Some issues with the new point completion

[ghost]

I upgraded my swiper install and found the new ivy point completion features. I really like the idea, but am running into a couple bugs with this feature, and have disabled it for myself at the moment:

Too aggressive replacement:

I spawn bash shells using shell. When using cd and hitting tab to get completion, ivy will complete the term as well as replace the starting input. This is good for replacing uncompleted content, but not always the whole term used for completion.

Example:

$ cd pat<tab>
$ cd path/to/dir<tab>
$ cd directory/

Expected behavior would be to replace pat in the first, then dir in the second. path/to/dir is replaced in the second, resulting in an invalid directory.

---

Completion doesn't work from within the minibuffer. I receive a recursive minibuffer error, since ivy handles it in the minibuffer as opposed to the *Completions* buffer.

Example: M-x customize-face def TAB

---

This is minor, but is there a reason for binding ivy-count-format in 569c11e with the compact value? I usually keep mine disabled, and it is re-enabled for the point completion.

[abo-abo]

Thanks, please test. All your concerns have been addressed. Please try to file separate issues for independent concerns.

[ghost]

Thanks for the fixes, I've just now got the opportunity to re-enable and test in my Emacs.

The first two cases work great now, but the ivy-count-format should check empty string rather than nil. I've opened #349 for the special cases needed with empty ivy-count-format in ivy-completion-in-region.

Please try to file separate issues for independent concerns.

Will do in the future.