaged implementation
Signature RVA to IL in executable file
Parameter Metadata Contents Name Parent method Order in method parameter list Attributes (CorParamAttr) In / Out Has default value Optional
Field Metadata Contents Name Parent class Attributes (CorFieldAttr) Visibility Static member
Signature (implicitly gives the size)
Property Metadata Contents Name Parent class Attributes (CorPropertyAttr) Signature (implicitly gives the size) Getter & Setter methods Default value
Event Metadata Contents Name Parent class Fired method AddOn / RemoveOn methods
Custom Attributes Custom attributes are special classes that provide additional information about a: Assembly / Module Class / Struct Method / Constructor / Property Enum Parameter / Field
Unmanaged and Reflection APIs for reading / creating custom attributes
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
About The Reflection API Full access to metadata using .NET runtime classes Assemblies, classes, methods, fields, etc… are all represented as .NET objects Layered over the unmanaged metadata interfaces
The Metadata Hierarchy As Seen By The Reflection API Assembly
ManifestResource
Module
Type
FieldInfo
PropertyInfo
MethodInfo
EventInfo
ParameterInfo
AssemblyName
Custom Attributes
Getting To The Reflection API There are a variety of entry points into the Reflection classes Assembly::LoadFrom loads a file based assembly Returns an Assembly *
For an existing .NET object, use the ::GetType method to get a Type* Type::Module property moves up the hierarchy Type::GetMethods, etc… moves down
Demo ReflectCS program
Example: Enumerating All Methods In An Assembly Use Assembly::LoadFrom to load a .NET assembly Returns an Assembly *
Assembly::GetModules returns an array of Module *’s For each Module *, call Module::GetTypes Returns an array of Type *’s
For each Type *, call Type::GetMethods Returns an array of MethodInfo *’s
For each MethodInfo *, the MethodInfo::Name property returns the method’s name
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
About The Unmanaged API Regular COM interfaces that don’t use the .NET runtime Relies heavily on “tokens” More granular than equivalent COM ITypeLib & ITypeInfo interfaces Information isn’t packed into structures
Declared in COR.H / CORHDR.H CLSIDs and IIDs defined in CORGUID.LIB
The Unmanaged Metadata Interfaces IMetaDataDispenser The “root” interface for obtaining all others Works on file or memory based metadata
IMetaDataImport The primary interface for reading class info
IMetaDataAssemblyImport IMetaDataEmit IMetaDataAssemblyEmit
Demo The Meta program
Obtaining an IMetaDataImport Interface Instance #include CoCreateInstance(
CLSID_CorMetaDataDispenser, 0, CLSCTX_INPROC_SERVER, IID_IMetaDataDispenser, (LPVOID *)&pIMetaDataDispenser );
pIMetaDataDispenser->OpenScope( szFileName,
// Name of assembly file
ofRead, IID_IMetaDataImport,// Desired interface (LPUNKNOWN *)&pIMetaDataImport );
Important Metadata Tokens mdTypeDef: A class mdMethodDef: A code member of a class mdParamDef: A parameter to a method mdFieldDef: A data member of a class mdPropertyDef: Similar to a field, but with accessor methods mdTypeRef: A reference to a class (possibly in another assembly) mdMemberRef: A reference to a code or data member in another module mdModuleRef: A reference to another module (possibly in another assembly)
IMetaDataImport Methods Token
Enumeration Method
Get Properties method
mdTypeDef
EnumTypeDefs
GetTypeDefProps
mdMethodDef
EnumMethods
GetMethodProps
mdFieldDef
EnumFields
GetFieldProps
mdEvent
EnumEvents
GetEventProps
mdParamDef
EnumParams
GetParamProps
mdToken
EnumMembers
GetMemberProps
mdProperty
EnumProperties
GetPropertyProps
mdMemberRef
EnumMemberRefs
GetMemberRefProps
mdModuleRef
EnumModuleRefs
GetModuleRefProps
mdTypeRef
EnumTypeRefs
GetTypeRefProps
mdPermission
EnumPermissionSets
GetPermissionSetProps
mdString
EnumUserStrings
GetUserString
mdCustomValue
EnumCustomAttributes
GetCustomAttributeProps
mdInterfaceImpl
EnumInterfaceImpls
GetInterfaceImplProps
This List Is Not Complete!
