Initial support for 'q' and 'Q' struct format codes: for now, only in

native mode, and only when config #defines HAVE_LONG_LONG.  Standard mode
will eventually treat them as 8-byte ints across all platforms, but that
likely requires a new set of routines in longobject.c first (while
sizeof(long) >= 4 is guaranteed by C, there's nothing in C we can rely
on x-platform to hold 8 bytes of int, so we'll have to roll our own;
I'm thinking of a simple pair of conversion functions, Python long
to/from sized vector of unsigned bytes; that may be useful for GMP
conversions too; std q/Q would call them with size fixed at 8).

test_struct.py:  In addition to adding some native-mode 'q' and 'Q' tests,
got rid of unused code, and repaired a non-portable assumption about
native sizeof(short) (it isn't 2 on some Cray boxes).

libstruct.tex:  In addition to adding a bit of 'q'/'Q' docs (more needed
later), removed an erroneous footnote about 'I' behavior.

Doc/lib/libstruct.tex

@@ -53,9 +53,11 @@ C and Python values should be obvious given their types:
   \lineiv{h}{\ctype{short}}{integer}{}
   \lineiv{H}{\ctype{unsigned short}}{integer}{}
   \lineiv{i}{\ctype{int}}{integer}{}
-  \lineiv{I}{\ctype{unsigned int}}{long}{(1)}
+  \lineiv{I}{\ctype{unsigned int}}{long}{}
   \lineiv{l}{\ctype{long}}{integer}{}
   \lineiv{L}{\ctype{unsigned long}}{long}{}
+  \lineiv{q}{\ctype{long long}}{long}{(1)}
+  \lineiv{Q}{\ctype{unsigned long long}}{long}{(1)}
   \lineiv{f}{\ctype{float}}{float}{}
   \lineiv{d}{\ctype{double}}{float}{}
   \lineiv{s}{\ctype{char[]}}{string}{}
@@ -68,10 +70,9 @@ Notes:
 
 \begin{description}
 \item[(1)]
-  The \character{I} conversion code will convert to a Python long if
-  the C \ctype{int} is the same size as a C \ctype{long}, which is
-  typical on most modern systems.  If a C \ctype{int} is smaller than
-  a C \ctype{long}, an Python integer will be created instead.
+  The \character{q} and \character{Q} conversion codes are available in
+  native mode only if the platform C compiler supports C \ctype{long long},
+  or, on Windows, \ctype{__int64}.
 \end{description}
 
 

Lib/test/test_struct.py

-from test_support import TestFailed, verbose
+from test_support import TestFailed, verbose, verify
 import struct
 ## import pdb
 
@@ -12,7 +12,7 @@ def simple_err(func, *args):
             func.__name__, args)
 ##      pdb.set_trace()
 
-simple_err(struct.calcsize, 'Q')
+simple_err(struct.calcsize, 'Z')
 
 sz = struct.calcsize('i')
 if sz * 3 != struct.calcsize('iii'):
@@ -93,14 +93,7 @@ tests = [
                '\000\000\000\000\000\000\000\300', 0),
 ]
 
-def badpack(fmt, arg, got, exp):
-    return
-
-def badunpack(fmt, arg, got, exp):
-    return "unpack(%s, %s) -> (%s,) # expected (%s,)" % (
-        `fmt`, `arg`, `got`, `exp`)
-
-isbigendian = struct.pack('=h', 1) == '\0\1'
+isbigendian = struct.pack('=i', 1)[0] == chr(0)
 
 for fmt, arg, big, lil, asy in tests:
     if verbose:
@@ -119,3 +112,47 @@ for fmt, arg, big, lil, asy in tests:
         if rev != arg and not asy:
             raise TestFailed, "unpack(%s, %s) -> (%s,) # expected (%s,)" % (
                 `fmt`, `res`, `rev`, `arg`)
+
+# Some q/Q sanity checks.
+
+has_native_qQ = 1
+try:
+    struct.pack("q", 5)
+except struct.error:
+    has_native_qQ = 0
+
+if verbose:
+    print "Platform has native q/Q?", has_native_qQ and "Yes." or "No."
+
+simple_err(struct.pack, "Q", -1)   # can't pack -1 as unsigned regardless
+simple_err(struct.pack, "q", "a")  # can't pack string as 'q' regardless
+simple_err(struct.pack, "Q", "a")  # ditto, but 'Q'
+
+def force_bigendian(value):
+    if isbigendian:
+        return value
+    chars = list(value)
+    chars.reverse()
+    return "".join(chars)
+
+if has_native_qQ:
+    bytes = struct.calcsize('q')
+    # The expected values here are in big-endian format, primarily because
+    # I'm on a little-endian machine and so this is the clearest way (for
+    # me) to force the code to get exercised.
+    for format, input, expected in (
+            ('q', -1, '\xff' * bytes),
+            ('q', 0, '\x00' * bytes),
+            ('Q', 0, '\x00' * bytes),
+            ('q', 1L, '\x00' * (bytes-1) + '\x01'),
+            ('Q', (1L << (8*bytes))-1, '\xff' * bytes),
+            ('q', (1L << (8*bytes-1))-1, '\x7f' + '\xff' * (bytes - 1))):
+        got = struct.pack(format, input)
+        bigexpected = force_bigendian(expected)
+        verify(got == bigexpected,
+               "%r-pack of %r gave %r, not %r" %
+                    (format, input, got, bigexpected))
+        retrieved = struct.unpack(format, got)[0]
+        verify(retrieved == input,
+               "%r-unpack of %r gave %r, not %r" %
+                    (format, got, retrieved, input))

