gdobbins / ido-ubiquitous

Fancy completion all over Emacs, not just for buffers and files.

ido-ubiquitous

Gimme some ido... everywhere! This package replaces stock emacs completion with ido completion wherever it is possible to do so without breaking things.

Get it from MELPA: https://stable.melpa.org/#/ido-ubiquitous

Version 3.0 changes

Long-time users should know that ido-ubiquitous version 3.0 is a major update, including a split into two packages, and some of the configuration options have changed in non-backwards-compatible ways. If you have customized any options ido-ubiquitous, be sure to check out M-x customize-group ido-ubiquitous and M-x customize-group ido-completing-read+ after updating to 3.0 and make sure the new settings are to your liking.

How to enable ido in as many places as possible

If you are using this package, you probably want to enable ido everywhere that it is possible to do so. Here are all the places to enable ido that I'm aware of. (Note that most of these variables can also be set via M-x customize-variable if you prefer that.)

Ido itself

First, enable ido-mode and ido-everywhere.

(ido-mode 1)
(ido-everywhere 1)

ido-ubiquitous (this package)

Install this package from MELPA and then turn on ido-ubiquitous-mode:

(require 'ido-ubiquitous)
(ido-ubiquitous-mode 1)

Smex

Smex allows you to use ido for completion of commands in M-x, with enhancements like putting your most-used commands at the front of the list. First install the smex package, then follow the directions to set up key-bindings for it.

ido-yes-or-no

If you want to use ido for yes-or-no questions, even though it's massive overkill, install my ido-yes-or-no package from MELPA, and then enable the mode:

(require 'ido-yes-or-no)
(ido-yes-or-no-mode 1)

ido for describe-face and certain other commands

Some commands, such as describe-face, use completing-read-multiple instead of completing-read. You can get ido completion for these commands with crm-custom-mode, which replaces completing-read-multiple with repeated calls to completing-read, which would then use ido thanks to ido-ubiquitous-mode. First, install the crm-custom package from MELPA, then enable the mode:

(require 'crm-custom)
(crm-custom-mode 1)

Remember that when using this mode, completion for affected commands will continue to read additional items until you use C-j to enter an empty input, which terminates the completion. (Due to this quirk, I do not find this mode to be very useful in conjunction with ido, but it does work.)

Packages with built-in ido support

Lastly, some packages already provide their own interfaces to ido, so ido-ubiquitous specifically avoids interfering with these. If you use any of the following packages, you need to enable ido for each of them separately.

• Magit: (setq magit-completing-read-function 'magit-ido-completing-read)
• Gnus: (setq gnus-completing-read-function 'gnus-ido-completing-read)

(You can also use M-x customize-variable to set all of these options.)

Frequently asked questions

How does ido-ubiquitous decide when to replace completing-read?
Why don't some commands use ido completion?

Emacs' completing-read is a complex function with many advanced features and some quirks that are only maintained for backwards compatibility. Not all of these features are supported by ido, so it is impossible to always replace completing-read with ido completion. Trying to use ido when these features are requested can completely break the completion system. So, by default ido-ubiquitous tries to get out of the way whenever it detects that these features might be used by a given call to completing-read. Furthermore, it's not always possible to detect based on the arguments to completing-read whether such ido-incompatible features are being used or not. ido-ubiquitous uses a series of heuristics to determine whether unsupported features might be used, and also supports an override feature to correct it when the heuristics get things wrong.

Each time a function invokes completing-read, ido-ubiquitous selects one of 3 modes:

• enable: use ido completion;
• enable-old: use ido completion, but with a compatibility fix for old-style default selection; and
• disable: no ido completion.

The following describes in detail how ido-ubiquitous selects the appropriate mode, and what to do when you think it is making the wrong choice.

Disabling ido completion when the collection is very large

Ido can get slow on very large collections, so by default ido-ubiquitous disables itself for collections with more than 30,000 items in them. This rule takes precedence over anything else, even overrides that explicitly enable ido completion for a command.

You can change the large collection threshold by customizing ido-cr+-max-items.

Disabling ido completion when the collection is a function

One feature of completing-read is that the collection argument can be a function. This function could simply return a list of all possible completions, in which case it would be safe to use ido, or it could implement a completely new completion system, in which case using ido would interfere with this new completion system (for an example of this, see the tmm command). But ido-ubiquitous cannot tell by looking at the function which kind it is, so it errs on the side of caution and disables itself whenever the collection is a function, unless an override exists telling it that the command is safe for ido completion. You can turn off this safeguard by customizing ido-ubiquitous-allow-on-functional-collection. Be aware that enabling this will likely break completion entirely in any command that uses this feature to implement non-standard completion.

If you run across a command that unexpectedly uses normal Emacs completion instead of ido completion, it's likely that either this or the previous option is to blame.

