__getitem__ error message hard to understand

[cvrebert]

BPO	5760
Nosy	@birkenfeld, @terryjreedy

^{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 = None
closed_at = <Date 2009-04-18.08:31:49.499>
created_at = <Date 2009-04-15.06:32:10.854>
labels = ['interpreter-core', 'type-bug']
title = '__getitem__ error message hard to understand'
updated_at = <Date 2009-04-18.08:31:49.188>
user = 'https://bugs.python.org/cvrebert'

bugs.python.org fields:

activity = <Date 2009-04-18.08:31:49.188>
actor = 'georg.brandl'
assignee = 'none'
closed = True
closed_date = <Date 2009-04-18.08:31:49.499>
closer = 'georg.brandl'
components = ['Interpreter Core']
creation = <Date 2009-04-15.06:32:10.854>
creator = 'cvrebert'
dependencies = []
files = []
hgrepos = []
issue_num = 5760
keywords = []
message_count = 3.0
messages = ['85983', '86105', '86110']
nosy_count = 3.0
nosy_names = ['georg.brandl', 'terry.reedy', 'cvrebert']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue5760'
versions = ['Python 3.0', 'Python 3.1', 'Python 2.7']

[cvrebert]

Prompted by
http://mail.python.org/pipermail/python-ideas/2009-April/004048.html

The current error message issued when trying to use the get item ([])
operator on an object that does not define __getitem__ can be hard to
understand:

>>> class A(object): pass
...
>>> A()['a']
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: 'A' object is unsubscriptable

Problems observed:

• "unsubscriptable" is easily misread in haste as "unscriptable", which
  can be quite confusing, especially to newbies
• "subscripting" is not frequently used to describe the [] operator,
  making the message more difficult to decipher (again, esp. for newbies)
• the underlying lack of a __getitem__ method is not mentioned, thus not
  making it obvious how to remedy the error

Suggestion:
Use exception chaining and rephrase the error message to get something like:

AttributeError: class 'A' has no attribute '__getitem__'
The above exception was the direct cause of the following exception:
TypeError: 'A' object does not support the 'get item' operator

Similar changes should be made to __setitem__ & __delitem__.

[terryjreedy]

As I said on the python-ideas discussion, which definitely did *not*
come to consensus, I disagree with this suggestion. To repeat and
expand on what I said there:

1. 'unsubscriptable' could instead be changed to 'not subscriptable'.

2. 'subscription' *is* the way Python describes the use of [].
   """A subscription selects an item of a sequence (string, tuple or list)
   or mapping (dictionary) object:

subscription ::= primary "[" expression_list "]"
"""
'array subscript' (from C, for example) is a standard term used for
decades in computer languages. It is appropriate since a[i] is one
standard single-text-line replacement for a<typographic subscript>i when
one cannot actually 'type' a subscript. It *is* a way of representing a
subscript, just as i**j is an in-line way of representing a superscript
exponent, which we still call an exponent in spite of it not being
'raised'. Both constitute visible markup that is part of the compromise
in representing 2D typography with a 1D stream of characters. Using '/'
for division (and fractions) is another.

3. Python error message generally do not and I think should not mention
   the special method implementation that underlies surface level errors.
   In this particular case, the remedy to mistakenly trying to subscript
   something is to not do that. I think mentioning __getitem__ would be a
   distraction. One generally *cannot* just go around adding it.

[birkenfeld]

I concur. I changed "unsubscriptable" to "not subscriptable" in r71696.