gh-153785: Generate `AttributeError` messages from context by johnslavik · Pull Request #153786 · python/cpython

johnslavik · 2026-07-16T01:03:13Z

Motivation

Code very often raises AttributeError(name) or, more rarely, a bare AttributeError from __getattr__ (there are ~40 such sites in CPython itself). In both cases the exception's obj and name attributes are populated automatically by _PyObject_SetAttributeErrorContext, but the message rendered in the traceback ignores them:

raise AttributeError(name) shows just the attribute name, losing which object it was accessed on.
bare raise AttributeError shows nothing at all.

So today the message throws away context that the interpreter already has on hand (it uses obj/name for the "Did you mean …?" suggestions, but not for the message itself).

This also produces the garbled output reported in #143811, where a bare AttributeError renders as an empty message glued to a suggestion:

AttributeError: . Did you mean: 'foobar'?

Change

This gives AttributeError a custom __str__ (AttributeError_str) that synthesizes the familiar message from the collected obj/name context:

AttributeError: '<qualname>' object has no attribute 'missing'
AttributeError: module '<name>' has no attribute 'missing'

It is the same trick KeyError already uses to wrap its key in repr() — see KeyError_str. Doing it in AttributeError.__str__ (rather than in traceback) means every consumer of the message benefits, not just tracebacks.

To avoid clobbering messages the caller actually meant, the generated message is only used when obj and name are both set and the exception was constructed either with no positional arguments or with a single positional argument equal to name. An explanatory string passed by the caller (e.g. AttributeError("custom")) is preserved unchanged.

As a side effect this also fixes #143811, which motivated the current issue — the functools.singledispatchmethod case now reports a proper message instead of the garbled one.

Issue: Generate AttributeError messages from their context #153785

read-the-docs-community · 2026-07-16T01:07:23Z

Documentation build overview

📚 cpython-previews | 🛠️ Build #33647017 | 📁 Comparing 9b00321 against main (b4662e8)

🔍 Preview build

21 files changed · ± 21 modified

± Modified

I'm not sure how to create a nameless module at runtime

This reverts commit 7378e30.

pablogsal · 2026-07-18T15:25:28Z

+            result = PyUnicode_FromFormat("module has no attribute '%U'", name);
+        }
+    } else {
+        result = PyUnicode_FromFormat("'%T' object has no attribute '%U'",


This formats class objects as instances of their metaclass. Can we preserve the standard type object 'C' has no attribute ... form here?

pablogsal · 2026-07-18T15:25:28Z

+    if (PyModule_Check(obj)) {
+        PyModuleObject *mod = _PyModule_CAST(obj);
+        /* In a typical case, module's __name__ is examined instead. */
+        if (mod->md_name) {


This uses cached md_name, so assigning module.__name__ makes the generated message report the old name. We should read the current module name here.

pablogsal · 2026-07-18T15:25:28Z

+
+    PyObject *result;
+    if (PyModule_Check(obj)) {
+        PyModuleObject *mod = _PyModule_CAST(obj);


Nit: we use _PyModule_CAST and md_name here but only get pycore_moduleobject.h transitively through pycore_object.h. Can we include it explicitly like the other headers?

pablogsal · 2026-07-18T15:25:28Z


+   When possible, :attr:`name` and :attr:`obj` are set automatically.
+
   .. versionchanged:: 3.10


We dropped the .. versionchanged:: next note that documented the generated message. This is a user-visible change to str(), so I think we should keep a versionchanged here, no?

vstinner

The following code writes 'lazy_import' object has no attribute 'missing1': it mentions lazy_import rather than module 'asyncio'. But I don't think that AttributeError_str() would be the right place to reify the asyncio module!

lazy import asyncio

class RaiseWithName:
    def __getattr__(self, name):
        raise AttributeError(name)

try:
    getattr(globals()['asyncio'], "missing1")
except AttributeError as exc:
#    assert exc.obj is ...
    assert exc.name == "missing1"
    print(str(exc))

vstinner · 2026-07-18T15:28:35Z


      The object that was accessed for the named attribute.

+   When possible, :attr:`name` and :attr:`obj` are set automatically.


This change surprised me but it's correct.

