python / python/cpython

Narrow source positions for exceptions during pattern matching

Aperta
#96,999 4 commenti 0 reazioni 2 assegnatari Vedi su GitHub

@iritkatriel ci sta già lavorando.

Dal 30/9/2022.

interpreter-core type-feature
Lingua principale
Python
Stelle
77.2k
Fork
36k
Metriche di merge delle PR
Metriche PR in attesa

Descrizione

The source range in the traceback is to wide in some cases, if an exception is thrown during matching of a pattern.

The Ranges

example for attribute lookup:

class A:
    def __getattr__(self, name):
        assert name[0] == "a"
        return 5


match A():
    case A(apple=5, banana=7):
        print("hi")

output (Python 3.11.0rc2+):

Traceback (most recent call last):
  File "/home/frank/projects/cpython/match_test.py", line 8, in <module>
    case A(apple=5, banana=7):
         ^^^^^^^^^^^^^^^^^^^^
  File "/home/frank/projects/cpython/match_test.py", line 3, in __getattr__
    assert name[0] == "a"
AssertionError

Annotating only the source positions where the attribute check failed (banana=7) would make the traceback easier to read.

The same problem exists for positional matching:

class A:
    __match_args__ = ("apple", "banana")

    def __getattr__(self, name):
        assert name == "apple"


match A():
    case A(5, 7):
        print("hi")

output (Python 3.11.0rc2+):

Traceback (most recent call last):
  File "/home/frank/projects/cpython/match_test2.py", line 9, in <module>
    case A(5, 7):
         ^^^^^^^
  File "/home/frank/projects/cpython/match_test2.py", line 5, in __getattr__
    assert name == "apple"
AssertionError

Equality checks are handled better:

class NeverEqual:
    def __eq__(self, other):
        assert other==5


class A:
    def __init__(self):
        self.a = NeverEqual()
        self.b = 3


match A():
    case A(a=5|3, b=3):
        print("hi")

output (Python 3.11.0rc2+):

Traceback (most recent call last):
  File "/home/frank/projects/cpython/match_test3.py", line 13, in <module>
    case A(a=5|3, b=3):
               ^
  File "/home/frank/projects/cpython/match_test3.py", line 3, in __eq__
    assert other==5
AssertionError

Instance checks are represented in the same way:

class MetaB(type):
    def __instancecheck__(self, instance):
        assert False, "do not use me"


class B(metaclass=MetaB):
    pass


class A:
    def __init__(self):
        self.a = 5
        self.b = 3


match A():
    case A(a=5, b=2|B()):
        print("hi")

output (Python 3.11.0rc2+):

Traceback (most recent call last):
  File "/home/frank/projects/cpython/match_test4.py", line 17, in <module>
    case A(a=5, b=2|B()):
                    ^^^
  File "/home/frank/projects/cpython/match_test4.py", line 3, in __instancecheck__
    assert False, "do not use me"
AssertionError: do not use me

Problem for the User

Consistent source ranges would improve the readability of the tracebacks.

Is it possible to highlight always only the failing Pattern (with attribute name if given) for failing attribute access?

example:

case A(a=5):
       ^^^
case A(a=B()):
       ^^^^^
case A(5):
       ^
case A(5|B()):
       ^^^^^

Instance checks and equality tests are already handled well:

case A(a=5):
         ^
case A(a=B()):
         ^^^
case A(5):
       ^
case A(B()):
       ^^^
case A(5|B())
         ^^^

Problem for libraries

Libraries are using the source ranges for all sorts of functionality.
Executing for example is using this ranges (for the new 3.11 support) to map from instructions back to ast-nodes. Useful ranges are here required to map to useful ast-nodes.

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.

Valutazione

Questa issue non è ancora stata valutata.

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.