1481 lines
61 KiB
Plaintext
1481 lines
61 KiB
Plaintext
*command-t.txt* Command-T plug-in for Vim *command-t*
|
|
|
|
CONTENTS *command-t-contents*
|
|
|
|
1. Introduction |command-t-intro|
|
|
2. Requirements |command-t-requirements|
|
|
3. Installation |command-t-installation|
|
|
4. Trouble-shooting |command-t-trouble-shooting|
|
|
5. Usage |command-t-usage|
|
|
6. Commands |command-t-commands|
|
|
7. Mappings |command-t-mappings|
|
|
8. Options |command-t-options|
|
|
9. FAQ |command-t-faq|
|
|
10. Tips |command-t-tips|
|
|
11. Authors |command-t-authors|
|
|
12. Development |command-t-development|
|
|
13. Website |command-t-website|
|
|
14. License |command-t-license|
|
|
15. History |command-t-history|
|
|
|
|
|
|
INTRODUCTION *command-t-intro*
|
|
|
|
The Command-T plug-in provides an extremely fast, intuitive mechanism for
|
|
opening files and buffers with a minimal number of keystrokes. It's named
|
|
"Command-T" because it is inspired by the "Go to File" window bound to
|
|
Command-T in TextMate.
|
|
|
|
Files are selected by typing characters that appear in their paths, and are
|
|
ordered by an algorithm which knows that characters that appear in certain
|
|
locations (for example, immediately after a path separator) should be given
|
|
more weight.
|
|
|
|
To search efficiently, especially in large projects, you should adopt a
|
|
"path-centric" rather than a "filename-centric" mentality. That is you should
|
|
think more about where the desired file is found rather than what it is
|
|
called. This means narrowing your search down by including some characters
|
|
from the upper path components rather than just entering characters from the
|
|
filename itself.
|
|
|
|
Screencasts demonstrating the plug-in can be viewed at:
|
|
|
|
https://wincent.com/products/command-t
|
|
|
|
|
|
REQUIREMENTS *command-t-requirements*
|
|
|
|
The plug-in requires Vim compiled with Ruby support, a compatible Ruby
|
|
installation at the operating system level, and a C compiler to build
|
|
the Ruby extension.
|
|
|
|
|
|
1. Vim compiled with Ruby support ~
|
|
|
|
You can check for Ruby support by launching Vim with the --version switch:
|
|
|
|
vim --version
|
|
|
|
If "+ruby" appears in the version information then your version of Vim has
|
|
Ruby support.
|
|
|
|
Another way to check is to simply try using the :ruby command from within Vim
|
|
itself:
|
|
|
|
:ruby 1
|
|
|
|
If your Vim lacks support you'll see an error message like this:
|
|
|
|
E319: Sorry, the command is not available in this version
|
|
|
|
The version of Vim distributed with OS X may not include Ruby support (for
|
|
example, Snow Leopard, which was the current version of OS X when Command-T
|
|
was first released, did not support Ruby in the system Vim, but the current
|
|
version of OS X at the time of writing, Mavericks, does). All recent versions
|
|
of MacVim come with Ruby support; it is available from:
|
|
|
|
http://github.com/b4winckler/macvim/downloads
|
|
|
|
For Windows users, the Vim 7.2 executable available from www.vim.org does
|
|
include Ruby support, and is recommended over version 7.3 (which links against
|
|
Ruby 1.9, but apparently has some bugs that need to be resolved).
|
|
|
|
|
|
2. Ruby ~
|
|
|
|
In addition to having Ruby support in Vim, your system itself must have a
|
|
compatible Ruby install. "Compatible" means the same version as Vim itself
|
|
links against. If you use a different version then Command-T is unlikely
|
|
to work (see |command-t-trouble-shooting| below).
|
|
|
|
On OS X Snow Leopard, Lion and Mountain Lion, the system comes with Ruby 1.8.7
|
|
and all recent versions of MacVim (the 7.2 snapshots and 7.3) are linked
|
|
against it.
|
|
|
|
On OS X Mavericks, the default system Ruby is 2.0, but MacVim continues to
|
|
link against 1.8.7, as does the Apple-provided Vim. Ruby 1.8.7 is present on
|
|
the system at:
|
|
|
|
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/bin/ruby
|
|
|
|
On OS X Yosemite, the default system Ruby is 2.0, and the Vim that comes with
|
|
the system links against it.
|
|
|
|
On Linux and similar platforms, the linked version of Ruby will depend on
|
|
your distribution. You can usually find this out by examining the
|
|
compilation and linking flags displayed by the |:version| command in Vim, and
|
|
by looking at the output of:
|
|
|
|
:ruby puts "#{RUBY_VERSION}-p#{RUBY_PATCHLEVEL}"
|
|
|
|
Or, for very old versions of Ruby which don't define `RUBY_PATCHLEVEL`:
|
|
|
|
:ruby puts RUBY_VERSION
|
|
|
|
A suitable Ruby environment for Windows can be installed using the Ruby
|
|
1.8.7-p299 RubyInstaller available at:
|
|
|
|
http://rubyinstaller.org/downloads/archives
|
|
|
|
If using RubyInstaller be sure to download the installer executable, not the
|
|
7-zip archive. When installing mark the checkbox "Add Ruby executables to your
|
|
PATH" so that Vim can find them.
|
|
|
|
|
|
3. C compiler ~
|
|
|
|
Part of Command-T is implemented in C as a Ruby extension for speed, allowing
|
|
it to work responsively even on directory hierarchies containing enormous
|
|
numbers of files. As such, a C compiler is required in order to build the
|
|
extension and complete the installation.
|
|
|
|
On OS X, this can be obtained by installing the Xcode Tools from the App
|
|
Store.
|
|
|
|
On Windows, the RubyInstaller Development Kit can be used to conveniently
|
|
install the necessary tool chain:
|
|
|
|
http://rubyinstaller.org/downloads/archives
|
|
|
|
At the time of writing, the appropriate development kit for use with Ruby
|
|
1.8.7 is DevKit-3.4.5r3-20091110.
|
|
|
|
To use the Development Kit extract the archive contents to your C:\Ruby
|
|
folder.
