Error message says dict has no len

[vedgar]

BPO	33933
Nosy	@rhettinger, @serhiy-storchaka, @vedgar
Superseder	bpo-32500: PySequence_Length() raises TypeError on dict type

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/serhiy-storchaka'
closed_at = <Date 2018-06-24.07:09:15.001>
created_at = <Date 2018-06-21.13:25:21.228>
labels = ['interpreter-core', '3.7']
title = 'Error message says dict has no len'
updated_at = <Date 2018-06-24.07:09:15.000>
user = 'https://github.com/vedgar'

bugs.python.org fields:

activity = <Date 2018-06-24.07:09:15.000>
actor = 'serhiy.storchaka'
assignee = 'serhiy.storchaka'
closed = True
closed_date = <Date 2018-06-24.07:09:15.001>
closer = 'serhiy.storchaka'
components = ['Interpreter Core']
creation = <Date 2018-06-21.13:25:21.228>
creator = 'veky'
dependencies = []
files = []
hgrepos = []
issue_num = 33933
keywords = []
message_count = 3.0
messages = ['320178', '320330', '320362']
nosy_count = 3.0
nosy_names = ['rhettinger', 'serhiy.storchaka', 'veky']
pr_nums = []
priority = 'low'
resolution = 'duplicate'
stage = 'resolved'
status = 'closed'
superseder = '32500'
type = None
url = 'https://bugs.python.org/issue33933'
versions = ['Python 3.7']

[vedgar]

Look at this:

>>> import bisect
>>> bisect.bisect({}, None)
Traceback (most recent call last):
  File "<pyshell#1>", line 1, in <module>
    bisect.bisect({}, None)
TypeError: object of type 'dict' has no len()

Of course, objects of type 'dict' do have len. The problem is that bisect considers its first argument a sequence, which is very sensible to do (although, with ordered dicts, it's not the only sensible choice), but it gives a very wrong error message given that context.

At https://bugs.python.org/issue32500, R. David Murray told me to open this.

[rhettinger]

This probably doesn't warrant another tracker entry. Serhiy already had a PR for this and attached it to the original bug report for https://bugs.python.org/issue32500 . See #7846

My suggestion would be to just change the error message in PySequence_Size() from "object of type '%.200s' has no len()" to "object of type '%.200s' has no len() or does not support the sequence protocol (consider using a list instead)".

Marking this as low priority because the behavior has existed for a very long time and hasn't bothered anyone until now (possibly because the bisect module is documented to only search lists). Also, this is a specific instance of a more general, pervasive, and hard problem where low level routines raise exceptions that have error messages than may not make much sense from the point of view someone calling higher level APIs.

Suggest closing this as a duplicate and continuing the discussion back in bpo-32500 where it arose. This would show-up in any C code that calls PySequence_Size() not just bisect.

[serhiy-storchaka]

I think this is an exact duplicate of bpo-32500.