Example: Enumerating All Methods In An Assembly ::EnumTypeDefs returns an array of mdTypeDefs Each representing one class
::GetTypeDefProps returns the name of the class (and other assorted info) Pass each mdTypeDef to ::EnumMethods to get an array of mdMethodDefs One per method in the class
::GetMethodProps returns the name of the method (and other assorted info)
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
Type Signatures Signatures are very similar in concept to C++ name mangling They both provide a minimal level assurance that the calling code matches up with the called code
In .NET, all methods and fields have a signature associated with them .NET signatures are a sequence of bytes that describe “type” information and other details For example, the return value and parameter types of a method
Type Signature Encoding Format for methods: Calling convention (BYTE) Number of parameters (BYTE) Return value type (Variable length) Parameter types (Variable length)
CorElementType enum in CORHDR.H defines “base” types More complex types (e.g., pointers) are built out of base types and modifiers
Type Signature Example long FirstFunction( long j, float k); becomes… 0x00
CorCallingConvention::IMAGE_CEE_CS_CALLCONV_DEFAULT
0x02
Number of parameters
0x08
Return type: CorElementType::ELEMENT_TYPE_I4
0x08
Param 1 type: CorElementType::ELEMENT_TYPE_I4
0x0C Param 2 type: CorElementType::ELEMENT_TYPE_R4
Drilling Into Metadata Tokens The DWORD tokens used by the unmanaged interfaces are easily decoded The top BYTE indicates the token type See CorTokenType in CORHDR.H
The bottom 3 BYTEs are an array index Indices start at 0, and increase monotonically through the assembly Thus, a maximum of 24MB of any particular token type in an assembly
Token Examples typedef enum CorTokenType { mdtModule
= 0x00000000,
mdtTypeRef
= 0x01000000,
mdtTypeDef
= 0x02000000,
mdtFieldDef
= 0x04000000,
mdtMethodDef
= 0x06000000,
…
Example tokens 0x02000034 = The 0x34th TypeDef in the assembly 0x06000159 = The 0x159th method in the assembly
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
ILDASM “Intermediate Language Disassembler” Displays IL and metadata for an assembly Displays metadata in a nice, graphical hierarchy But information is not necessarily complete
Semi-undocumented /ADV option Demo
Aisto’s Reflector A much more complete Metadata viewer than ILDasm Demo Lutz Roeder’s “Reflector” shows assembly resources, and lets you save them to a file Demo 2: Documentation.xsl from Reflector http://www.aisto.com/roeder/dotnet/
MetaInfo Written in C++ and the unmanaged API Microsoft’s most complete metadata tool Sources in <.NET SDK> .\tool developers guide\samples\metainfo
Demo
TypeRefViewer Demo Shows imported namespaces / classes / methods From the November 2001 MSDN magazine, including source
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
Wrap-up And Resources .NET SDK .\include\cor.h, .\include\corhdr.h .\tool developers guide\samples\metainfo .\tool developers guide\docs\ • “Assembly Metadata Unmanaged API.doc” • “Metadata Unmanaged API.doc”
Reflector program: http://www.aisto.com/roeder/dotnet/ Avoiding DLL Hell, Matt Pietrek, October 2000 MSDN magazine (Meta program) March 20001 MSDN magazine (MetaViewer) November 2001 MSDN magazine (TypeRefViewer)
Housekeeping Who is this guy? Co-founder, Mindreef LLC (www.mindreef.com) President, Wheaty Productions Inc. MSDN’s “Under the Hood” columnist
web: http://www.wheaty.net email: [email protected] Source availability
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
Metadata: The Big Picture Metadata is information stored in .NET assemblies that makes them self-describing Metadata describes: Classes (types) provided by the assembly Method and field information for classes Version information Dependencies on other assemblies and modules Security attributes required
Metadata provides the information that the .NET runtime needs to ensure type safety and security
Demo: Many Views Of Metadata
Metadata: Evolution Of A COM Concept COM IDL files describe interfaces, methods, parameters, and types in sufficient detail to marshal calls between processes COM Type Libraries are just a binary representation of IDL information Metadata is effectively type libraries on steroids Microsoft provides tools to convert between metadata and type libraries
Metadata Usage In .NET The .NET runtime relies heavily on metadata to link together calls and data references at runtime Metadata is stored in a binary format, and subject to change Microsoft provides .NET classes and classic COM interfaces to read and write metadata
The Advantage Of Metadata A shared standard of representing information between any .NET compatible language Cool benefit: Intellisense for any .NET component
Metadata is automatically generated and consumed by .NET compilers No need to create .IDL files to expose your code to others No need to mess around with .H files and import libraries
Emitting Metadata Metadata (and associated IL) can be created using the .NET runtime Compilers aren’t the only ones that get to play!
