Report a bug
If you spot a problem with this page, click here to create a Bugzilla issue.
Improve this page
Quickly fork, edit online, and submit a pull request for this page. Requires a signed-in GitHub account. This works well for small changes. If you'd like to make larger changes you may want to consider using a local clone.

dmd.clone

Compiler implementation of the D programming language.
Authors:

Source clone.d

pure StorageClass mergeFuncAttrs(StorageClass s1, const FuncDeclaration f);
Merge function attributes pure, nothrow, @safe, @nogc, and @disable from f into s1.
Parameters:
StorageClass s1 storage class to merge into
FuncDeclaration f function
Returns:
merged storage class
FuncDeclaration hasIdentityOpAssign(AggregateDeclaration ad, Scope* sc);
Check given aggregate actually has an identity opAssign or not.
Parameters:
AggregateDeclaration ad struct or class
Scope* sc current scope
Returns:
if found, returns FuncDeclaration of opAssign, otherwise null
FuncDeclaration buildOpAssign(StructDeclaration sd, Scope* sc);
Build opAssign for a struct.
The generated opAssign function has the following signature:
ref S opAssign(S s)    // S is the name of the `struct`
The opAssign function will be built for a struct S if the following constraints are met:
1. S does not have an identity opAssign defined.
2. S has at least one of the following members: a postblit (user-defined or generated for fields that have a defined postblit), a destructor (user-defined or generated for fields that have a defined destructor) or at least one field that has a defined opAssign.
3. S does not have any non-mutable fields.
If S has a disabled destructor or at least one field that has a disabled opAssign, S.opAssign is going to be generated, but marked with @disable
If S defines a destructor, the generated code for opAssign is:
S __swap = void;
__swap = this;   // bit copy
this = s;        // bit copy
__swap.dtor();
Otherwise, if S defines a postblit, the generated code for opAssign is:
this = s;
Note that the parameter to the generated opAssign is passed by value, which means that the postblit is going to be called (if it is defined) in both of the above situations before entering the body of opAssign. The assignments in the above generated function bodies are blit expressions, so they can be regarded as memcpys (opAssign is not called as this will result in an infinite recursion; the postblit is not called because it has already been called when the parameter was passed by value).
If S does not have a postblit or a destructor, but contains at least one field that defines an opAssign function (which is not disabled), then the body will make member-wise assignments:
this.field1 = s.field1;
this.field2 = s.field2;
...;
In this situation, the assignemnts are actual assign expressions (opAssign is used if defined).
Parameters:
StructDeclaration sd struct to generate opAssign for
Scope* sc context
Returns:
generated opAssign function
bool needOpEquals(StructDeclaration sd);
We need an opEquals for the struct if any fields has an opEquals. Generate one if a user-specified one does not exist.
FuncDeclaration buildOpEquals(StructDeclaration sd, Scope* sc);
Build opEquals for struct. const bool opEquals(const S s) { ... }
By fixing https://issues.dlang.org/show_bug.cgi?id=3789 opEquals is changed to be never implicitly generated. Now, struct objects comparison s1 == s2 is translated to: s1.tupleof == s2.tupleof to calculate structural equality. See EqualExp.op_overload.
FuncDeclaration buildXopEquals(StructDeclaration sd, Scope* sc);
Build __xopEquals for TypeInfo_Struct static bool __xopEquals(ref const S p, ref const S q) { return p == q; }
This is called by TypeInfo.equals(p1, p2). If the struct does not support const objects comparison, it will throw "not implemented" Error in runtime.
FuncDeclaration buildXopCmp(StructDeclaration sd, Scope* sc);
Build __xopCmp for TypeInfo_Struct static bool __xopCmp(ref const S p, ref const S q) { return p.opCmp(q); }
This is called by TypeInfo.compare(p1, p2). If the struct does not support const objects comparison, it will throw "not implemented" Error in runtime.
FuncDeclaration buildXtoHash(StructDeclaration sd, Scope* sc);
Build _xtoHash for non-bitwise hashing static hash_t xtoHash(ref const S p) nothrow @trusted;
DtorDeclaration buildDtor(AggregateDeclaration ad, Scope* sc);
Create inclusive destructor for struct/class by aggregating all the destructors in dtors[] with the destructors for all the members.
Parameters:
AggregateDeclaration ad struct or class to build destructor for
Scope* sc context
Returns:
generated function, null if none needed

Note Close similarity with StructDeclaration::buildPostBlit(), and the ordering changes (runs backward instead of forwards).

DtorDeclaration buildExternDDtor(AggregateDeclaration ad, Scope* sc);
build a shim function around the compound dtor that translates a C++ destructor to a destructor with extern(D) calling convention
Parameters:
AggregateDeclaration ad the aggregate that contains the destructor to wrap
Scope* sc the scope in which to analyze the new function
Returns:
the shim destructor, semantically analyzed and added to the class as a member
FuncDeclaration buildInv(AggregateDeclaration ad, Scope* sc);
Create inclusive invariant for struct/class by aggregating all the invariants in invs[]. void __invariant() const [pure nothrow @trusted] { invs[0](), invs[1](), ...; }