|
|
|
|
|
|
INSTALLATION *command-t-installation*
|
|
|
|
You install Command-T by obtaining the source files and building the C
|
|
extension.
|
|
|
|
The recommended way to get the source is by using a plug-in management system.
|
|
There are several such systems available, and my preferred one is Pathogen
|
|
(https://github.com/tpope/vim-pathogen) due to its simplicity and robustness.
|
|
|
|
Other plug-in managers include:
|
|
|
|
- Vundle: https://github.com/gmarik/Vundle.vim (see |command-t-vundle|)
|
|
- NeoBundle: https://github.com/Shougo/neobundle.vim (see
|
|
|command-t-neobundle|)
|
|
- VAM: https://github.com/MarcWeber/vim-addon-manager (see |command-t-vam|)
|
|
|
|
The following sections outline how to use each of these managers to download
|
|
Command-T, and finally |command-t-compile| describes how to compile it.
|
|
|
|
*command-t-pathogen*
|
|
Obtaining the source using Pathogen ~
|
|
|
|
Pathogen is a plugin that allows you to maintain plugin installations in
|
|
separate, isolated subdirectories under the "bundle" directory in your
|
|
|'runtimepath'|. The following examples assume that you already have
|
|
Pathogen installed and configured, and that you are installing into
|
|
`~/.vim/bundle`.
|
|
|
|
If you manage your entire `~/.vim` folder using Git then you can add the
|
|
Command-T repository as a submodule:
|
|
|
|
cd ~/.vim
|
|
git submodule add git://git.wincent.com/command-t.git bundle/command-t
|
|
git submodule init
|
|
|
|
Or if you just wish to do a simple clone instead of using submodules:
|
|
|
|
cd ~/.vim
|
|
git clone git://git.wincent.com/command-t.git bundle/command-t
|
|
|
|
Once you have a local copy of the repository you can update it at any time
|
|
with:
|
|
|
|
cd ~/.vim/bundle/command-t
|
|
git pull
|
|
|
|
Or you can switch to a specific release with:
|
|
|
|
cd ~/.vim/bundle/command-t
|
|
git checkout 1.10
|
|
|
|
To generate the help tags under Pathogen it is necessary to do so explicitly
|
|
from inside Vim:
|
|
|
|
:call pathogen#helptags()
|
|
|
|
For more information about Pathogen, see:
|
|
|
|
https://github.com/tpope/vim-pathogen
|
|
|
|
*command-t-vundle*
|
|
Obtaining the source using Vundle ~
|
|
|
|
Anywhere between the calls to `vundle#begin` and `vundle#end` in your
|
|
`~/.vimrc`, add a `Plugin` directive telling Vundle of your desire to use
|
|
Command-T:
|
|
|
|
call vundle#begin()
|
|
Plugin 'wincent/command-t'
|
|
call vundle#end()
|
|
|
|
To actually install the plug-in run `:PluginInstall` from inside Vim. After
|
|
this, you can proceed to compile Command-T (see |command-t-compile|).
|
|
|
|
For more information about Vundle, see:
|
|
|
|
https://github.com/gmarik/Vundle.vim
|
|
|
|
*command-t-neobundle*
|
|
Obtaining the source using NeoBundle ~
|
|
|
|
Anywhere between the calls to `neobundle#begin` and `neobundle#end` in your
|
|
`~/.vimrc`, add a `NeoBundle` directive telling NeoBundle of your desire to use
|
|
Command-T:
|
|
|
|
call neobundle#begin(expand('~/.vim/bundle/'))
|
|
NeoBundle 'wincent/command-t'
|
|
call neobundle#end()
|
|
|
|
To actually install the plug-in run `:NeoBundleInstall` from inside Vim. After
|
|
this, you can proceed to compile Command-T (see |command-t-compile|).
|
|
|
|
For more information about NeoBundle, see:
|
|
|
|
https://github.com/Shougo/neobundle.vim
|
|
*command-t-vam*
|
|
Obtaining the source using VAM ~
|
|
|
|
After the call to `vam#ActivateAddons` in your `~/.vimrc`, add Command-T to
|
|
the `VAMActivate` call:
|
|
|
|
call vam#ActivateAddons([])
|
|
VAMActivate github:wincent/command-t
|
|
|
|
After VAM has downloaded Command-T, you can proceed to compile it (see
|
|
|command-t-compile|).
|
|
|
|
For more information about VAM, see:
|
|
|
|
https://github.com/MarcWeber/vim-addon-manager
|
|
|
|
*command-t-compile*
|
|
Compiling Command-T ~
|
|
|
|
The C extension must be built, which can be done from the shell. If you use a
|
|
typical Pathogen, Vundle or NeoBundle set-up then the files were installed inside
|
|
`~/.vim/bundle/command-t`. A typical VAM installation path might be
|
|
`~/.vim/vim-addons/command-t`.
|
|
|
|
Wherever the Command-T files were installed, you can build the extension by
|
|
changing to the `ruby/command-t` subdirectory and running a couple of commands
|
|
as follows:
|
|
|
|
cd ~/.vim/bundle/command-t/ruby/command-t
|
|
ruby extconf.rb
|
|
make
|
|
|
|
Note: If you are an RVM or rbenv user, you must perform the build using the
|
|
same version of Ruby that Vim itself is linked against. This will often be the
|
|
system Ruby, which can be selected before issuing the "make" command with one
|
|
of the following commands:
|
|
|
|
rvm use system
|
|
rbenv local system
|
|
|
|
Note: If you are on OS X Mavericks and compiling against MacVim, the default
|
|
system Ruby is 2.0 but MacVim still links against the older 1.8.7 Ruby that is
|
|
also bundled with the system; in this case the build command becomes:
|
|
|
|
cd ~/.vim/bundle/command-t/ruby/command-t
|
|
/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/bin/ruby extconf.rb
|
|
make
|
|
|
|
Note: Make sure you compile targeting the same architecture Vim was built for.
|
|
For instance, MacVim binaries are built for i386, but sometimes GCC compiles
|
|
for x86_64. First you have to check the platform Vim was built for:
|
|
|
|
vim --version
|
|
...
|
|
Compilation: gcc ... -arch i386 ...
|
|
...
