pythongh-121645: Add PyBytes_Join() function

Author: vstinner

Doc/c-api/bytes.rst

@@ -189,6 +189,16 @@ called with a non-bytes parameter.
    to *newpart* (i.e. decrements its reference count).
 
 
+.. c:function:: PyObject* PyBytes_Join(PyObject *sep, PyObject *iterable)
+
+    Similar to ``sep.join(iterable)`` in Python.
+
+    *sep* must be Python :class:`bytes` object.
+    *iterable* must be an iterable object.
+
+   .. versionadded: 3.14
+
+
 .. c:function:: int _PyBytes_Resize(PyObject **bytes, Py_ssize_t newsize)
 
    Resize a bytes object. *newsize* will be the new length of the bytes object.

Doc/whatsnew/3.14.rst

@@ -357,6 +357,10 @@ New Features
 
   (Contributed by Victor Stinner in :gh:`119182`.)
 
+* Add :c:func:`PyBytes_Join(seq, iterable) <PyBytes_Join>` function,
+  similar to ``sep.join(iterable)`` in Python.
+  (Contributed by Victor Stinner in :gh:`121645`.)
+
 Porting to Python 3.14
 ----------------------
 

Include/cpython/bytesobject.h

@@ -31,3 +31,7 @@ static inline Py_ssize_t PyBytes_GET_SIZE(PyObject *op) {
     return Py_SIZE(self);
 }
 #define PyBytes_GET_SIZE(self) PyBytes_GET_SIZE(_PyObject_CAST(self))
+
+// PyBytes_Join(sep, x) is like sep.join(x).  sep must be PyBytesObject*,
+// x must be an iterable object.
+PyAPI_FUNC(PyObject*) PyBytes_Join(PyObject *sep, PyObject *x);

Include/internal/pycore_bytesobject.h

@@ -23,10 +23,6 @@ extern PyObject* _PyBytes_FromHex(
 PyAPI_FUNC(PyObject*) _PyBytes_DecodeEscape(const char *, Py_ssize_t,
                                             const char *, const char **);
 
-/* _PyBytes_Join(sep, x) is like sep.join(x).  sep must be PyBytesObject*,
-   x must be an iterable object. */
-extern PyObject* _PyBytes_Join(PyObject *sep, PyObject *x);
-
 
 // Substring Search.
 //

Lib/test/test_capi/test_bytes.py

@@ -247,6 +247,29 @@ def test_resize(self):
         # CRASHES resize(NULL, 0, False)
         # CRASHES resize(NULL, 3, False)
 
+    def test_join(self):
+        """Test PyBytes_Join()"""
+        bytes_join = _testcapi.bytes_join
+
+        self.assertEqual(bytes_join(b'', []), b'')
+        self.assertEqual(bytes_join(b'sep', []), b'')
+
+        self.assertEqual(bytes_join(b'', [b'a', b'b', b'c']), b'abc')
+        self.assertEqual(bytes_join(b'-', [b'a', b'b', b'c']), b'a-b-c')
+        self.assertEqual(bytes_join(b' - ', [b'a', b'b', b'c']), b'a - b - c')
+
+        # invalid 'sep' argument type
+        with self.assertRaises(TypeError):
+            bytes_join('unicode', [])
+        with self.assertRaises(TypeError):
+            bytes_join(123, [])
+
+        # invalid 'iterable' argument type
+        with self.assertRaises(TypeError):
+            bytes_join(b'', [b'bytes', 'unicode'])
+        with self.assertRaises(TypeError):
+            bytes_join(b'', [b'bytes', 123])
+
 
 if __name__ == "__main__":
     unittest.main()

Misc/NEWS.d/next/C_API/2024-07-12-13-40-59.gh-issue-121645.96QvD3.rst

@@ -0,0 +1,2 @@
+Add :c:func:`PyBytes_Join(seq, iterable) <PyBytes_Join>` function, similar to
+``sep.join(iterable)`` in Python. Patch by Victor Stinner.

Modules/_io/bufferedio.c

@@ -8,7 +8,6 @@
 */
 
 #include "Python.h"
-#include "pycore_bytesobject.h"         // _PyBytes_Join()
 #include "pycore_call.h"                // _PyObject_CallNoArgs()
 #include "pycore_object.h"              // _PyObject_GC_UNTRACK()
 #include "pycore_pyerrors.h"            // _Py_FatalErrorFormat()
@@ -1284,7 +1283,7 @@ _buffered_readline(buffered *self, Py_ssize_t limit)
         Py_CLEAR(res);
         goto end;
     }
-    Py_XSETREF(res, _PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), chunks));
+    Py_XSETREF(res, PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), chunks));
 
 end:
     LEAVE_BUFFERED(self)
@@ -1737,7 +1736,7 @@ _bufferedreader_read_all(buffered *self)
                 goto cleanup;
             }
             else {
-                tmp = _PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), chunks);
+                tmp = PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), chunks);
                 res = tmp;
                 goto cleanup;
             }

Modules/_io/iobase.c

@@ -999,7 +999,7 @@ _io__RawIOBase_readall_impl(PyObject *self)
             return NULL;
         }
     }
-    result = _PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), chunks);
+    result = PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), chunks);
     Py_DECREF(chunks);
     return result;
 }

Modules/_sre/sre.c

@@ -1287,7 +1287,7 @@ pattern_subx(_sremodulestate* module_state,
     }
     else {
         if (state.isbytes)
-            item = _PyBytes_Join(joiner, list);
+            item = PyBytes_Join(joiner, list);
         else
             item = PyUnicode_Join(joiner, list);
         Py_DECREF(joiner);
@@ -2918,7 +2918,7 @@ expand_template(TemplateObject *self, MatchObject *match)
     }
     else {
         Py_SET_SIZE(list, count);
-        result = _PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), list);
+        result = PyBytes_Join((PyObject *)&_Py_SINGLETON(bytes_empty), list);
     }
 
 cleanup:

Modules/_testcapi/bytes.c

@@ -37,8 +37,21 @@ bytes_resize(PyObject *Py_UNUSED(module), PyObject *args)
 }
 
 
+/* Test PyBytes_Join() */
+static PyObject *
+bytes_join(PyObject *Py_UNUSED(module), PyObject *args)
+{
+    PyObject *sep, *seq;
+    if (!PyArg_ParseTuple(args, "OO", &sep, &seq)) {
+        return NULL;
+    }
+    return PyBytes_Join(sep, seq);
+}
+
+
 static PyMethodDef test_methods[] = {
     {"bytes_resize", bytes_resize, METH_VARARGS},
+    {"bytes_join", bytes_join, METH_VARARGS},
     {NULL},
 };
 

Objects/bytesobject.c

@@ -1867,10 +1867,15 @@ bytes_join(PyBytesObject *self, PyObject *iterable_of_bytes)
 }
 
 PyObject *
-_PyBytes_Join(PyObject *sep, PyObject *x)
+PyBytes_Join(PyObject *sep, PyObject *x)
 {
-    assert(sep != NULL && PyBytes_Check(sep));
+    assert(sep != NULL);
     assert(x != NULL);
+    if (!PyBytes_Check(sep)) {
+        PyErr_Format(PyExc_TypeError,
+                     "expected bytes for sep argument, got %T", sep);
+        return NULL;
+    }
     return bytes_join((PyBytesObject*)sep, x);
 }