python / python/cpython

Exception notes discarded from TypeError within __hash__()

Aperta
#149,313 1 commento 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

interpreter-core type-bug
Lingua principale
Python
Stelle
77.2k
Fork
35.9k
Metriche di merge delle PR
Metriche PR in attesa

Descrizione

Bug report

Bug description:

Context

I'm maintaining this library that requires users to construct deeply nested structures that are used as dict keys. To alleviate the annoyance of figuring out where a non-hashable value has been introduced into the key structure, I started using dedicated data structures with a __hash__(), which adds location information to the exception's notes:

from dataclasses import dataclass
from dataclasses import field

@dataclass
class KeyStruct:
    name: str
    first: object = field(repr=False)
    second: object = field(repr=False)

    def __hash__(self) -> int:
        try:
            h1 = hash(self.first)
        except TypeError as e:
            e.add_note(f"within {self}.first")
            raise

        try:
            h2 = hash(self.second)
        except TypeError as e:
            e.add_note(f"within {self}.second")
            raise

        return hash((self.name, h1, h2))

So when such a structure is used, information about the location of the offending value can be provided:

s = KeyStruct("foo", 1, KeyStruct("bar", 2, KeyStruct("spam", 3, [])))
s_dict = {s: "hello"}
Traceback (most recent call last):
  File "/Users/michi/Documents/Projects/Declarative_Python/declarative-python/src/hello.py", line 43, in <module>
    s_dict = {s: "hello"}
             ^^^^^^^^^^^^
  File "/Users/michi/Documents/Projects/Declarative_Python/declarative-python/src/hello.py", line 34, in __hash__
    h2 = hash(self.second)
  File "/Users/michi/Documents/Projects/Declarative_Python/declarative-python/src/hello.py", line 34, in __hash__
    h2 = hash(self.second)
  File "/Users/michi/Documents/Projects/Declarative_Python/declarative-python/src/hello.py", line 28, in __hash__
    h1 = hash(self.first)
TypeError: unhashable type: 'list'
within KeyStruct(name='spam').first
within KeyStruct(name='bar').second
within KeyStruct(name='foo').second

It is not very beautiful, but very helpful. :)

The Bug

The above output is from Python 3.13. The output changes with Python 3.14 (3.14.4 on macOS 14.8.5):

Traceback (most recent call last):
  File "/Users/michi/Documents/Projects/Declarative_Python/declarative-python/src/hello.py", line 43, in <module>
    s_dict = {s: "hello"}
             ^^^^^^^^^^^^
TypeError: cannot use 'KeyStruct' as a dict key (unhashable type: 'list')

I suspect that the source of this is the improved error messages introduced with https://github.com/python/cpython/issues/132825. If I interpret this code correctly, the dict and set methods that need to hash a key will catch and discard the raised TypeError and only re-use the exception message.

IMHO his is unfortunate and I consider it a regression.

I'm not sure what the best way to solve this would be:

  • As an experienced Python programmer, I would have expected the implementation of dict and set to attach the existing exception as __cause__ on the new exception. That would solve my use case.

  • An alternative would be to copy the notes from the old to the new exception. If you consider the notes a part of something I'll call the "extended exception message", the code would conceptually replace the exception, modifying, but retaining its "extended exception message" in full.

    But this sill has the drawback of discarding the original stack trace, which I also consider valuable in many cases.

  • Update the passage describing the improvement in What’s new in Python 3.14 to document the change in behavior and provide a workaround. E.g.:

    • Improved error message when trying to add an instance of an unhashable type to a dict or set. (Contributed by CF Bolz-Tereick and Victor Stinner in gh-132828.)

      (code block)

      In those instances, the original TypeError and its notes and stack trace are replaced. You can raise a subclass of TypeError from a __hash__() implementation to opt-out of this feature.

    AFAICT, the improvement is not documented elsewhere in Python's documentation.

CPython versions tested on:

3.14

Operating systems tested on:

macOS

Linked PRs
  • gh-149505

Guida per i contributori

Apri la guida per i contributori

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Direzione di ricerca

Inizia esaminando i metodi di hashing di dict e set coinvolti nel report, insieme all'entry point hash e alla PR collegata gh-149505. Confronta il comportamento di Python 3.13 e 3.14, incluse le note sulle eccezioni e la gestione dei traceback. Il lavoro è completo quando viene implementata la risoluzione accettata per la regressione oppure viene aggiornato What's New se il comportamento viene mantenuto, insieme alla relativa copertura della regressione o alla documentazione.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Valutazione

Stack tecnologico
python
Ambito
backend
Tipo di issue
Bug
Difficoltà
4/5
Tempo stimato
3-5 giorni
Stato di attività
Ferma
Chiarezza
Abbastanza chiara
Idoneità per principianti
35/100

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.