|
|
|
|
and make sure you use the correct ARCHFLAGS during compilation:
|
|
|
|
export ARCHFLAGS="-arch i386"
|
|
make
|
|
|
|
Note: If you are on Fedora 17+, you can install Command-T from the system
|
|
repository with:
|
|
|
|
su -c 'yum install vim-command-t'
|
|
|
|
*command-t-appstream*
|
|
AppStream Metadata ~
|
|
|
|
When preparing a Command-T package for distribution on Linux using Gnome
|
|
Software or another AppStream compatible application, there is a metafile in
|
|
appstream directory.
|
|
|
|
You can find more about AppStream specification at:
|
|
|
|
http://www.freedesktop.org/software/appstream/docs/
|
|
|
|
|
|
TROUBLE-SHOOTING *command-t-trouble-shooting*
|
|
|
|
Most installation problems are caused by a mismatch between the version of
|
|
Ruby on the host operating system, and the version of Ruby that Vim itself
|
|
linked against at compile time. For example, if one is 32-bit and the other is
|
|
64-bit, or one is from the Ruby 1.9 series and the other is from the 1.8
|
|
series, then the plug-in is not likely to work.
|
|
|
|
On OS X, Apple tends to change the version of Ruby that comes with the system
|
|
with each major release. See |command-t-requirements| above for details about
|
|
specific versions. If you wish to use custom builds of Ruby or of MacVim then
|
|
you will have to take extra care to ensure that the exact same Ruby
|
|
environment is in effect when building Ruby, Vim and the Command-T extension.
|
|
|
|
For Windows, the following combination is known to work:
|
|
|
|
- Vim 7.2 from http://www.vim.org/download.php:
|
|
ftp://ftp.vim.org/pub/vim/pc/gvim72.exe
|
|
- Ruby 1.8.7-p299 from http://rubyinstaller.org/downloads/archives:
|
|
http://rubyforge.org/frs/download.php/71492/rubyinstaller-1.8.7-p299.exe
|
|
- DevKit 3.4.5r3-20091110 from http://rubyinstaller.org/downloads/archives:
|
|
http://rubyforge.org/frs/download.php/66888/devkit-3.4.5r3-20091110.7z
|
|
|
|
If a problem occurs the first thing you should do is inspect the output of:
|
|
|
|
ruby extconf.rb
|
|
make
|
|
|
|
During the installation, and:
|
|
|
|
vim --version
|
|
|
|
And compare the compilation and linker flags that were passed to the
|
|
extension and to Vim itself when they were built. If the Ruby-related
|
|
flags or architecture flags are different then it is likely that something
|
|
has changed in your Ruby environment and the extension may not work until
|
|
you eliminate the discrepancy.
|
|
|
|
From inside Vim, you can confirm the version of Ruby that it is using by
|
|
issuing this command:
|
|
|
|
:ruby puts "#{RUBY_VERSION}-p#{RUBY_PATCHLEVEL}"
|
|
|
|
Or, for very old versions of Ruby which don't define `RUBY_PATCHLEVEL`:
|
|
|
|
:ruby puts RUBY_VERSION
|
|
|
|
Finally, beware that if you change your installation method for Command-T (for
|
|
example, switching from one plugin manager to another) you should verify that
|
|
you remove all of the files installed by the previous installation method; if
|
|
you fail to do this, Vim may end up executing the old code, invalidating all
|
|
your attempts to get Vim and Command-T using the same version of Ruby.
|
|
|
|
|
|
USAGE *command-t-usage*
|
|
|
|
Bring up the Command-T file window by typing:
|
|
|
|
<Leader>t
|
|
|
|
This mapping is set up automatically for you, provided you do not already have
|
|
a mapping for <Leader>t or |:CommandT|. You can also bring up the file window
|
|
by issuing the command:
|
|
|
|
:CommandT
|
|
|
|
A prompt will appear at the bottom of the screen along with a file window
|
|
showing all of the files in the current directory (as returned by the
|
|
|:pwd| command).
|
|
|
|
For the most efficient file navigation within a project it's recommended that
|
|
you |:cd| into the root directory of your project when starting to work on it.
|
|
If you wish to open a file from outside of the project folder you can pass in
|
|
an optional path argument (relative or absolute) to |:CommandT|:
|
|
|
|
:CommandT ../path/to/other/files
|
|
|
|
Type letters in the prompt to narrow down the selection, showing only the
|
|
files whose paths contain those letters in the specified order. Letters do not
|
|
need to appear consecutively in a path in order for it to be classified as a
|
|
match.
|
|
|
|
Once the desired file has been selected it can be opened by pressing <CR>.
|
|
(By default files are opened in the current window, but there are other
|
|
mappings that you can use to open in a vertical or horizontal split, or in
|
|
a new tab.) Note that if you have |'nohidden'| set and there are unsaved
|
|
changes in the current window when you press <CR> then opening in the current
|
|
window would fail; in this case Command-T will open the file in a new split.
|
|
|
|
The following mappings are active when the prompt has focus:
|
|
|
|
<BS> delete the character to the left of the cursor
|
|
<Del> delete the character at the cursor
|
|
<Left> move the cursor one character to the left
|
|
<C-h> move the cursor one character to the left
|
|
<Right> move the cursor one character to the right
|
|
<C-l> move the cursor one character to the right
|
|
<C-a> move the cursor to the start (left)
|
|
<C-e> move the cursor to the end (right)
|
|
<C-u> clear the contents of the prompt
|
|
<Tab> change focus to the file listing
|
|
|
|
The following mappings are active when the file listing has focus:
|
|
|
|
<Tab> change focus to the prompt
|
|
|
|
The following mappings are active when either the prompt or the file listing
|
|
has focus:
|
|
|
|
<CR> open the selected file
|
|
<C-CR> open the selected file in a new split window
|
|
<C-s> open the selected file in a new split window
|
|
<C-v> open the selected file in a new vertical split window
|
|
<C-t> open the selected file in a new tab
|
|
<C-j> select next file in the file listing
|
|
<C-n> select next file in the file listing
|
|
<Down> select next file in the file listing
|
|
<C-k> select previous file in the file listing
|
|
<C-p> select previous file in the file listing
|
|
<Up> select previous file in the file listing
|
|
<C-f> flush the cache (see |:CommandTFlush| for details)
|
|
<C-q> place the current matches in the quickfix window
|
|
<C-c> cancel (dismisses file listing)
|
|
|
|
The following is also available on terminals which support it:
|
|
|
|
<Esc> cancel (dismisses file listing)
|
|
|
|
Note that the default mappings can be overriden by setting options in your
|
|
`~/.vimrc` file (see the OPTIONS section for a full list of available options).
