Incomplete code for C tokenizer textbuffer.

mwparserfromhell/parser/ctokenizer/common.h

@@ -43,6 +43,18 @@ SOFTWARE.
 #define malloc PyObject_Malloc  // XXX: yuck
 #define free   PyObject_Free
 
+/* Unicode support macros */
+
+#if defined(IS_PY3K) && PYTHON_MINOR_VERSION >= 3
+#define PEP_393
+#endif
+
+#ifdef PEP_393
+#define Unicode Py_UCS4
+#else
+#define Unicode Py_UNICODE
+#endif
+
 /* Error handling macros */
 
 #define BAD_ROUTE            self->route_state
@@ -63,18 +75,15 @@ extern PyObject* definitions;
 
 /* Structs */
 
-struct Textbuffer {
+typedef struct {
     Py_ssize_t size;
     Py_UNICODE* data;
-    struct Textbuffer* prev;
-    struct Textbuffer* next;
-};
-typedef struct Textbuffer Textbuffer;
+} Textbuffer;
 
 struct Stack {
     PyObject* stack;
     uint64_t context;
-    struct Textbuffer* textbuffer;
+    Textbuffer* textbuffer;
     struct Stack* next;
 };
 typedef struct Stack Stack;

mwparserfromhell/parser/ctokenizer/tok_parse.c

@@ -30,11 +30,6 @@ SOFTWARE.
 #define HEXDIGITS "0123456789abcdefABCDEF"
 #define ALPHANUM  "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
 
-static const char MARKERS[] = {
-    '{', '}', '[', ']', '<', '>', '|', '=', '&', '\'', '#', '*', ';', ':', '/',
-    '-', '!', '\n', '\0'};
-
-#define NUM_MARKERS 19
 #define MAX_BRACES 255
 #define MAX_ENTITY_SIZE 8
 
@@ -45,12 +40,6 @@ static const char MARKERS[] = {
 #define IS_SCHEME(scheme, slashes, reverse) \
     (call_def_func("is_scheme", scheme, slashes ? Py_True : Py_False, reverse ? Py_True : Py_False))
 
-#ifdef IS_PY3K
-    #define NEW_INT_FUNC      PyLong_FromSsize_t
-#else
-    #define NEW_INT_FUNC      PyInt_FromSsize_t
-#endif
-
 typedef struct {
     PyObject* title;
     int level;
@@ -798,7 +787,11 @@ static int Tokenizer_parse_heading(Tokenizer* self)
         self->global ^= GL_HEADING;
         return 0;
     }
-    level = NEW_INT_FUNC(heading->level);
+#ifdef IS_PY3K
+    level = PyLong_FromSsize_t(heading->level);
+#else
+    level = PyInt_FromSsize_t(heading->level);
+#endif
     if (!level) {
         Py_DECREF(heading->title);
         free(heading);

mwparserfromhell/parser/ctokenizer/tok_parse.h

@@ -24,6 +24,12 @@ SOFTWARE.
 
 #include "common.h"
 
+static const char MARKERS[] = {
+    '{', '}', '[', ']', '<', '>', '|', '=', '&', '\'', '#', '*', ';', ':', '/',
+    '-', '!', '\n', '\0'};
+
+#define NUM_MARKERS 19
+
 /* Functions */
 
 PyObject* Tokenizer_parse(Tokenizer*, uint64_t, int);

tests/tokenizer/text.mwtest

@@ -27,6 +27,6 @@ output: [Text(text="𐌲𐌿𐍄𐌰𐍂𐌰𐌶𐌳𐌰")]
 ---
 
 name:   large
-label:  a lot of text, requiring multiple textbuffer blocks in the C tokenizer
+label:  a lot of text, requiring proper storage in the C tokenizer
 input:  "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"
 output: [Text(text="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")]