unit-test to catch a TimSort issue

[duzun]

Description

This is a failing test to prove TimSort is badly implemented.

Motivation and Context

Current implementation of TimSort works for arrays up to 32 items only.

How Has This Been Tested?

I've just generated a list of 42 numbers, from 1 to 42, shuffled it and sorted it with TimSort. The result should equal the initial sorted array.

[doganoo]

Hey,

thank you for this PR.

I will check this.

Do you also have a suggestion for the TimSort implementation?

EDIT

could you also please add your name and email to the sources you have changed?

[doganoo]

again, thank you for this PR.

Unfortunately I will not merge this branch into master since we do not want to prove broken code. You are very welcome to contribute to the TimSort class in order to fix whatever you think is broken (including the corresponding Unit Test).

[duzun]

You are welcome!

I am glad to contribute when I have more free time.

Until then I don't think it is the right thing to do to close this PR and leave the TimSort implementation broken without any notice.

[doganoo]

Hey there,

it has been a while :)

I think it is not a good idea to prove broken code since it could let people assume that this behaviour is wanted.

Instead, I added a var_dump to the class' constructor. It is not a good solution, I know. But it clearly says that it is broken instead of allowing people to assume.

PR's are highly welcome. Please refer to #23 when creating a PR. Thank you in advance :)