Misc/NEWS

@@ -140,6 +140,14 @@ Library
 
 - pprint functions now much faster for large containers (tuple, list, dict).
 
+- New 'q' and 'Q' format codes in the struct module, corresponding to C
+  types "long long" and "unsigned long long" (on Windows, __int64).  In
+  native mode, these can be used only when the platform C compiler supports
+  these types (when HAVE_LONG_LONG is #define'd by the Python config
+  process), and then they inherit the sizes and alignments of the C types.
+  XXX TODO In standard mode, 'q' and 'Q' are supported on all platforms, and
+  XXX TODO are 8-byte integral types.
+
 Tests
 
 - New test_mutants.py runs dict comparisons where the key and value

Modules/structmodule.c

@@ -22,9 +22,11 @@ these can be preceded by a decimal repeat count:\n\
  h:short; H:unsigned short; i:int; I:unsigned int;\n\
  l:long; L:unsigned long; f:float; d:double.\n\
 Special cases (preceding decimal count indicates length):\n\
- s:string (array of char); p: pascal string (w. count byte).\n\
+ s:string (array of char); p: pascal string (with count byte).\n\
 Special case (only available in native format):\n\
  P:an integer type that is wide enough to hold a pointer.\n\
+Special case (not in native mode unless 'long long' in platform C):\n\
+ q:long long; Q:unsigned long long\n\
 Whitespace between formats is ignored.\n\
 \n\
 The variable struct.error is an exception raised on errors.";
@@ -65,6 +67,18 @@ typedef struct { char c; void *x; } s_void_p;
 #define DOUBLE_ALIGN (sizeof(s_double) - sizeof(double))
 #define VOID_P_ALIGN (sizeof(s_void_p) - sizeof(void *))
 
+/* We can't support q and Q in native mode unless the compiler does;
+   in std mode, they're 8 bytes on all platforms. */
+#ifdef HAVE_LONG_LONG
+typedef struct { char c; LONG_LONG x; } s_long_long;
+#define LONG_LONG_ALIGN (sizeof(s_long_long) - sizeof(LONG_LONG))
+
+#else
+static char qQ_error_msg[] =
+"q and Q unavailable in native mode on this platform; use a standard mode.\0";
+
+#endif
+
 #define STRINGIFY(x)    #x
 
 #ifdef __powerc
@@ -106,6 +120,93 @@ get_ulong(PyObject *v, unsigned long *p)
 	}
 }
 
+#ifdef HAVE_LONG_LONG
+
+/* Same, but handling native long long. */
+
+static int
+get_longlong(PyObject *v, LONG_LONG *p)
+{
+	LONG_LONG x;
+	int v_needs_decref = 0;
+
+	if (PyInt_Check(v)) {
+		x = (LONG_LONG)PyInt_AS_LONG(v);
+		*p = x;
+		return 0;
+	}
+	if (!PyLong_Check(v)) {
+		PyNumberMethods *m = v->ob_type->tp_as_number;
+		if (m != NULL && m->nb_long != NULL) {
+			v = m->nb_long(v);
+			if (v == NULL)
+				return -1;
+			v_needs_decref = 1;
+		}
+		if (!PyLong_Check(v)) {
+			PyErr_SetString(StructError,
+					"cannot convert argument to long");
+			if (v_needs_decref)
+				Py_DECREF(v);
+			return -1;
+		}
+	}
+	assert(PyLong_Check(v));
+	x = PyLong_AsLongLong(v);
+	if (v_needs_decref)
+		Py_DECREF(v);
+	if (x == (LONG_LONG)-1 && PyErr_Occurred())
+		return -1;
+	*p = x;
+	return 0;
+}
+
+/* Same, but handling native unsigned long long. */
+
+static int
+get_ulonglong(PyObject *v, unsigned LONG_LONG *p)
+{
+	unsigned LONG_LONG x;
+	int v_needs_decref = 0;
+
+	if (PyInt_Check(v)) {
+		long i = PyInt_AS_LONG(v);
+		if (i < 0) {
+			PyErr_SetString(StructError, "can't convert negative "
+					"int to unsigned");
+			return -1;
+		}
+		x = (unsigned LONG_LONG)i;
+		*p = x;
+		return 0;
+	}
+	if (!PyLong_Check(v)) {
+		PyNumberMethods *m = v->ob_type->tp_as_number;
+		if (m != NULL && m->nb_long != NULL) {
+			v = m->nb_long(v);
+			if (v == NULL)
+				return -1;
+			v_needs_decref = 1;
+		}
+		if (!PyLong_Check(v)) {
+			PyErr_SetString(StructError,
+					"cannot convert argument to long");
+			if (v_needs_decref)
+				Py_DECREF(v);
+			return -1;
+		}
+	}
+	assert(PyLong_Check(v));
+	x = PyLong_AsUnsignedLongLong(v);
+	if (v_needs_decref)
+		Py_DECREF(v);
+	if (x == (unsigned LONG_LONG)-1 && PyErr_Occurred())
+		return -1;
+	*p = x;
+	return 0;
+}
+
+#endif
 