|
|
|
|
In addition, when the file listing has focus, typing a character will cause
|
|
the selection to jump to the first path which begins with that character.
|
|
Typing multiple characters consecutively can be used to distinguish between
|
|
paths which begin with the same prefix.
|
|
|
|
|
|
COMMANDS *command-t-commands*
|
|
|
|
*:CommandT*
|
|
|:CommandT| Brings up the Command-T file window, starting in the
|
|
current working directory as returned by the|:pwd|
|
|
command.
|
|
|
|
*:CommandTBuffer*
|
|
|:CommandTBuffer| Brings up the Command-T buffer window.
|
|
This works exactly like the standard file window,
|
|
except that the selection is limited to files that
|
|
you already have open in buffers.
|
|
|
|
*:CommandTMRU*
|
|
|:CommandTMRU| Brings up the Command-T buffer window, except that matches
|
|
are shown in MRU (most recently used) order. If you prefer to
|
|
use this over the normal buffer finder, I suggest overwriting
|
|
the standard mapping with a command like:
|
|
|
|
:nnoremap <silent> <leader>b :CommandTMRU<CR>
|
|
|
|
Note that Command-T only starts recording most recently used
|
|
buffers when you first use a Command-T command or mapping;
|
|
this is an optimization to improve startup time.
|
|
|
|
*:CommandTJumps*
|
|
|:CommandTJump| Brings up the Command-T jumplist window.
|
|
This works exactly like the standard file window,
|
|
except that the selection is limited to files that
|
|
you already have in the jumplist. Note that jumps
|
|
can persist across Vim sessions (see Vim's |jumplist|
|
|
documentation for more info).
|
|
|
|
*:CommandTTag*
|
|
|:CommandTTag| Brings up the Command-T window tags window, which can
|
|
be used to select from the tags, if any, returned by
|
|
Vim's |taglist()| function. See Vim's |tag| documentation
|
|
for general info on tags.
|
|
|
|
*:CommandTFlush*
|
|
|:CommandTFlush| Instructs the plug-in to flush its path cache, causing
|
|
the directory to be rescanned for new or deleted paths
|
|
the next time the file window is shown (pressing <C-f> when
|
|
a match listing is visible flushes the cache immediately; this
|
|
mapping is configurable via the |g:CommandTRefreshMap|
|
|
setting). In addition, all configuration settings are
|
|
re-evaluated, causing any changes made to settings via the
|
|
|:let| command to be picked up.
|
|
|
|
|
|
MAPPINGS *command-t-mappings*
|
|
|
|
By default Command-T comes with only two mappings:
|
|
|
|
<Leader>t bring up the Command-T file window
|
|
<Leader>b bring up the Command-T buffer window
|
|
|
|
However, Command-T won't overwrite a pre-existing mapping so if you prefer
|
|
to define different mappings use lines like these in your `~/.vimrc`:
|
|
|
|
nnoremap <silent> <Leader>t :CommandT<CR>
|
|
nnoremap <silent> <Leader>b :CommandTBuffer<CR>
|
|
|
|
Replacing "<Leader>t" or "<Leader>b" with your mapping of choice.
|
|
|
|
Note that in the case of MacVim you actually can map to Command-T (written
|
|
as <D-t> in Vim) in your `~/.gvimrc` file if you first unmap the existing menu
|
|
binding of Command-T to "New Tab":
|
|
|
|
if has("gui_macvim")
|
|
macmenu &File.New\ Tab key=<nop>
|
|
map <D-t> :CommandT<CR>
|
|
endif
|
|
|
|
When the Command-T window is active a number of other additional mappings
|
|
become available for doing things like moving between and selecting matches.
|
|
These are fully described above in the USAGE section, and settings for
|
|
overriding the mappings are listed below under OPTIONS.
|
|
|
|
|
|
OPTIONS *command-t-options*
|
|
|
|
A number of options may be set in your `~/.vimrc` to influence the behaviour of
|
|
the plug-in. To set an option, you include a line like this in your `~/.vimrc`:
|
|
|
|
let g:CommandTMaxFiles=20000
|
|
|
|
To have Command-T pick up new settings immediately (that is, without having
|
|
to restart Vim) you can issue the |:CommandTFlush| command after making
|
|
changes via |:let|.
|
|
|
|
Following is a list of all available options:
|
|
|
|
*g:CommandTMaxFiles*
|
|
|g:CommandTMaxFiles| number (default 30000)
|
|
|
|
The maximum number of files that will be considered when scanning the
|
|
current directory. Upon reaching this number scanning stops. This
|
|
limit applies only to file listings and is ignored for buffer
|
|
listings.
|
|
|
|
*g:CommandTMaxDepth*
|
|
|g:CommandTMaxDepth| number (default 15)
|
|
|
|
The maximum depth (levels of recursion) to be explored when scanning the
|
|
current directory. Any directories at levels beyond this depth will be
|
|
skipped.
|
|
|
|
*g:CommandTMaxCachedDirectories*
|
|
|g:CommandTMaxCachedDirectories| number (default 1)
|
|
|
|
The maximum number of directories whose contents should be cached when
|
|
recursively scanning. With the default value of 1, each time you change
|
|
directories the cache will be emptied and Command-T will have to
|
|
rescan. Higher values will make Command-T hold more directories in the
|
|
cache, bringing performance at the cost of memory usage. If set to 0,
|
|
there is no limit on the number of cached directories.
|
|
|
|
*g:CommandTMaxHeight*
|
|
|g:CommandTMaxHeight| number (default: 0)
|
|
|
|
The maximum height in lines the match window is allowed to expand to.
|
|
If set to 0, the window will occupy as much of the available space as
|
|
needed to show matching entries.
