gh-129205: Add os.readinto API for reading data into a caller provided buffer

Lib/test/_test_eintr.py

@@ -152,6 +152,37 @@ def test_read(self):
                 self.assertEqual(data, os.read(rd, len(data)))
             self.assertEqual(proc.wait(), 0)
 
+    def test_readinto(self):
+        rd, wr = os.pipe()
+        self.addCleanup(os.close, rd)
+        # wr closed explicitly by parent
+
+        # the payload below are smaller than PIPE_BUF, hence the writes will be
+        # atomic
+        datas = [b"hello", b"world", b"spam"]
+        bufs = [bytearray(5), bytearray(5), bytearray(4)]
+
+        code = '\n'.join((
+            'import os, sys, time',
+            '',
+            'wr = int(sys.argv[1])',
+            'datas = %r' % datas,
+            'sleep_time = %r' % self.sleep_time,
+            '',
+            'for data in datas:',
+            '    # let the parent block on read()',
+            '    time.sleep(sleep_time)',
+            '    os.write(wr, data)',
+        ))
+
+        proc = self.subprocess(code, str(wr), pass_fds=[wr])
+        with kill_on_error(proc):
+            os.close(wr)
+            for data, buffer in zip(datas, bufs):
+                os.readinto(rd, buffer)
+                self.assertEqual(data, buffer)
+            self.assertEqual(proc.wait(), 0)
+
     def test_write(self):
         rd, wr = os.pipe()
         self.addCleanup(os.close, wr)

Lib/test/test_os.py

@@ -230,6 +230,54 @@ def test_read(self):
             self.assertEqual(type(s), bytes)
             self.assertEqual(s, b"spam")
 
+    def test_readinto(self):
+        with open(os_helper.TESTFN, "w+b") as fobj:
+            fobj.write(b"spam")
+            fobj.flush()
+            fd = fobj.fileno()
+            os.lseek(fd, 0, 0)
+            buffer = bytearray(8)
+            s = os.readinto(fd, buffer)
+            self.assertEqual(type(s), int)
+            self.assertEqual(s, 4)
+            # Should overwrite the first 4 bytes of the buffer.
+            self.assertEqual(bytes(buffer), b"spam\0\0\0\0")
+
+            # Readinto at EOF shold return 0 and not touch buffer
+            buffer[:] = b"notspam\0"
+            s = os.readinto(fd, buffer)
+            self.assertEqual(type(s), int)
+            self.assertEqual(s, 0)
+            self.assertEqual(bytes(buffer), b"notspam\0")
+            s = os.readinto(fd, buffer)
+            self.assertEqual(s, 0)
+            self.assertEqual(bytes(buffer), b"notspam\0")
+
+            # Readinto a 0 length bytearray when at EOF should return 0
+            self.assertEqual(os.readinto(fd, bytearray()), 0)
+
+            # Readinto a 0 length bytearray with data available should return 0.
+            os.lseek(fd, 0, 0)
+            self.assertEqual(os.readinto(fd, bytearray()), 0)
+
+    def test_readinto_badbuffer(self):
+        with open(os_helper.TESTFN, "w+b") as fobj:
+            fobj.write(b"spam")
+            fobj.flush()
+            fd = fobj.fileno()
+            os.lseek(fd, 0, 0)
+
+            for bad_arg in ("test", bytes(), 14):
+                with self.subTest(f"{type(bad_arg)}"):
+                    with self.assertRaises(TypeError):
+                        os.readinto(fd, bad_arg)
+
+            # No data should have been read with the bad arguments.
+            buffer = bytearray(8)
+            s = os.readinto(fd, buffer)
+            self.assertEqual(s, 4)
+            self.assertEqual(bytes(buffer), b"spam\0\0\0\0")
+
     @support.cpython_only
     # Skip the test on 32-bit platforms: the number of bytes must fit in a
     # Py_ssize_t type
@@ -249,6 +297,29 @@ def test_large_read(self, size):
         # operating system is free to return less bytes than requested.
         self.assertEqual(data, b'test')
 
