From efd764b9077b101fb02b4b484766be8ae81f2a51 Mon Sep 17 00:00:00 2001
From: Ludovic 'Archivist' Lagouardette <ludovic@nekoit.xyz>
Date: Tue, 21 May 2024 19:24:55 +0200
Subject: [PATCH] Added garbage collection (and a memory corruption bug)

---
 ink.h          |  83 +++++++++++++++++--
 lib.c          | 221 ++++++++++++++++++++++++++++++++++++++++++++++++-
 main.c         |   3 +-
 test/test02.nk |   2 +
 test/test03.nk |  10 ++-
 test/test04.nk |   2 +-
 6 files changed, 308 insertions(+), 13 deletions(-)

diff --git a/ink.h b/ink.h
index 8846c9b..0ef9278 100644
--- a/ink.h
+++ b/ink.h
@@ -1,6 +1,7 @@
 #pragma once
 #include "stddef.h"
 
+
 /**
  * Represents the natively defined type of integers
  */
@@ -85,16 +86,31 @@ struct ink_routine {
 	void *routine_userdata;
 };
 
+/**
+ * Contains a list of element on which garbage collection is to not be performed
+ */
 struct ink_collection_list {
-	struct elem* elems;
-	size_t size;
+    struct elem* elements;
+    size_t count;
 };
 
+struct element_slab {
+    void* data;
+    int uses;
+    int in_use;
+};
+
+/**
+ * Contains all the data for every element of any type and its garbage collection information.
+ */
 struct ink_type {
-	const char* name;
-	void* elements;
-	void (*collect)(void*);
-	struct ink_collection_list (*gc)(void*);
+	const char* name; //< The name of the type
+    int element_size; //< The size of individual elements of the type, 0 for int adjacent, negative for unmanaged size
+	struct element_slab* elements; //< The elements that are still live
+    int elements_top; //< The top of the elements list
+    int elements_capacity; //< The allocated capacity of the elements list
+	void (*collect)(struct context*,void*); //< The "destructor" of the object
+	struct ink_collection_list (*gc)(struct context*,void*); //< A function that returns an in-interpreter allocated list of elem references within the object
 };
 
 /**
@@ -136,8 +152,9 @@ struct context {
 	char **lex_reserved_words;
 	int lex_reserved_words_capacity;
 	int lex_reserved_words_top;
-	
-	unsigned int steps;
+
+    unsigned int steps;
+    unsigned int collections;
 
     /**
      * Can be set to any data that is convenient to the user to track and use within natively defined functions
@@ -251,6 +268,56 @@ void ink_pop_fn(struct context *ctx);
  */
 void ink_pop(struct context *ctx);
 
+/**
+ * Declares a new type that can be stored within the interpreter
+ * @param ctx The context in which to add the file
+ * @param type_name The name of the type we want to add
+ * @param size The size in bytes of the type to add, size of 0 mean no size, in which case the type is adjacent to C int, negative size means that the memory is not managed by the interpreter.
+ * @param collect A "destructor" function for the data
+ * @param gc A function that returns a list (allocated with the `inner_malloc`) of all the elements this element holds references to
+ * @return if positive, a new type id, if negative an error value
+ * @internal user defined type ids minimal value is 15, we keep the first 16 types as reserved, just like negative type ids
+ */
+int ink_new_type(
+        struct context* ctx,
+        const char* type_name,
+        int size,
+        void (*collect)(struct context*,void*),
+        struct ink_collection_list (*gc)(struct context*,void*)
+);
+
+/**
+ *
+ * @param ctx The context of the interpreter
+ * @param ref The in-interpreter reference
+ * @return A pointer to the created value or NULL if it can't be found or has a size of 0
+ */
+void* ink_get_value(struct context* ctx, struct elem ref);
+
+/**
+ * Builds a native type from the provided memory by copying it using memcpy
+ * @param ctx The context in which we operate
+ * @param type_id The type_id to use
+ * @param ptr The pointer from which to copy
+ * @return The in-interpreter reference of the newly created element
+ */
+struct elem ink_make_native(struct context* ctx, int type_id, void* ptr);
+
+/**
+ * Builds a transparent type from the provided pointer
+ * @param ctx The context in which we operate
+ * @param type_id The type_id to use
+ * @param ptr The pointer
+ * @return The in-interpreter reference of the newly created element
+ */
+struct elem ink_make_transparent(struct context* ctx, int type_id, void* ptr);
+
+/**
+ * Launch the mark and sweep garbage collection
+ * @param ctx The context to clean
+ */
+void ink_gc(struct context* ctx);
+
 #ifdef __cplusplus
 };
 #endif
\ No newline at end of file
diff --git a/lib.c b/lib.c
index e0cc0a4..c5c5835 100644
--- a/lib.c
+++ b/lib.c
@@ -21,6 +21,9 @@
 #define _KEYWORD_INK_END "end"
 #define _KEYWORD_INK_RETURN "return"
 