|
|
|
|
*g:CommandTInputDebounce*
|
|
|g:CommandTInputDebounce| number (default: 50)
|
|
|
|
The number of milliseconds to wait before updating the match listing
|
|
following a key-press. This can be used to avoid wasteful recomputation
|
|
when making a rapid series of key-presses in a directory with many tens
|
|
(or hundreds) of thousands of files.
|
|
|
|
*g:CommandTFileScanner*
|
|
|g:CommandTFileScanner| string (default: 'ruby')
|
|
|
|
The underlying scanner implementation that should be used to explore the
|
|
filesystem. Possible values are:
|
|
|
|
- "ruby": uses built-in Ruby and should work everywhere, albeit slowly
|
|
on large (many tens of thousands of files) hierarchies.
|
|
|
|
- "find": uses the command-line tool of the same name, which can be much
|
|
faster on large projects because it is written in pure C, but may not
|
|
work on systems without the tool or with an incompatible version of
|
|
the tool.
|
|
|
|
- "git": uses `git ls-files` to quickly produce a list of files; when
|
|
Git isn't available or the path being searched is not inside a Git
|
|
repository falls back to "find".
|
|
|
|
- "watchman": uses Watchman (https://github.com/facebook/watchman) if
|
|
available; otherwise falls back to "find". Note that this scanner is
|
|
intended for use with very large hierarchies (hundreds of thousands of
|
|
files) and so the task of deciding which files should be included is
|
|
entirely delegated to Watchman; this means that settings which
|
|
Command-T would usually consult, such as 'wildignore' and
|
|
|g:CommandTScanDotDirectories| are ignored.
|
|
|
|
*g:CommandTTraverseSCM*
|
|
|g:CommandTTraverseSCM| string (default: 'file')
|
|
|
|
Instructs Command-T how to chose a root path when opening a file finder
|
|
without an explicit path argument. Possible values are:
|
|
|
|
- "file": starting from the file currently being edited, traverse
|
|
upwards through the filesystem hierarchy until you find an SCM root
|
|
(as indicated by the presence of a ".git", ".hg" or similar directory)
|
|
and use that as the base path. If no such root is found, fallback to
|
|
using Vim's present working directory as a root. The list of SCM
|
|
directories that Command-T uses to detect an SCM root can be
|
|
customized with the |g:CommandTSCMDirectories| option.
|
|
|
|
- "dir": traverse upwards looking for an SCM root just like the "file"
|
|
setting (above), but instead of starting from the file currently being
|
|
edited, start from Vim's present working directory instead.
|
|
|
|
- "pwd": use Vim's present working directory as a root (ie. attempt no
|
|
traversal).
|
|
|
|
*g:CommandTSCMDirectories*
|
|
|g:CommandTSCMDirectories| string (default: '.git,.hg,.svn,.bzr,_darcs')
|
|
|
|
The marker directories that Command-T will use to identify SCM roots
|
|
during traversal (see |g:CommandTTraverseSCM| above).
|
|
|
|
|
|
*g:CommandTMinHeight*
|
|
|g:CommandTMinHeight| number (default: 0)
|
|
|
|
The minimum height in lines the match window is allowed to shrink to.
|
|
If set to 0, will default to a single line. If set above the max height,
|
|
will default to |g:CommandTMaxHeight|.
|
|
|
|
*g:CommandTAlwaysShowDotFiles*
|
|
|g:CommandTAlwaysShowDotFiles| boolean (default: 0)
|
|
|
|
When showing the file listing Command-T will by default show dot-files
|
|
only if the entered search string contains a dot that could cause a
|
|
dot-file to match. When set to a non-zero value, this setting instructs
|
|
Command-T to always include matching dot-files in the match list
|
|
regardless of whether the search string contains a dot. See also
|
|
|g:CommandTNeverShowDotFiles|. Note that this setting only influences
|
|
the file listing; the buffer listing treats dot-files like any other
|
|
file.
|
|
|
|
*g:CommandTNeverShowDotFiles*
|
|
|g:CommandTNeverShowDotFiles| boolean (default: 0)
|
|
|
|
In the file listing, Command-T will by default show dot-files if the
|
|
entered search string contains a dot that could cause a dot-file to
|
|
match. When set to a non-zero value, this setting instructs Command-T to
|
|
never show dot-files under any circumstances. Note that it is
|
|
contradictory to set both this setting and
|
|
|g:CommandTAlwaysShowDotFiles| to true, and if you do so Vim will suffer
|
|
from headaches, nervous twitches, and sudden mood swings. This setting
|
|
has no effect in buffer listings, where dot files are treated like any
|
|
other file.
|
|
|
|
*g:CommandTScanDotDirectories*
|
|
|g:CommandTScanDotDirectories| boolean (default: 0)
|
|
|
|
Normally Command-T will not recurse into "dot-directories" (directories
|
|
whose names begin with a dot) while performing its initial scan. Set
|
|
this setting to a non-zero value to override this behavior and recurse.
|
|
Note that this setting is completely independent of the
|
|
|g:CommandTAlwaysShowDotFiles| and |g:CommandTNeverShowDotFiles|
|
|
settings; those apply only to the selection and display of matches
|
|
(after scanning has been performed), whereas
|
|
|g:CommandTScanDotDirectories| affects the behaviour at scan-time.
|
|
|
|
Note also that even with this setting off you can still use Command-T to
|
|
open files inside a "dot-directory" such as `~/.vim`, but you have to use
|
|
the |:cd| command to change into that directory first. For example:
|
|
|
|
:cd ~/.vim
|
|
:CommandT
|
|
|
|
*g:CommandTMatchWindowAtTop*
|
|
|g:CommandTMatchWindowAtTop| boolean (default: 0)
|
|
|
|
When this setting is off (the default) the match window will appear at
|
|
the bottom so as to keep it near to the prompt. Turning it on causes the
|
|
match window to appear at the top instead. This may be preferable if you
|
|
want the best match (usually the first one) to appear in a fixed location
|
|
on the screen rather than moving as the number of matches changes during
|
|
typing.
