[Swig-user] [java] Is my memory management ok?

[Hendrik M. <qwi...@sn...>]

Hi,

I think my interface file is almost perfect now.  However, I am unsure=20
whether the memory is managed properly.  Can someone shed some light on=20
this?

These are the relevant parts of my interface file:

%module Mona

%{
static char* printTreeNodeRecursively(mgTreeNode* self, int labelLength) =
{
   \\ malloc a string and return it
}
%}

struct mgTreeNode { /* tree node */
	mA a; /* alphabet element */
	struct mgTreeNode *left, *right; /* successors */
	mgId id; /* state space id */
	mgState state; /* automaton state */
};

%extend mgTreeNode {

	mgTreeNode(char *label, mgTreeNode *leftDaughter, mgTreeNode=20
*rightDaughter) {
	  \\ malloc an mgTreeNode, as well as a, left and right, 	=09
	  \\ initialize and return it
	}

	char *toString(int labelLength) {
		return printTreeNodeRecursively($self, labelLength);
	}

};

struct mgGta { /* Guided Tree Automaton */
	unsigned numSS; /* number of state spaces */
	mgStateSpace *stateSpace; /* array of state spaces */
	unsigned numUnivs; /* number of universes */
	mgUniverse *universe; /* array of universes */
	unsigned numVars; /* number of free variables */
	mgVariable *var; /* array of variables */
	mgKind *final; /* initial state space state kinds */
	unsigned numTypes; /* number of types (0: not using types) */
	mgType *type; /* array of types (NULL if not using types) */
};

/* load GTA from text file */
/* This ensures Java takes care of garbage collecting automata created=20
by this method */
%newobject mgLoad;
mgGta *mgLoad(const char *filename);

%newobject printTreeNodeRecursively;
%newobject toString;

I only need few GTAs, so it wouldn=E2=80=99t be that big a disaster if so=
mething=20
went wrong there, but I create masses of TreeNodes, so I want to make=20
sure they get reclaimed properly.  I checked one tree constructed with=20
the nodes.  The top-most node had swigCMemOwn true, but its left=20
daughter hadn=E2=80=99t.  Now I don=E2=80=99t see the memory of the daugh=
ters being=20
freed in the delete() function.

Maybe I need
%newobject mgTreeNode(char, mgTreeNode, mgTreeNode);
or do I?  Since that is converted to a Java constructor automatically, I =

thought swig would be smart enought to know that it creates a new node=20
as well?

Do I have to worry about the parts of the GTA not being reclaimed as=20
well?  There is a function mgFree(mgGta*) in the library I am wrapping,=20
should I put that in the finalize() function?  And write my own=20
freeTree(mgTreeNode*)?

Cheers, H.
--=20
Hendrik Maryns
http://tcl.sfs.uni-tuebingen.de/~hendrik/
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
http://aouw.org
Ask smart questions, get good answers:
http://www.catb.org/~esr/faqs/smart-questions.html