This addition surprised me since the PR doesn't change that. The change documents that _PyObject_SetAttributeErrorContext() is called by many functions such as PyObject_GetAttr() to set name and obj attributes of an AttributeError.

vstinner · 2026-07-18T15:30:46Z

+        && ((PyTuple_GET_SIZE(self->args) == 1
+             && PyUnicode_Check(arg = PyTuple_GET_ITEM(self->args, 0))
+             && _PyUnicode_Equal(arg, self->name))
+            || PyTuple_GET_SIZE(self->args) == 0)


Can you please add a comment explaining why you use exc.name only if it's equal to exc.args[0] or if exc.args is empty? IMO it's a good thing to implement this change, but it wasn't obvious the first time that I read this code.

vstinner · 2026-07-18T15:43:11Z

+AttributeError_str(PyObject *op)
+{
+    PyAttributeErrorObject *self = PyAttributeErrorObject_CAST(op);
+    PyObject *arg, *obj = NULL, *name;


I dislike the arg = ... assignment in the middle of the big if condition, but you can keep it if you like it.

Suggested change

PyObject *arg, *obj = NULL, *name;

PyObject *arg; // borrowed ref

PyObject *obj = NULL, *name = NULL;

vstinner · 2026-07-18T15:43:32Z

+                                          mod->md_name,
+                                          name);


Suggested change

mod->md_name,

name);

mod->md_name, name);

vstinner · 2026-07-18T15:50:31Z

            self.assertEqual("bluch", exc.name)
            self.assertEqual(obj, exc.obj)

+    def test_getattr_error_message(self):


Please check also obj and name attributes of AttributeError:

Details

def test_getattr_error_message(self): def fqn(type): return f'{type.__module__}.{type.__qualname__}' class RaiseWithName: def __getattr__(self, name): raise AttributeError(name) obj = RaiseWithName() with self.assertRaises(AttributeError) as cm: getattr(obj, "missing1") self.assertEqual(str(cm.exception), f"'{fqn(RaiseWithName)}' object has no attribute 'missing1'") self.assertIs(cm.exception.obj, obj) self.assertIs(cm.exception.name, "missing1") class BareRaise: def __getattr__(self, name): raise AttributeError obj = BareRaise() with self.assertRaises(AttributeError) as cm: getattr(obj, "missing2") self.assertEqual(str(cm.exception), f"'{fqn(BareRaise)}' object has no attribute 'missing2'") self.assertIs(cm.exception.obj, obj) self.assertIs(cm.exception.name, "missing2") class RaiseCustom: def __getattr__(self, name): raise AttributeError("custom") obj = RaiseCustom() with self.assertRaises(AttributeError) as cm: getattr(obj, "missing3") self.assertEqual(str(cm.exception), "custom") self.assertIs(cm.exception.obj, obj) self.assertIs(cm.exception.name, "missing3") def test_module_getattr_error_message(self): raisewithname_mod = ModuleType("raisewithname") def raise_with_name(name): raise AttributeError(name) raisewithname_mod.__getattr__ = raise_with_name with self.assertRaises(AttributeError) as cm: getattr(raisewithname_mod, "missing1") self.assertEqual(str(cm.exception), "module 'raisewithname' has no attribute 'missing1'") self.assertIs(cm.exception.obj, raisewithname_mod) self.assertIs(cm.exception.name, "missing1") bareraise_mod = ModuleType("bareraise") def bare_raise(name): raise AttributeError bareraise_mod.__getattr__ = bare_raise with self.assertRaises(AttributeError) as cm: getattr(bareraise_mod, "missing2") self.assertEqual(str(cm.exception), "module 'bareraise' has no attribute 'missing2'") self.assertIs(cm.exception.obj, bareraise_mod) self.assertIs(cm.exception.name, "missing2") custom_mod = ModuleType("custom") def raise_custom(name): raise AttributeError("custom") custom_mod.__getattr__ = raise_custom with self.assertRaises(AttributeError) as cm: getattr(custom_mod, "missing3") self.assertEqual(str(cm.exception), "custom") self.assertIs(cm.exception.obj, custom_mod) self.assertIs(cm.exception.name, "missing4") nameless_mod = ModuleType.__new__(ModuleType) nameless_mod.__getattr__ = raise_with_name with self.assertRaises(AttributeError) as cm: getattr(nameless_mod, "missing4") self.assertEqual(str(cm.exception), "module has no attribute 'missing4'") self.assertIs(cm.exception.obj, nameless_mod) self.assertIs(cm.exception.name, "missing4")

vstinner · 2026-07-18T15:52:21Z

+The default error message is now generated from ``name`` and ``obj`` attributes
+of :exc:`AttributeError` when both are set and the exception was constructed with