+#define min(x, y) ((x) > (y) ? (y) : (x))
+#define max(x, y) ((x) < (y) ? (y) : (x))
+
 struct label {
 	int active;
 	int dest;
@@ -271,6 +274,9 @@ struct context* ink_make_context(void*(*malloc)(size_t), void*(*realloc)(void*,
 	ctx->routines = NULL;
 	ctx->routines_capacity = 0;
 	ctx->routines_top = 0;
+    ctx->types = NULL;
+    ctx->types_capacity = 0;
+    ctx->types_top = 0;
 	ctx->native_words = NULL;
 	ctx->native_words_capacity = 0;
 	ctx->native_words_top = 0;
@@ -280,7 +286,8 @@ struct context* ink_make_context(void*(*malloc)(size_t), void*(*realloc)(void*,
 	ctx->lex_reserved_words = NULL;
 	ctx->lex_reserved_words_capacity = 0;
 	ctx->lex_reserved_words_top = 0;
-	ctx->steps = 0;
+    ctx->collections = 0;
+    ctx->steps = 0;
 	return ctx;
 }
 
@@ -791,6 +798,172 @@ int ink_step_everyone(struct context* pContext) {
 	return 0;
 }
 
+int ink_new_type(
+    struct context* ctx,
+    const char* type_name,
+    int size,
+    void (*collect)(struct context*,void*),
+    struct ink_collection_list (*gc)(struct context*,void*)
+) {
+    if(ctx->panic) return -128;
+    if(ctx->types == NULL) {
+        ctx->types = ctx->inner_malloc(sizeof(struct ink_type) * 8);
+        ctx->types_top = 0;
+        ctx->types_capacity = 8;
+    } else if(ctx->types_top == ctx->types_capacity) {
+        int new_count = (ctx->types_capacity + ctx->types_capacity/2);
+        void* renewed = ctx->inner_realloc(ctx->types, sizeof(struct ink_type) * new_count);
+        if(renewed == NULL) {
+            return -129;
+        } else {
+            ctx->types = renewed;
+            ctx->types_capacity = new_count;
+        }
+    }
+    ctx->types[ctx->types_top].name = type_name;
+    ctx->types[ctx->types_top].element_size = size;
+    ctx->types[ctx->types_top].elements = NULL;
+    ctx->types[ctx->types_top].elements_top = 0;
+    ctx->types[ctx->types_top].elements_capacity = 0;
+    ctx->types[ctx->types_top].collect = collect;
+    ctx->types[ctx->types_top].gc = gc;
+
+    ctx->types_top++;
+    // Satisfying the minimal value requirement
+    return ctx->types_top - 1 + 16;
+}
+
+static struct element_slab* ink_get_value_link(struct context* ctx, struct elem ref) {
+    if(ref.type < 16) return NULL;
+    int type_id = ref.type - 16;
+    if(type_id >= ctx->types_top) return NULL;
+    if(ctx->types[type_id].element_size == 0) return NULL;
+    if(ref.value < 0) return NULL;
+    if(ref.value >= ctx->types[type_id].elements_top) return NULL;
+    if(! ctx->types[type_id].elements[ref.value].in_use) return NULL;
+    return ctx->types[type_id].elements + ref.value;
+}
+
+void* ink_get_value(struct context* ctx, struct elem ref) {
+    struct element_slab* s;
+    s = ink_get_value_link(ctx, ref);
+    if(s == NULL) return NULL;
+    return s->data;
+}
+
+struct elem ink_make_native(struct context* ctx, int type, void* ptr) {
+    if(type < 16) {
+        struct elem ret;
+        ret.type = 0;
+        ret.value = -130;
+        return ret;
+    }
+    int type_id = type - 16;
+    if(type_id >= ctx->types_top) {
+        struct elem ret;
+        ret.type = 0;
+        ret.value = -129;
+        return ret;
+    }
+
+    if(ctx->panic) {
+        struct elem ret;
+        ret.type = 0;
+        ret.value = -135;
+        return ret;
+    }
+    if(ctx->types[type_id].elements == NULL) {
+        ctx->types[type_id].elements = ctx->inner_malloc(sizeof(struct element_slab) * 8);
+        ctx->types[type_id].elements_top = 0;
+        ctx->types[type_id].elements_capacity = 8;
+        memset(ctx->types[type_id].elements + ctx->types[type_id].elements_top, 0, ctx->types[type_id].elements_capacity - ctx->types[type_id].elements_top);
