pythongh-109523: Raise a BlockingIOError if reading text from a non-blocking stream cannot immediately return bytes. (pythonGH-122933)

Author: giosiragusa

Doc/library/io.rst

@@ -64,6 +64,12 @@ In-memory text streams are also available as :class:`StringIO` objects::
 
    f = io.StringIO("some initial text data")
 
+.. note::
+
+   When working with a non-blocking stream, be aware that read operations on text I/O objects
+   might raise a :exc:`BlockingIOError` if the stream cannot perform the operation
+   immediately.
+
 The text stream API is described in detail in the documentation of
 :class:`TextIOBase`.
 
@@ -770,6 +776,11 @@ than raw I/O does.
       Read and return *size* bytes, or if *size* is not given or negative, until
       EOF or if the read call would block in non-blocking mode.
 
+      .. note::
+
+         When the underlying raw stream is non-blocking, a :exc:`BlockingIOError`
+         may be raised if a read operation cannot be completed immediately.
+
    .. method:: read1(size=-1, /)
 
       Read and return up to *size* bytes with only one call on the raw stream.
@@ -779,6 +790,10 @@ than raw I/O does.
       .. versionchanged:: 3.7
          The *size* argument is now optional.
 
+      .. note::
+
+         When the underlying raw stream is non-blocking, a :exc:`BlockingIOError`
+         may be raised if a read operation cannot be completed immediately.
 
 .. class:: BufferedWriter(raw, buffer_size=DEFAULT_BUFFER_SIZE)
 
@@ -1007,6 +1022,11 @@ Text I/O
    .. versionchanged:: 3.10
       The *encoding* argument now supports the ``"locale"`` dummy encoding name.
 
+   .. note::
+
+      When the underlying raw stream is non-blocking, a :exc:`BlockingIOError`
+      may be raised if a read operation cannot be completed immediately.
+
    :class:`TextIOWrapper` provides these data attributes and methods in
    addition to those from :class:`TextIOBase` and :class:`IOBase`:
 

Doc/whatsnew/3.14.rst

@@ -404,6 +404,15 @@ inspect
   (Contributed by Zhikang Yan in :gh:`125634`.)
 
 
+
+io
+--
+
+* Reading text from a non-blocking stream with ``read`` may now raise a
+  :exc:`BlockingIOError` if the operation cannot immediately return bytes.
+  (Contributed by Giovanni Siragusa in :gh:`109523`.)
+
+
 json
 ----
 

Lib/_pyio.py

@@ -2545,9 +2545,12 @@ def read(self, size=None):
                 size = size_index()
         decoder = self._decoder or self._get_decoder()
         if size < 0:
+            chunk = self.buffer.read()
+            if chunk is None:
+                raise BlockingIOError("Read returned None.")
             # Read everything.
             result = (self._get_decoded_chars() +
-                      decoder.decode(self.buffer.read(), final=True))
+                      decoder.decode(chunk, final=True))
             if self._snapshot is not None:
                 self._set_decoded_chars('')
                 self._snapshot = None

Lib/test/test_io.py

@@ -3932,6 +3932,22 @@ def test_issue35928(self):
         f.write(res)
         self.assertEqual(res + f.readline(), 'foo\nbar\n')
 
+    @unittest.skipUnless(hasattr(os, "pipe"), "requires os.pipe()")
+    def test_read_non_blocking(self):
+        import os
+        r, w = os.pipe()
+        try:
+            os.set_blocking(r, False)
+            with self.io.open(r, 'rt') as textfile:
+                r = None
+                # Nothing has been written so a non-blocking read raises a BlockingIOError exception.
+                with self.assertRaises(BlockingIOError):
+                    textfile.read()
+        finally:
+            if r is not None:
+                os.close(r)
+            os.close(w)
+
 
 class MemviewBytesIO(io.BytesIO):
     '''A BytesIO object whose read method returns memoryviews

Misc/ACKS

@@ -1736,6 +1736,7 @@ Ng Pheng Siong
 Yann Sionneau
 George Sipe
 J. Sipprell
+Giovanni Siragusa
 Ngalim Siregar
 Kragen Sitaker
 Kaartic Sivaraam

Misc/NEWS.d/next/C_API/2024-08-12-10-15-19.gh-issue-109523.S2c3fi.rst

@@ -0,0 +1 @@
+Reading text from a non-blocking stream with ``read`` may now raise a :exc:`BlockingIOError` if the operation cannot immediately return bytes.

Modules/_io/textio.c

@@ -1992,6 +1992,12 @@ _io_TextIOWrapper_read_impl(textio *self, Py_ssize_t n)
         if (bytes == NULL)
             goto fail;
 
+        if (bytes == Py_None){
+            Py_DECREF(bytes);
+            PyErr_SetString(PyExc_BlockingIOError, "Read returned None.");
+            return NULL;
+        }
+
         _PyIO_State *state = self->state;
         if (Py_IS_TYPE(self->decoder, state->PyIncrementalNewlineDecoder_Type))
             decoded = _PyIncrementalNewlineDecoder_decode(self->decoder,