|
|
|
|
*g:CommandTMatchWindowReverse*
|
|
|g:CommandTMatchWindowReverse| boolean (default: 0)
|
|
|
|
When this setting is off (the default) the matches will appear from
|
|
top to bottom with the topmost being selected. Turning it on causes the
|
|
matches to be reversed so the best match is at the bottom and the
|
|
initially selected match is the bottom most. This may be preferable if
|
|
you want the best match to appear in a fixed location on the screen
|
|
but still be near the prompt at the bottom.
|
|
|
|
*g:CommandTTagIncludeFilenames*
|
|
|g:CommandTTagIncludeFilenames| boolean (default: 0)
|
|
|
|
When this setting is off (the default) the matches in the |:CommandTTag|
|
|
listing do not include filenames.
|
|
|
|
*g:CommandTHighlightColor*
|
|
|g:CommandTHighlightColor| string (default: 'PmenuSel')
|
|
|
|
Specifies the highlight color that will be used to show the currently
|
|
selected item in the match listing window.
|
|
|
|
*g:CommandTWildIgnore*
|
|
|g:CommandTWildIgnore| string (default: none)
|
|
|
|
Optionally override Vim's global |'wildignore'| setting during Command-T
|
|
searches. If you wish to supplement rather than replace the global
|
|
setting, you can use a syntax like:
|
|
|
|
let g:CommandTWildIgnore=&wildignore . ",**/bower_components/*"
|
|
|
|
See also |command-t-wildignore|.
|
|
|
|
*g:CommandTIgnoreCase*
|
|
|g:CommandTIgnoreCase| boolean (default: 1)
|
|
|
|
Ignore case when searching. Defaults to on, which means that searching
|
|
is case-insensitive by default. See also |g:CommandTSmartCase|.
|
|
|
|
*g:CommandTSmartCase*
|
|
|g:CommandTSmartCase| boolean (default: none)
|
|
|
|
Override the |g:CommandTIgnoreCase| setting if the search pattern
|
|
contains uppercase characters, forcing the match to be case-sensitive.
|
|
If unset (which is the default), the value of the Vim |'smartcase'|
|
|
setting will be used instead.
|
|
|
|
*g:CommandTAcceptSelectionCommand*
|
|
|g:CommandTAcceptSelectionCommand| string (default: 'e')
|
|
|
|
The Vim command that will be used to open a selection from the match
|
|
listing (via |g:CommandTAcceptSelectionMap|).
|
|
|
|
For an example of how this can be used to apply arbitrarily complex
|
|
logic, see the example in |g:CommandTAcceptSelectionTabCommand| below.
|
|
|
|
*g:CommandTAcceptSelectionTabCommand*
|
|
|g:CommandTAcceptSelectionTabCommand| string (default: 'tabe')
|
|
|
|
The Vim command that will be used to open a selection from the match
|
|
listing in a new tab (via |g:CommandTAcceptSelectionSplitMap|).
|
|
|
|
For example, this can be used to switch to an existing buffer (rather
|
|
than opening a duplicate buffer with the selection in a new tab) with
|
|
configuration such as the following:
|
|
|
|
set switchbuf=usetab
|
|
|
|
function! GotoOrOpen(...)
|
|
for file in a:000
|
|
if bufwinnr(file) != -1
|
|
exec "sb " . file
|
|
else
|
|
exec "tabe " . file
|
|
endif
|
|
endfor
|
|
endfunction
|
|
|
|
command! -nargs=+ GotoOrOpen call GotoOrOpen("<args>")
|
|
|
|
let g:CommandTAcceptSelectionTabCommand = 'GotoOrOpen'
|
|
|
|
For a slightly more comprehensive example, see: https://wt.pe/e
|
|
|
|
*g:CommandTAcceptSelectionSplitCommand*
|
|
|g:CommandTAcceptSelectionSplitCommand| string (default: 'sp')
|
|
|
|
The Vim command that will be used to open a selection from the match
|
|
listing in a split (via |g:CommandTAcceptSelectionVSplitMap|).
|
|
|
|
For an example of how this can be used to apply arbitrarily complex
|
|
logic, see the example in |g:CommandTAcceptSelectionTabCommand| above.
|
|
|
|
*g:CommandTAcceptSelectionVsplitCommand*
|
|
string (default: 'vs')
|
|
|g:CommandTAcceptSelectionVSplitCommand|
|
|
|
|
The Vim command that will be used to open a selection from the match
|
|
listing in a vertical split (via |g:CommandTAcceptSelectionVSplitMap|).
|
|
|
|
For an example of how this can be used to apply arbitrarily complex
|
|
logic, see the example in |g:CommandTAcceptSelectionTabCommand| above.
