`bytes.index()`, `bytes.rindex()`, `bytearray.index()` and `bytearray.rindex()` get a different error message

[hyperkai]

Bug report

Bug description:

str.index() and str.rindex() get the error message as shown below:

v = str('Hello')

print(v.index('abc'))
print(v.rindex('abc'))
# ValueError: substring not found

But bytes.index() and bytes.rindex() get the different error message as shown below:

v = bytes(b'Hello')

print(v.index(b'abc'))
print(v.rindex(b'abc'))
# ValueError: subsection not found

And, bytearray.index() and bytearray.rindex() also get the different error message as shown below:

v = bytearray(b'Hello')

print(v.index(b'abc'))
print(v.rindex(b'abc'))
# ValueError: subsection not found

So, the error message of them should be the same for consistency as shown below:

ValueError: substring not found

Or:

ValueError: subsection not found

CPython versions tested on:

3.12

Operating systems tested on:

Windows

[ryan-duve]

Are you saying it would be an improvement for bytes(b'Hello').index(b'abc') to return the message:

# ValueError: substring not found

to be consistent with the analogous operation on strings? If so, would it be confusing for the error message to say a "substring" is not found when it is searching for bytes, not a string? I ask because it seems to me we want exactly this difference in wording to contrast searching for bytes and searching for a string.

[avengerofpi]

I think the messages are communicating what they’re supposed to. “Substring” would be incorrect for bytes, but “subsection” is reasonable for bytes (the method docstrings actually use the term “subsequence”). “Subsection” could be confusing for strings where the term “substring” is common and reasonable. So while I’m all for standardizing error messages, i think these particular error messages are already standardized to an appropriate degree.

[sobolevn]

I agree, they look good enough. Let's keep this as-is. Since any change can break something for someone (like doctests), we need a very good reason to do it.