System.Reflection.Emit classes are the basis for dynamically creating code: AssemblyBuilder MethodBuilder ILGenerator Many more!
Focus in this presentation is consuming metadata
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
The Metadata Hierarchy I Assembly (e.g., mscorlib)
Collections Single Instances
Modules
Module (i.e., a DLL)
Classes Global Methods & Variables
The Metadata Hierarchy II Class (e.g., System.Object)
Methods
Fields
Method (e.g., “Foo”)
Parameters
Properties
Parameter “int Param1”
Events
Parameter “String Param2”
Assembly Metadata Name Public key Version # (major & minor) Build # Revision # Dependencies on other assemblies Locale
Class Metadata Contents Name Flags (CorTypeAttr) Visibility (Public, private, assembly, …) Layout (auto, sequential, explicit…) Type (class, interface, value type, …)
Base class (if applicable)
Method Metadata Contents Name Parent class Flags (CorMethodAttr) Visibility Static / Virtual Platform Invoke (interop to legacy code)
Implementation Flags (CorMethodImpl) IL / OptIL / Native Managed / Unmanaged implementation
Signature RVA to IL in executable file
Parameter Metadata Contents Name Parent method Order in method parameter list Attributes (CorParamAttr) In / Out Has default value Optional
Field Metadata Contents Name Parent class Attributes (CorFieldAttr) Visibility Static member
Signature (implicitly gives the size)
Property Metadata Contents Name Parent class Attributes (CorPropertyAttr) Signature (implicitly gives the size) Getter & Setter methods Default value
Event Metadata Contents Name Parent class Fired method AddOn / RemoveOn methods
Custom Attributes Custom attributes are special classes that provide additional information about a: Assembly / Module Class / Struct Method / Constructor / Property Enum Parameter / Field
Unmanaged and Reflection APIs for reading / creating custom attributes
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
About The Reflection API Full access to metadata using .NET runtime classes Assemblies, classes, methods, fields, etc… are all represented as .NET objects Layered over the unmanaged metadata interfaces
The Metadata Hierarchy As Seen By The Reflection API Assembly
ManifestResource
Module
Type
FieldInfo
PropertyInfo
MethodInfo
EventInfo
ParameterInfo
AssemblyName
Custom Attributes
Getting To The Reflection API There are a variety of entry points into the Reflection classes Assembly::LoadFrom loads a file based assembly Returns an Assembly *
For an existing .NET object, use the ::GetType method to get a Type* Type::Module property moves up the hierarchy Type::GetMethods, etc… moves down
Demo ReflectCS program
Example: Enumerating All Methods In An Assembly Use Assembly::LoadFrom to load a .NET assembly Returns an Assembly *
Assembly::GetModules returns an array of Module *’s For each Module *, call Module::GetTypes Returns an array of Type *’s
For each Type *, call Type::GetMethods Returns an array of MethodInfo *’s
For each MethodInfo *, the MethodInfo::Name property returns the method’s name
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
About The Unmanaged API Regular COM interfaces that don’t use the .NET runtime Relies heavily on “tokens” More granular than equivalent COM ITypeLib & ITypeInfo interfaces Information isn’t packed into structures
Declared in COR.H / CORHDR.H CLSIDs and IIDs defined in CORGUID.LIB
The Unmanaged Metadata Interfaces IMetaDataDispenser The “root” interface for obtaining all others Works on file or memory based metadata
IMetaDataImport The primary interface for reading class info
IMetaDataAssemblyImport IMetaDataEmit IMetaDataAssemblyEmit
Demo The Meta program
Obtaining an IMetaDataImport Interface Instance #include
CLSID_CorMetaDataDispenser, 0, CLSCTX_INPROC_SERVER, IID_IMetaDataDispenser, (LPVOID *)&pIMetaDataDispenser );
pIMetaDataDispenser->OpenScope( szFileName,
// Name of assembly file
ofRead, IID_IMetaDataImport,// Desired interface (LPUNKNOWN *)&pIMetaDataImport );
Important Metadata Tokens mdTypeDef: A class mdMethodDef: A code member of a class mdParamDef: A parameter to a method mdFieldDef: A data member of a class mdPropertyDef: Similar to a field, but with accessor methods mdTypeRef: A reference to a class (possibly in another assembly) mdMemberRef: A reference to a code or data member in another module mdModuleRef: A reference to another module (possibly in another assembly)
IMetaDataImport Methods Token
Enumeration Method
Get Properties method
mdTypeDef
EnumTypeDefs
GetTypeDefProps
mdMethodDef
EnumMethods
GetMethodProps
mdFieldDef
EnumFields
GetFieldProps
mdEvent
EnumEvents
GetEventProps
mdParamDef
EnumParams
GetParamProps
mdToken
EnumMembers
GetMemberProps
mdProperty
EnumProperties
GetPropertyProps
mdMemberRef
EnumMemberRefs
GetMemberRefProps
mdModuleRef
EnumModuleRefs
GetModuleRefProps
mdTypeRef
EnumTypeRefs
GetTypeRefProps
mdPermission
EnumPermissionSets
GetPermissionSetProps
mdString
EnumUserStrings
GetUserString
mdCustomValue
EnumCustomAttributes
GetCustomAttributeProps
mdInterfaceImpl
EnumInterfaceImpls
GetInterfaceImplProps
This List Is Not Complete!