|
|
|
|
As well as the basic options listed above, there are a number of settings that
|
|
can be used to override the default key mappings used by Command-T. For
|
|
example, to set <C-x> as the mapping for cancelling (dismissing) the Command-T
|
|
window, you would add the following to your `~/.vimrc`:
|
|
|
|
let g:CommandTCancelMap='<C-x>'
|
|
|
|
Multiple, alternative mappings may be specified using list syntax:
|
|
|
|
let g:CommandTCancelMap=['<C-x>', '<C-c>']
|
|
|
|
Following is a list of all map settings and their defaults:
|
|
|
|
Setting Default mapping(s)
|
|
|
|
*g:CommandTBackspaceMap*
|
|
|g:CommandTBackspaceMap| <BS>
|
|
|
|
*g:CommandTDeleteMap*
|
|
|g:CommandTDeleteMap| <Del>
|
|
|
|
*g:CommandTAcceptSelectionMap*
|
|
|g:CommandTAcceptSelectionMap| <CR>
|
|
|
|
*g:CommandTAcceptSelectionSplitMap*
|
|
|g:CommandTAcceptSelectionSplitMap| <C-CR>
|
|
<C-s>
|
|
|
|
*g:CommandTAcceptSelectionTabMap*
|
|
|g:CommandTAcceptSelectionTabMap| <C-t>
|
|
|
|
*g:CommandTAcceptSelectionVSplitMap*
|
|
|g:CommandTAcceptSelectionVSplitMap| <C-v>
|
|
|
|
*g:CommandTToggleFocusMap*
|
|
|g:CommandTToggleFocusMap| <Tab>
|
|
|
|
*g:CommandTCancelMap*
|
|
|g:CommandTCancelMap| <C-c>
|
|
<Esc> (not on all terminals)
|
|
|
|
*g:CommandTSelectNextMap*
|
|
|g:CommandTSelectNextMap| <C-n>
|
|
<C-j>
|
|
<Down>
|
|
|
|
*g:CommandTSelectPrevMap*
|
|
|g:CommandTSelectPrevMap| <C-p>
|
|
<C-k>
|
|
<Up>
|
|
|
|
*g:CommandTClearMap*
|
|
|g:CommandTClearMap| <C-u>
|
|
|
|
*g:CommandTClearPrevWordMap*
|
|
|g:CommandTClearPrevWordMap| <C-w>
|
|
|
|
*g:CommandTRefreshMap*
|
|
|g:CommandTRefreshMap| <C-f>
|
|
|
|
*g:CommandTQuickfixMap*
|
|
|g:CommandTQuickfixMap| <C-q>
|
|
|
|
*g:CommandTCursorLeftMap*
|
|
|g:CommandTCursorLeftMap| <Left>
|
|
<C-h>
|
|
|
|
*g:CommandTCursorRightMap*
|
|
|g:CommandTCursorRightMap| <Right>
|
|
<C-l>
|
|
|
|
*g:CommandTCursorEndMap*
|
|
|g:CommandTCursorEndMap| <C-e>
|
|
|
|
*g:CommandTCursorStartMap*
|
|
|g:CommandTCursorStartMap| <C-a>
|
|
|
|
In addition to the options provided by Command-T itself, some of Vim's own
|
|
settings can be used to control behavior:
|
|
|
|
*command-t-wildignore*
|
|
|'wildignore'| string (default: '')
|
|
|
|
Vim's |'wildignore'| setting is used to determine which files should be
|
|
excluded from listings. This is a comma-separated list of glob patterns.
|
|
It defaults to the empty string, but common settings include "*.o,*.obj"
|
|
(to exclude object files) or "**/.git/*,**/.svn/*" (to exclude SCM
|
|
metadata directories). For example:
|
|
|
|
:set wildignore+=*.o,*.obj
|
|
|
|
A pattern such as "vendor/rails/**" would exclude all files and
|
|
subdirectories inside the "vendor/rails" directory (relative to
|
|
directory Command-T starts in).
|
|
|
|
See the |'wildignore'| documentation for more information.
|
|
|
|
If you want to influence Command-T's file exclusion behavior without
|
|
changing your global |'wildignore'| setting, you can use the
|
|
|g:CommandTWildIgnore| setting to apply an override that takes effect
|
|
only during Command-T searches.
|
|
|
|
Note that there are some differences among file scanners
|
|
(see |g:CommandTFileScanner|) with respect to 'wildignore' handling:
|
|
|
|
- The default "ruby" scanner explores the filesystem recursively using a
|
|
depth-first search, and any directory (or subdirectory) which matches
|
|
the 'wildignore' pattern is not explored. So, if your 'wildignore'
|
|
contains "node_modules" then that entire sub-hierarchy will be
|
|
ignored. Additionally, wildcard patterns like "node_modules/**" or
|
|
"**/node_modules/*" will cause the entire sub-hierarchy to be ignored.
|
|
|
|
- The "git" and "find" scanners apply 'wildignore' filtering only after
|
|
completing their scans. Filtering only applies to files and not
|
|
directories. This means that in the "node_modules" example case, the
|
|
"node_modules" directory is not considered itself, and when we examine
|
|
a file like "node_modules/foo/bar" the "node_modules" pattern does
|
|
not match it (because "bar" does not match it). To exclude
|
|
any "node_modules" directory anywhere in the hierarchy and all of its
|
|
descendants we must use a pattern like "**/node_modules/*". To do this
|
|
only for a top-level "node_modules", use "node_modules/**".
|
|
|
|
- The "watchman" scanner is intended for use with massive hierarchies
|
|
where speed is of the utmost import, so it doesn't consult
|
|
'wildignore' at all.
|
|
|
|
|
|
FAQ *command-t-faq*
|
|
|
|
Why does my build fail with "unknown argument -multiply_definedsuppress"? ~
|
|
|
|
You may see this on OS X Mavericks when building with the Clang compiler
|
|
against the system Ruby. This is an unfortunate Apple bug that breaks
|
|
compilation of many Ruby gems with native extensions on Mavericks. It has been
|
|
worked around in the upstream Ruby version, but won't be fixed in OS X until
|
|
Apple updates their supplied version of Ruby (most likely this won't be until
|
|
the next major release):
|
|
|
|
https://bugs.ruby-lang.org/issues/9624
|
|
|
|
Workarounds include building your own Ruby (and then your own Vim and
|
|
Command-T), or more simply, building with the following `ARCHFLAGS` set:
|
|
|
|
ARCHFLAGS=-Wno-error=unused-command-line-argument-hard-error-in-future ruby extconf.rb
|
|
make
|
|
|
|
Why can't I open in a split with <C-CR> and <C-s> in the terminal? ~
|
|
|
|
It's likely that <C-CR> won't work in most terminals, because the keycode that
|
|
is sent to processes running inside them is identical to <CR>; when you type
|
|
<C-CR>, terminal Vim literally "sees" <CR>. Unfortunately, there is no
|
|
workaround for this.
|
|
|
|
If you find that <C-s> also doesn't work the most likely explanation is that
|
|
XON/XOFF flow control is enabled; this is the default in many environments.
|
|
This means that when you press <C-s> all input to the terminal is suspended
|
|
until you release it by hitting <C-q>. While input is suspended you may think
|
|
your terminal has frozen, but it hasn't.
|
|
|
|
To disable flow control, add the following to your `.zshrc` or
|
|
`.bash_profile`:
|
|
|
|
stty -ixon
|
|
|
|
See the `stty` man page for more details.