Suggested change

The default error message is now generated from ``name`` and ``obj`` attributes

of :exc:`AttributeError` when both are set and the exception was constructed with

:exc:`AttributeError`: The default error message is now generated from ``name``

and ``obj`` attributes when both are set and the exception was constructed with

jaraco · 2026-07-18T16:17:06Z

I think this is the best place to do this. We do the same trick in KeyError.

This also fixed issue #143811 that motivated the current one.
* Issue: [Generate `AttributeError` messages from their context #153785](https://github.com/python/cpython/issues/153785)

"this" and "current one" is unclear here. Please update the PR description to describe the motivation. Also consider linking to the code for the trick in KeyError.

jaraco · 2026-07-18T16:21:36Z

I think this is the best place to do this. We do the same trick in KeyError.
This also fixed issue #143811 that motivated the current one.
* Issue: [Generate `AttributeError` messages from their context #153785](https://github.com/python/cpython/issues/153785)
"this" and "current one" is unclear here. Please update the PR description to describe the motivation. Also consider linking to the code for the trick in KeyError.

I used Claude to infer the motivations and update the description.

jaraco · 2026-07-18T16:27:48Z

It is the same trick KeyError already uses to wrap its key in repr() — see KeyError_str.

This trick has been applied to KeyError and AttributeError; should it be applied to other exceptions? Should it be implemented in such a way that the concern is shared across exceptions (explicitly or implicitly)?

jaraco · 2026-07-18T16:31:05Z

This trick has been applied to KeyError and AttributeError; should it be applied to other exceptions? Should it be implemented in such a way that the concern is shared across exceptions (explicitly or implicitly)?

Short answer, not really.

Details

(Analysis by Claude, on my behalf.)

A worthwhile question. Worth noting first that the two "tricks" only look alike at the surface (both override tp_str); underneath they differ:

KeyError_str reformats what's already in args — wrapping args[0] in repr() so {}[''] prints KeyError: ''. Pure presentation.
AttributeError_str synthesizes a message from side-channel context (obj/name) that the raiser never put in args, reconstructing what the interpreter attached out-of-band.

Only the second is the interesting reusable pattern ("I have structured context attributes; render them lazily instead of forcing every raise site to build the string"). The natural cohort there is NameError/UnboundLocalError (name), ImportError/ModuleNotFoundError (name, path). But both of those already build their full messages eagerly at the C raise site, so the problem AttributeError has — user __getattr__ implementations conventionally raising bare or name-only — barely exists for them.

So I'd lean against building a shared framework here:

Two thin, differing data points; abstracting them would abstract over a coincidence of mechanism, not a shared concern.
The templates ("object has no attribute" vs. "name is not defined" vs. "cannot import name from") are irreducibly per-type — only the dispatch around them could be shared.
Every __str__ change is a compatibility surface (people do match on message text), and each type wants its own guard about when overriding is safe.

The one genuinely reusable piece is the guard predicate — "was this constructed with a default/absent message, or one that merely duplicates a structured field, so I'm safe to override?" (here: args of size 0, or size 1 equal to name). If anything gets factored out, I'd make it that small internal helper and leave each template local. Otherwise I'd revisit sharing only when a third genuine case actually lands.

johnslavik added 7 commits July 16, 2026 02:15

Implement AttributeError_str()

ac555bc

Add tests

4f22a16

Check name too

b30d676

Add docs

28e04d3

Add news entry

e5c4922

Improve formatting

3854563

Simplify?

1e88331

johnslavik requested a review from serhiy-storchaka July 16, 2026 01:03

johnslavik requested a review from iritkatriel as a code owner July 16, 2026 01:03

This comment was marked as resolved.

