Broken the menu display (emacs 24.5 only)

The menu display breaks when 'auto-complete' is executed in Emacs 24.5.1. I called ac-linum-workaround, but the effect wasn't obtained.

Emacs 24.5.1でauto-completeを行うとメニュー表示が壊れます メニューが表示される場所に文字列が存在するとおかしくなるようです。 24.4までは正常表示されていました。 最初linum-modeが有効だからなのかと思い、(ac-linum-workaround)を使用しましたが効果がありませんでした。

environment

OS:Windows7/Ubuntu14.10 Emacs: 24.4.1/24.5.1

I tested both version.

両方をテストしました。

24.4 ( proper )

24.5 ( broke )

simple test case

I tested on simple environment. 24.4 & 24.5 launch by $ emacs -Q. You open popup.el auto-complete.el, and execute eval-buffer. Switch buffer to scratch, and evaluate following code.

よりシンプルな状況でテストしました。 24.4/24.5をそれぞれ $ emacs -Q で起動します。 popup.el auto-complete.el を開き、それぞれ eval-bufferを行います。 scratch へ移動して以下を評価します。

(let ()
  (ac-linum-workaround)
  (linum-mode)
  (auto-complete-mode))
font-
0
01
012
0123

Evaluate let, you execute auto-complete at after font-. 24.5 only broke. popup.el have same symptom.

letを評価して、font-でauto-completeを実行します。 すると24.5のみ表示が壊れます。 popup.elだけでも同様の状況です。

(let ((hoge "01234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789"))
  (linum-mode)
  (popup-tip hoge))

before eval

afeter eval(1st-figure 24.4, 2nd-figure 24.5)

以上報告でした。

After entering a string like "foo.ATTR" in python-mode (builtin), no completions will be shown if ATTR is a python keyword. E.g. foo.in on a list does not complete to foo.index. The completion is only shown again once the attribute is not a keyword anymore.

Hey m2ym, what would you think about moving auto-complete/popup/fuzzy to an auto-complete organization, and giving a few others the ability to push/merge pull requests/close tickets? Turn around on issues and pull requests has historically been a bit slow; I'm sure you are really busy, so I think you should let us help you out.

I know the ac and popup codebases pretty well so I am volunteering to help out a couple of days of the week. One thing I would like to do with the organization is group together all of the various sources (perhaps by mirroring, in some cases) and kill off auto-complete-config.el, which is typically not in sync with many sources of interest (eclim, yasnippet, ropemacs).

I'm sure others would be willing to help out as well. I'd like to hear your thoughts on the matter.

My environment:

• Windows 7, 64 bit
• cygwin 64bit

After I enabled ac, and type /, it gives a list of folders in my system, starting from root. But if I type two / continuously, Emacs got frozen. After a long, it will give a list of folders, like / did.

Aside from the completions themselves, the popup is useful for the accompanying documentation.

Could we have a command that spawned again the popup in any already complete word?

Thank you and keep up the good work.

If (point) is part-way through a word which is also a candidate, the old expansion behaviour is unexpected and undesirable.

Given

Foo|bar

where '|' indicates the point and the ac-candidates are "Foobar" and "Foobarometer", pressing TAB to expand would result in

Foobar|bar

With this commit, the result is now more expected:

Foobar|

Additionally, given

Foobar|om

then if we expanded the candidate "Foobarometer", the result would previously have been:

Foobarometer|om

With this commit, the result is now:

Foobarometer|

I want to discuss here about the architecture of AC 2.0. The main advance of 2.0 is to abolish the property-list style source definition. Instead, I want to adopt the defun style source definition like compay-mode.

Plus, I want to make source definitions more flexible and extensible with good maintainability. The problem is that I don't have much experience about source development. So it is hard for me to decide how flexible an extensible it should be.

Any suggestions are welcome.

The current policy is too disruptive. Any source could break all other sources because it wants to implement its own prefix parser. But there is no point in excluding, say, words in current or all buffers, yasnippet, etc. Why not letting each source parse with its own prefix?

PS: here is an example of the problems caused by the current source selection logic.

I'm tying to use auto-complete (and popup.el) with org-mode. I've the following problem. If a pop-up overlay a folded entry the menu is broken. I obtain some ... instead of the suggested word and sometime the alignment of the entries are wrong. The problem is already described in this post http://lists.gnu.org/archive/html/emacs-orgmode/2011-02/msg00305.html

When at eob, ac menu is distorted. The cursor and inline overlay are moved beyond the ac-menu. Consequently ac-expand (tab) doesn't work when it is invoked repeatedly. This happens because the point is moved out of the ac-prefix-overlay regio. This is extremely harmful in the shell and comint buffers when editing is always at eob.

There is already a workaround in auto-complete.el for the case when ac-point == (point-min) == (point-max). This workaround should be also used when point is at eob.

when add belowing setting:

(setq-default ac-sources '(ac-source-words-in-same-mode-buffers
                           ac-source-abbrev
                           ac-source-dictionary
                           ac-source-yasnippet 
                           ))

shadow tips feature is not worked !

if I comment ac-source-yasnippet, It rework for me.

But in ac pop menu, can not show the yasnippet candidate

How I can do setting yasnippet-candidate with ac ?

In the previous PR (#476) the author deleted the source repository and left us with raw patches only. I made a branch and applied the changeset from that PR. I'm not sure if it should be merged as-is. One thing that seems strange is the inclusion of parens (and commas) in some candidates, like matrix(,,,,,). From the diff it looks like we didn't have candidates like this before. Also, I imagine completing such a function would put the point after the closing paren - you'd then have to jump back to the beginning of the arguments list, which could be more distracting than completing to matrix| and pressing ( from there. Things like matrix(,,,,,) are probably better suited for yasnippet than AC. In any case - we now have a PR that is merge-able, whether we do merge it or not :)

Hi, Looks like auto-complete ignores completion-styles when building the candidate list... Problem shows itself on non substring based completion styles, eg. partial-completion.

To reproduce: (Emacs 27.1, auto-complete #aafd3f5)

; <emacs --no-init-file>
; <set up path to auto-complete>
(require 'auto-complete)
(require 'auto-complete-config)
(ac-config-default)
(define-key global-map (kbd "TAB") 'ac-trigger-key-command)
(setq completion-styles '(basic partial-completion))
; switch to new buffer
; activate emacs-lisp-mode
; type "(ad-to-lis"<TAB>; observe completion failure
; call M-x completion-at-point or company completion; observe input expansion to "add-to-list"

From a quick look at ac-candidates, it seems they are assuming that the initial input is the candidates' common substring?

Thanks

auto-complete, popup-el and fuzzy-el are looking for new maintainers. If you are interested, then please say so.

Currently I am the "owner" of these repositories but I have only volunteered to do that so that these packages can be handed over smoothly once we have found some volunteers. I am not qualified myself because I have never even just used these packages.

I used to use it a year ago but seeing people switching to company-mode makes me feel uncomfortable. What is it that auto-complete isn't providing because of which people are switching to company-mode?

I have a database of symbols used in a language. I want to have an auto-completion based on those words as well. Is there any way to achieve this using auto-complete?

.hmm,i use ycmd with ac, i want to use tab to complete word in popmenu,but now tab is use to select next one in popmenu then i need to to complete select one. i want to use tab to complete at the same time. what i need do? THANKS