+    } else if(ctx->types[type_id].elements_top == ctx->types[type_id].elements_capacity) {
+        int new_count = (ctx->types[type_id].elements_capacity + ctx->types[type_id].elements_capacity/2);
+        void* renewed = ctx->inner_realloc(ctx->types[type_id].elements, sizeof(struct element_slab) * new_count);
+        if(renewed == NULL) {
+            struct elem ret;
+            ret.type = 0;
+            ret.value = -129;
+            return ret;
+        } else {
+            ctx->types[type_id].elements = renewed;
+            ctx->types[type_id].elements_capacity = new_count;
+            memset(ctx->types[type_id].elements + ctx->types[type_id].elements_top, 0, ctx->types[type_id].elements_capacity - ctx->types[type_id].elements_top);
+        }
+    }
+    int g = ctx->types[type_id].elements_capacity;
+    int i;
+    for(i = 0; i < g; ++i) {
+        if(! ctx->types[type_id].elements[i].in_use) {
+            ctx->types[type_id].elements[i].in_use = 1;
+            ctx->types[type_id].elements[i].uses = 1;
+            if(ctx->types[type_id].element_size < 0) {
+                ctx->types[type_id].elements[i].data = ptr;
+            } else {
+                void* new_ptr = ctx->malloc(ctx->types[type_id].element_size);
+                if(new_ptr == NULL) {
+                    struct elem ret;
+                    ret.type = 0;
+                    ret.value = -139;
+                    return ret;
+                }
+                memcpy(new_ptr, ptr, ctx->types[type_id].element_size);
+                ctx->types[type_id].elements[i].data = ptr;
+            }
+            ctx->types[type_id].elements_top = max(ctx->types[type_id].elements_top+1, i+1);
+            struct elem ret;
+            ret.type = type;
+            ret.value = i;
+            return ret;
+        }
+    }
+    struct elem ret;
+    ret.type = 0;
+    ret.value = -140;
+    return ret;
+}
+
+void ink_gc(struct context* ctx) {
+    int i, j, k;
+    for(i = 0; i < ctx->types_top; ++i) {
+        for(j = 0; j < ctx->types[i].elements_top; ++j) {
+            ctx->types[i].elements[j].uses = 0;
+        }
+    }
+
+    for(i = 0; i < ctx->types_top; ++i) {
+        for(j = 0; j < ctx->types[i].elements_top; ++j) {
+            struct ink_collection_list c = ctx->types[i].gc(ctx, ctx->types[i].elements[j].data);
+            for(k = 0; k < c.count; ++k) {
+                struct element_slab* v = ink_get_value_link(ctx, c.elements[k]);
+                if(v != NULL) ++v->uses;
+            }
+            if(c.elements != NULL) ctx->inner_free(c.elements);
+        }
+    }
+    for(i = 0; i < ctx->routines_top; ++i) {
+        for(j = 0; j < ctx->routines[i].top; ++j) {
+            struct element_slab* v = ink_get_value_link(ctx, ctx->routines[i].stack[j]);
+            if(v != NULL) ++v->uses;
+        }
+    }
+
+    for(i = 0; i < ctx->types_top; ++i) {
+        for(j = 0; j < ctx->types[i].elements_top; ++j) {
+            if(ctx->types[i].elements[j].uses == 0) {
+                ctx->collections++;
+                ctx->types[i].collect(ctx, ctx->types[i].elements[j].data);
+                if(ctx->types[i].element_size > 0) {
+                    ctx->free(ctx->types[i].elements[j].data);
+                }
+                ctx->types[i].elements[j].data = NULL;
+                ctx->types[i].elements[j].uses = 0;
+                ctx->types[i].elements[j].in_use = 0;
+            }
+        }
+    }
+}
+
 /**********************************************************************************************************************/
 