+
+    @support.cpython_only
+    # Skip the test on 32-bit platforms: the number of bytes must fit in a
+    # Py_ssize_t type
+    @unittest.skipUnless(INT_MAX < PY_SSIZE_T_MAX,
+                         "needs INT_MAX < PY_SSIZE_T_MAX")
+    @support.bigmemtest(size=INT_MAX + 10, memuse=1, dry_run=False)
+    def test_large_readinto(self, size):
+        self.addCleanup(os_helper.unlink, os_helper.TESTFN)
+        create_file(os_helper.TESTFN, b'test')
+
+        # Issue #21932: For readinto the buffer contains the length rather than
+        # a length being passed explicitly to read, should still get capped to a
+        # valid size / not raise an OverflowError for sizes larger than INT_MAX.
+        buffer = bytearray(INT_MAX+10)
+        with open(os_helper.TESTFN, "rb") as fp:
+            length = os.readinto(fp.fileno(), buffer)
+
+        # The test does not try to read more than 2 GiB at once because the
+        # operating system is free to return less bytes than requested.
+        self.assertEqual(length, 4)
+        self.assertEqual(buffer[:4], b'test')
+
     def test_write(self):
         # os.write() accepts bytes- and buffer-like objects but not strings
         fd = os.open(os_helper.TESTFN, os.O_CREAT | os.O_WRONLY)
@@ -2467,6 +2538,10 @@ def test_lseek(self):
     def test_read(self):
         self.check(os.read, 1)
 
+    @unittest.skipUnless(hasattr(os, 'readinto'), 'test needs os.readinto()')
+    def test_readinto(self):
+        self.check(os.readinto, bytearray(5))
+
     @unittest.skipUnless(hasattr(os, 'readv'), 'test needs os.readv()')
     def test_readv(self):
         buf = bytearray(10)

Misc/NEWS.d/next/Library/2025-01-22-16-54-25.gh-issue-129205.FMqrUt.rst

@@ -0,0 +1 @@
+Add ``os.readinto`` to read into a :ref:`buffer protocol <bufferobjects>` from a file descriptor.

Modules/posixmodule.c

@@ -11433,6 +11433,32 @@ os_read_impl(PyObject *module, int fd, Py_ssize_t length)
     return buffer;
 }
 
+/*[clinic input]
+os.readinto -> Py_ssize_t
+    fd: int
+    buffer: Py_buffer(accept={rwbuffer})
+    /
+
+Read into a :ref:`buffer protocol <bufferobjects>` object from a file descriptor.
+
+The buffer should be mutable and accept bytes. On success, returns the number of
+bytes read. Less bytes may be read than the size of the buffer. Will retry the
+underlying system call when interrupted by a signal. For other errors, the
+system call will not be retried.
+[clinic start generated code]*/
+
+static Py_ssize_t
+os_readinto_impl(PyObject *module, int fd, Py_buffer *buffer)
+/*[clinic end generated code: output=8091a3513c683a80 input=810c820f4d9b1c6b]*/
+{
+    if (buffer->len < 0) {
+        errno = EINVAL;
+        PyErr_SetFromErrno(PyExc_OSError);
+        return -1;
+    }
+    return _Py_read(fd, buffer->buf, buffer->len);
+}
+
 #if (defined(HAVE_SENDFILE) && (defined(__FreeBSD__) || defined(__DragonFly__) \
                                 || defined(__APPLE__))) \
     || defined(HAVE_READV) || defined(HAVE_PREADV) || defined (HAVE_PREADV2) \
@@ -16973,6 +16999,7 @@ static PyMethodDef posix_methods[] = {
     OS_LOCKF_METHODDEF
     OS_LSEEK_METHODDEF
     OS_READ_METHODDEF
+    OS_READINTO_METHODDEF
     OS_READV_METHODDEF
     OS_PREAD_METHODDEF
     OS_PREADV_METHODDEF