Example: Enumerating All Methods In An Assembly ::EnumTypeDefs returns an array of mdTypeDefs Each representing one class
::GetTypeDefProps returns the name of the class (and other assorted info) Pass each mdTypeDef to ::EnumMethods to get an array of mdMethodDefs One per method in the class
::GetMethodProps returns the name of the method (and other assorted info)
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
Type Signatures Signatures are very similar in concept to C++ name mangling They both provide a minimal level assurance that the calling code matches up with the called code
In .NET, all methods and fields have a signature associated with them .NET signatures are a sequence of bytes that describe “type” information and other details For example, the return value and parameter types of a method
Type Signature Encoding Format for methods: Calling convention (BYTE) Number of parameters (BYTE) Return value type (Variable length) Parameter types (Variable length)
CorElementType enum in CORHDR.H defines “base” types More complex types (e.g., pointers) are built out of base types and modifiers
Type Signature Example long FirstFunction( long j, float k); becomes… 0x00
CorCallingConvention::IMAGE_CEE_CS_CALLCONV_DEFAULT
0x02
Number of parameters
0x08
Return type: CorElementType::ELEMENT_TYPE_I4
0x08
Param 1 type: CorElementType::ELEMENT_TYPE_I4
0x0C Param 2 type: CorElementType::ELEMENT_TYPE_R4
Drilling Into Metadata Tokens The DWORD tokens used by the unmanaged interfaces are easily decoded The top BYTE indicates the token type See CorTokenType in CORHDR.H
The bottom 3 BYTEs are an array index Indices start at 0, and increase monotonically through the assembly Thus, a maximum of 24MB of any particular token type in an assembly
Token Examples typedef enum CorTokenType { mdtModule
= 0x00000000,
mdtTypeRef
= 0x01000000,
mdtTypeDef
= 0x02000000,
mdtFieldDef
= 0x04000000,
mdtMethodDef
= 0x06000000,
…
Example tokens 0x02000034 = The 0x34th TypeDef in the assembly 0x06000159 = The 0x159th method in the assembly
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
ILDASM “Intermediate Language Disassembler” Displays IL and metadata for an assembly Displays metadata in a nice, graphical hierarchy But information is not necessarily complete
Semi-undocumented /ADV option Demo
Aisto’s Reflector A much more complete Metadata viewer than ILDasm Demo Lutz Roeder’s “Reflector” shows assembly resources, and lets you save them to a file Demo 2: Documentation.xsl from Reflector http://www.aisto.com/roeder/dotnet/
MetaInfo Written in C++ and the unmanaged API Microsoft’s most complete metadata tool Sources in <.NET SDK> .\tool developers guide\samples\metainfo
Demo
TypeRefViewer Demo Shows imported namespaces / classes / methods From the November 2001 MSDN magazine, including source
Agenda What Is Metadata? Metadata Object Overview The Reflection API The Unmanaged API Type Signatures & Tokens Tools And Demo Wrap-up And Resources
Wrap-up And Resources .NET SDK .\include\cor.h, .\include\corhdr.h .\tool developers guide\samples\metainfo .\tool developers guide\docs\ • “Assembly Metadata Unmanaged API.doc” • “Metadata Unmanaged API.doc”
Reflector program: http://www.aisto.com/roeder/dotnet/ Avoiding DLL Hell, Matt Pietrek, October 2000 MSDN magazine (Meta program) March 20001 MSDN magazine (MetaViewer) November 2001 MSDN magazine (TypeRefViewer)