 static void print_stacktrace(struct context* _) {
@@ -1069,10 +1242,54 @@ static void print_as_utf8(struct context* ctx) {
 	ink_pop(ctx);
 }
 
+struct ink_array {
+    int top;
+    int capacity;
+    struct elem* elements;
+};
+
+static int get_type_by_name(struct context* ctx, const char* name) {
+    int i;
+    for(i = 0; i < ctx->types_top; ++i) {
+        if(strcmp(ctx->types[i].name, name) == 0) {
+            return i + 16;
+        }
+    }
+    return -1;
+}
+
+static void collect_array(struct context* ctx, void* array) {
+    struct ink_array* ary = array;
+    ctx->free(ary->elements);
+}
+
+static struct ink_collection_list gc_array(struct context* ctx, void* array) {
+    struct ink_array* ary = array;
+    struct ink_collection_list c;
+    c.elements = ctx->inner_malloc(sizeof(struct elem)*ary->top);
+    c.count = ary->top;
+    memcpy(c.elements, ary->elements, sizeof(struct elem)*ary->top);
+    return c;
+}
+
+static void new_array(struct context* ctx) {
+    int tid =  get_type_by_name(ctx, "array");
+    struct ink_array ary;
+    ary.elements = NULL;
+    ary.top = 0;
+    ary.capacity = 0;
+    struct elem e = ink_make_native(ctx, tid, &ary);
+    ink_push(ctx, e);
+}
+
+
+
 int ink_std_library(struct context* ctx) {
 	int v;
 	v = 0;
-	v += ink_add_native(ctx, "trace", print_stacktrace);
+    int array_t = ink_new_type(ctx, "array", sizeof(struct ink_array), collect_array, gc_array);
+    v += ink_add_native(ctx, "array", new_array);
+    v += ink_add_native(ctx, "trace", print_stacktrace);
 	v += ink_add_native(ctx, "print_int", print_int);
 	v += ink_add_native(ctx, "print_utf8", print_as_utf8);
 	v += ink_add_native(ctx, "+", add_int);
diff --git a/main.c b/main.c
index b0ae3f8..cbe4d79 100644
--- a/main.c
+++ b/main.c
@@ -24,6 +24,7 @@ int main(int argc, char** argv) {
 	while(ink_can_run(ctx)) {
 		ink_step_everyone(ctx);
 	}
-	printf("\nExecuted in %u steps\n", ctx->steps);
+    ink_gc(ctx);
+	printf("\nExecuted in %u steps\nCollected %u times\n", ctx->steps, ctx->collections);
 	return ctx->panic;
 }
\ No newline at end of file
diff --git a/test/test02.nk b/test/test02.nk
index 3fc8712..5dea0ed 100644
--- a/test/test02.nk
+++ b/test/test02.nk
@@ -1,3 +1,5 @@
 fn print_n_utf8 do
     print_n_utf8_impl drop
 end
+
+array drop array drop
\ No newline at end of file
diff --git a/test/test03.nk b/test/test03.nk
index 51f2f9f..c850e4b 100644
--- a/test/test03.nk
+++ b/test/test03.nk
@@ -1 +1,9 @@
-4 potato2
\ No newline at end of file
+4 potato2
+
+fn u do
+
+loop:
+    65 print_int
+loop 1 jump_if
+
+end
\ No newline at end of file
diff --git a/test/test04.nk b/test/test04.nk
index c31557e..79f5b5e 100644
--- a/test/test04.nk
+++ b/test/test04.nk
@@ -1 +1 @@
-4 potato3
\ No newline at end of file
+4 potato3