 /* Floating point helpers */
 
@@ -395,6 +496,17 @@ typedef struct _formatdef {
 		    const struct _formatdef *);
 } formatdef;
 
+/* A large number of small routines follow, with names of the form
+
+	[bln][up]_TYPE
+
+   [bln] distiguishes among big-endian, little-endian and native.
+   [pu] distiguishes between pack (to struct) and unpack (from struct).
+   TYPE is one of char, byte, ubyte, etc.
+*/
+
+/* Native mode routines. */
+
 static PyObject *
 nu_char(const char *p, const formatdef *f)
 {
@@ -450,6 +562,34 @@ nu_ulong(const char *p, const formatdef *f)
 	return PyLong_FromUnsignedLong(*(unsigned long *)p);
 }
 
+/* Native mode doesn't support q or Q unless the platform C supports
+   long long (or, on Windows, __int64). */
+
+#ifdef HAVE_LONG_LONG
+
+static PyObject *
+nu_longlong(const char *p, const formatdef *f)
+{
+	return PyLong_FromLongLong(*(LONG_LONG *)p);
+}
+
+static PyObject *
+nu_ulonglong(const char *p, const formatdef *f)
+{
+	return PyLong_FromUnsignedLongLong(*(unsigned LONG_LONG *)p);
+}
+
+#else
+ 
+static PyObject *
+nu_qQerror(const char *p, const formatdef *f)
+{
+	PyErr_SetString(StructError, qQ_error_msg);
+	return NULL;
+}
+
+#endif
+
 static PyObject *
 nu_float(const char *p, const formatdef *f)
 {
@@ -585,6 +725,39 @@ np_ulong(char *p, PyObject *v, const formatdef *f)
 	return 0;
 }
 
+#ifdef HAVE_LONG_LONG
+
+static int
+np_longlong(char *p, PyObject *v, const formatdef *f)
+{
+	LONG_LONG x;
+	if (get_longlong(v, &x) < 0)
+		return -1;
+	* (LONG_LONG *)p = x;
+	return 0;
+}
+
+static int
+np_ulonglong(char *p, PyObject *v, const formatdef *f)
+{
+	unsigned LONG_LONG x;
+	if (get_ulonglong(v, &x) < 0)
+		return -1;
+	* (unsigned LONG_LONG *)p = x;
+	return 0;
+}
+
+#else
+
+static int
+np_qQerror(char *p, PyObject *v, const formatdef *f)
+{
+	PyErr_SetString(StructError, qQ_error_msg);
+ 	return -1;
+}
+
+#endif
+
 static int
 np_float(char *p, PyObject *v, const formatdef *f)
 {
@@ -642,6 +815,18 @@ static formatdef native_table[] = {
 	{'f',	sizeof(float),	FLOAT_ALIGN,	nu_float,	np_float},
 	{'d',	sizeof(double),	DOUBLE_ALIGN,	nu_double,	np_double},
 	{'P',	sizeof(void *),	VOID_P_ALIGN,	nu_void_p,	np_void_p},
+#ifdef HAVE_LONG_LONG
+	{'q',	sizeof(LONG_LONG), LONG_LONG_ALIGN, nu_longlong, np_longlong},
+	{'Q',	sizeof(LONG_LONG), LONG_LONG_ALIGN, nu_ulonglong,np_ulonglong},
+#else
+	/* n[pu]_qQerror just raise errors, but give them "the expected" size
+	   and alignment anyway so that calcsize returns something reasonable,
+	   and so unpack code that works on a 'long long' platform ends up in
+	   the error routine instead of with a mysterious "unpack str size
+	   does not match format" msg when run on a non-'long long' box. */
+	{'q',	8,		8, 		nu_qQerror,	 np_qQerror},
+	{'Q',	8,		8, 		nu_qQerror,	 np_qQerror},
+#endif
 	{0}
 };