Sign in to view

bedevere-app Bot added the awaiting review label Jul 16, 2026

johnslavik changed the title ~~Generate AttributeError messages from context~~ gh-153785: Generate AttributeError messages from context Jul 16, 2026

bedevere-app Bot mentioned this pull request Jul 16, 2026

Generate AttributeError messages from their context #153785

Open

Cooler formatting of the news entry

2fd216a

johnslavik added 4 commits July 16, 2026 03:13

Test custom messages aren't overwritten

356f56e

Fix news problem

8db0394

Shorter test name

ebcc795

Fix inconsistent style

b6fda10

johnslavik requested a review from DinoV July 16, 2026 01:23

johnslavik marked this pull request as draft July 16, 2026 02:11

bedevere-app Bot removed the awaiting review label Jul 16, 2026

Make thread-safe, presumably

d624835

johnslavik marked this pull request as ready for review July 16, 2026 02:21

bedevere-app Bot added the awaiting review label Jul 16, 2026

johnslavik added 5 commits July 16, 2026 04:23

Rarest case goes last

c50a356

} else {

7ad5204

Remove unnecessary whitespace

b65a7fd

Fix style

e99d875

Avoid nesting critical sections!

8c29de3

johnslavik added 6 commits July 17, 2026 11:11

Truncate AttributeError.obj representation to 200 chars

e905ab2

Remove extra empty line

7d720ac

Add "re-acquires lock" comment

337cef4

I desperately need a formatter

6aa47db

Missing5.

4ff6a48

.

e00db5f

serhiy-storchaka reviewed Jul 17, 2026

View reviewed changes

Comment thread Objects/exceptions.c Outdated

Comment thread Objects/exceptions.c Outdated

johnslavik added 6 commits July 17, 2026 11:35

Use _PyUnicode_Equal

b5ce3c9

Use typename or module name

da8c723

Rearrange tests

c941657

I'm not sure how to create a nameless module at runtime

Cover nameless module branch

9ea6c8a

Clearer variable names

f825f27

Localize tests visually

7b73fb8

johnslavik requested a review from pablogsal July 17, 2026 10:15

serhiy-storchaka reviewed Jul 17, 2026

View reviewed changes

Comment thread Objects/exceptions.c Outdated

johnslavik added 2 commits July 17, 2026 17:42

Use %T

7378e30

Revert "Use %T"

faf6e53

This reverts commit 7378e30.

johnslavik added the sprint label Jul 18, 2026

github-project-automation Bot added this to Sprint Jul 18, 2026

github-project-automation Bot moved this to Todo in Sprint Jul 18, 2026

github-project-automation Bot moved this from Todo to In Progress in Sprint Jul 18, 2026

Use '%T'

a86fde5

vstinner reviewed Jul 18, 2026

View reviewed changes

Comment thread Lib/test/test_exceptions.py Outdated

Test for fully qualified names

9b00321

pablogsal reviewed Jul 18, 2026

View reviewed changes

vstinner reviewed Jul 18, 2026

View reviewed changes


		When possible, :attr:`name` and :attr:`obj` are set automatically.

		.. versionchanged:: 3.10


		The object that was accessed for the named attribute.

		When possible, :attr:`name` and :attr:`obj` are set automatically.

	PyObject arg, obj = NULL, *name;
	PyObject *arg; // borrowed ref
	PyObject obj = NULL, name = NULL;

		The default error message is now generated from ``name`` and ``obj`` attributes
		of :exc:`AttributeError` when both are set and the exception was constructed with

Uh oh!

Uh oh!

Conversation

johnslavik commented Jul 16, 2026 • edited by jaraco Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Change

Uh oh!

This comment was marked as resolved.

read-the-docs-community Bot commented Jul 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Documentation build overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vstinner left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jaraco commented Jul 18, 2026

Uh oh!

jaraco commented Jul 18, 2026

Uh oh!

jaraco commented Jul 18, 2026

Uh oh!

jaraco commented Jul 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

johnslavik commented Jul 16, 2026 •

edited by jaraco

Loading

read-the-docs-community Bot commented Jul 16, 2026 •

edited

Loading

jaraco commented Jul 18, 2026 •

edited

Loading