Disabling or enabling ido completion by overrides

If ido-ubiquitous is not running for a command where is should be due to the functional collection rule, you can add an override for that command by using M-x customize-variable ido-ubiquitous-command-overrides. Conversely, if ido completion causes problems for a command, you can also use this to disable ido-ubiquitous for that command.

What is old-style default selection?

The enable-old mode of operation is required because the old way for completing-read to indicate that the user simply pressed RET and selected the default option was to return an empty string. When this old-style mode is used, completing-read doesn't even know what the default is supposed to be -- the calling code handles all of that. But in ido, simply pressing RET will return the first item of the list, not an empty string. The way to enter an empty string in ido is C-j. The enable-old mode enables ido completion, but swaps the meaning of C-j and RET if you haven't entered any text or cycled the options yet (once you do either of those, C-j and RET regain their standard meanings). This allows you to select the default by pressing RET as soon as the completion prompt appears, as intended (C-j would select the first item).

Unfortunately, there is no way for ido-ubiquitous to detect when a command is using this old-style default selection, so instead it uses a built-in set of overrides telling it about commands that are known to use old-style defaults. If you discover a command where pressing RET or C-j at an empty prompt is not doing what you expect it to, there's a good chance that you need to add an enable-old override for that command to ido-ubiquitous-command-overrides. Luckily, since this is an obsolete usage pattern, it is unlikely that any Elisp functions written since 1990 or so will need to be added to this list. Hopefully all uses of old-style completion will eventually be eliminated, and with them, the need for this feature of ido-ubiquitous.

How can I troubleshoot when ido-ubiquitous isn't doing what I want?

First, invoke the ido-ubiquitous-debug-mode and ido-cr+-debug-mode commands. Then, with these two modes active, run the command(s) that you are having trouble with, and when the completion prompt appears, make a selection to complete the process. Then, examine the Messages buffer, where ido-ubiquitous will explain which mode of operation it selected and why. Based on this, you can add an override to ido-ubiquitous-command-overrides. If you are not familiar with the structure of this variable, I recommend using M-x customize-variable to edit it, which will help you get it right. If ido completion was skipped ido completion because the collection was too large, try giving ico-cr+-max-items a larger value, or set it to nil if your computer is fast enough to handle any size of collection.

Updates to ido-ubiquitous often include new overrides, but Emacs will not edit your override variables if you have already customized them. So, if you have recently upgraded ido-ubiquitous, remember to invoke ido-ubiquitous-restore-default-overrides to add in any new overrides. (In the future, this process may become automatic: DarwinAwardWinner#90)

Where can I report bugs or suggest new overrides?

If you end up adding any overrides, please also report them at https://github.com/DarwinAwardWinner/ido-ubiquitous/issues so I can incorporate them into the defaults for future versions. You can also report any bugs you find in ido-ubiquitous.

I'm getting some weird warnings from ido-ubiquitous when Emacs starts.

I've gotten numerous reports about nonsensical warnings produced by ido-ubiquitous, such as "free variable" warnings about variables that are most definitely properly declared, or warnings that only appear when ido-ubiquitous is loaded after another unrelated package. For many of these warnings, I've never been able to discover the cause or reproduce the warnings myself, and I've given up trying to figure it out. Please don't report any new bugs about variable warnings unless you can tell me how to consistently reproduce them starting from emacs -Q. If you are an Emacs expert who knows how to fix these warnings, please let me know.

You can see the bug reports about weird warnings here.

What is the "bleeding-edge" branch?

All users should just use the master branch, or better yet, install from MELPA. The bleeding-edge branch is where I test experimental and unfinished features. Because ido-ubiquitous hooks deeply into the bowels of Emacs, a bug in ido-ubiquitous could easily freeze or crash Emacs entirely. Additionally, some bug only show up when ido-ubiquitous is installed and compiled as a package. So I test every new feature myself for some time on this branch before pushing to the master branch. If you report a bug, I might develop a fix for it on the bleeding edge branch and ask then you to try this branch. Otherwise, normal users don't need to think about this branch.

ido-completing-read+

As of version 3.0, most of the core functionality of ido-ubiquitous has been spun off into a separate library called ido-completing-read+, or "ido-cr+" for short. ido-cr+ incorporates all the features of ido-ubiquitous that are actually just generic improvements to ido that should probably always be enabled. It implements these fixes in a single function ido-completing-read+, which should be suitable as a drop-in replacement for either ido-completing-read or completing-read. Notably, unlike the original ido-completing-read, ido-completing-read+ should always do something useful, by falling back to normal completion when ido completion won't work. Additionally, it allows you to manually fall back using C-f and C-b, in the same way you can use those keys to switch between file and buffer completion in ido. As a user, you don't really need to know anything about ido-cr+. However, if you are writing an Emacs package and would like to incorporate ido completion, you may wish to use ido-cr+ to get more robust completion with fewer weird edge cases.