|
|
|
|
Why doesn't the Escape key close the match listing in terminal Vim? ~
|
|
|
|
In some terminals such as xterm the Escape key misbehaves, so Command-T
|
|
doesn't set up a mapping for it. If you want to try using the escape key
|
|
anyway, you can add something like the following to your `~/.vimrc` file:
|
|
|
|
if &term =~ "xterm" || &term =~ "screen"
|
|
let g:CommandTCancelMap = ['<ESC>', '<C-c>']
|
|
endif
|
|
|
|
This configuration has worked for me with recent versions of Vim on multiple
|
|
platforms (OS X, CentOS etc).
|
|
|
|
|
|
TIPS *command-t-tips*
|
|
|
|
Working with very large repositories ~
|
|
|
|
One of the primary motivations for writing Command-T was to get fast, robust
|
|
high-quality matches even on large hierarchies. The larger the hierarchy, the
|
|
more important having good file navigation becomes. This is why Command-T's
|
|
performance-critical sections are written in C. This requires a compilation
|
|
step and makes Command-T harder to install than similar plug-ins which are
|
|
written in pure Vimscript, and can be a disincentive against use. This is a
|
|
conscious trade-off; the goal isn't to have as many users as possible, but
|
|
rather to provide the best performance at the highest quality.
|
|
|
|
The speed of the core is high enough that Command-T can afford to burn a bunch
|
|
of extra cycles -- using its recursive matching algorithm -- looking for a
|
|
higher-quality, more intuitive ranking of search results. Again, the larger
|
|
the hierarchy, the more important the quality of result ranking becomes.
|
|
|
|
Nevertheless, for extremely large hierarchies (of the order of 500,000 files)
|
|
some tuning is required in order to get useful and usable performance levels.
|
|
Here are some useful example settings:
|
|
|
|
let g:CommandTMaxHeight = 30
|
|
|
|
You want the match listing window to be large enough that you can get useful
|
|
feedback about how your search query is going; in large hierarchies there may
|
|
be many, many matches for a given query. At the same time, you don't want Vim
|
|
wasting valuable cycles repainting a large portion of the screen area,
|
|
especially on a large display. Setting the limit to 30 or similar is a
|
|
reasonable compromise.
|
|
|
|
let g:CommandTMaxFiles = 500000
|
|
|
|
The default limit of 30,000 files prevents Command-T from "seeing" many of the
|
|
files in a large directory hierarchy so you need to increase this limit.
|
|
|
|
let g:CommandTInputDebounce = 200
|
|
|
|
Wait for 200ms of keyboard inactivity before computing search results. For
|
|
example, if you are enter "foobar" quickly (ie. within 1 second), there is
|
|
little sense in fetching the results for "f", "fo", "foo", "foob", "fooba" and
|
|
finally "foobar". Instead, we can just fetch the results for "foobar". This
|
|
setting trades off some immediate responsiveness at the micro level for
|
|
better performance (real and perceived) and a better search experience
|
|
overall.
|
|
|
|
let g:CommandTFileScanner = 'watchman'
|
|
|
|
On a large hierarchy with of the order of 500,000 files, scanning a directory
|
|
tree with a tool like the `find` executable may take literally minutes with a
|
|
cold cache. Once the cache is warm, the same `find` run may take only a second
|
|
or two. Command-T provides a "find" scanner to leverage this performance, but
|
|
there is still massive overhead in passing the results through Vim internal
|
|
functions that apply 'wildignore' settings and such, so for truly immense
|
|
repos the "watchman" scanner is the tool of choice.
|
|
|
|
This scanner delegates the task of finding files to Facebook's `watchman` tool
|
|
(https://github.com/facebook/watchman), which can return results for a 500,000
|
|
file hierarchy within about a second.
|
|
|
|
Note that Watchman has a range of configuration options that can be applied by
|
|
files such as `/etc/watchman.json` or per-direcory `.watchmanconfig` files and
|
|
which may affect how Command-T works. For example, if your configuration has a
|
|
`root_restrict_files` setting that makes Watchman only work with roots that
|
|
look like Git or Mercurial repos, then Command-T will fall back to using the
|
|
"find" scanner any time you invoke it on a non-repo directory. For
|
|
simplicity's sake, it is probably a good idea to use Vim and Command-T
|
|
anchored at the root level of your repository in any case.
|
|
|
|
let g:CommandTMaxCachedDirectories = 10
|
|
|
|
Command-T will internally cache up to 10 different directories, so even if you
|
|
|cd| repeatedly, it should only need to scan each directory once.
|
|
|
|
It's advisable to keep a long-running Vim instance in place and let it cache
|
|
the directory listings rather than repeatedly closing and re-opening Vim in
|
|
order to edit every file. On those occasions when you do need to flush the
|
|
cache (ie. with |CommandTFlush| or <C-f> in the match listing window), use of
|
|
the Watchman scanner should make the delay barely noticeable.
|
|
|
|
As noted in the introduction, Command-T works best when you adopt a
|
|
"path-centric" mentality. This is especially true on very large hierarchies.
|
|
For example, if you're looking for a file at:
|
|
|
|
lib/third-party/adapters/restful-services/foobar/foobar-manager.js
|
|
|
|
you'll be able to narrow your search results down more narrowly if you search
|
|
with a query like "librestfoofooman" than "foobar-manager.js". This evidently
|
|
requires that you know where the file you're wanting to open exists, but
|
|
again, this is a concious design decision: Command-T is made to enable people
|
|
who know what they want to open and where it is to open it as quickly as
|
|
possible; other tools such as NERDTree exist for visually exploring an unknown
|
|
hierarchy.
|
|
|
|
Over time, you will get a feel for how economical you can afford to be with
|
|
your search queries in a given repo. In the example above, if "foo" is not a
|
|
very common pattern in your hierarchy, then you may find that you can find
|
|
what you need with a very concise query such as "foomanjs". With time, this
|
|
kind of ongoing calibration will come quite naturally.
|
|
|
|
Finally, it is important to be on a relatively recent version of Command-T to
|
|
fully benefit from the available performance enhancements:
|
|
|
|
- version 1.9 (May 2014) tweaked memoization algorithm for a 10% speed boost
|
|
- version 1.8 (March 2014) sped up the Watchman file scanner by switching its
|
|
communication from the JSON to the binary Watchman protocol
|
|
- version 1.7 (February 2014) added the |g:CommandTInputDebounce| and
|
|
|g:CommandTFileScanner| settings, along with support for the Watchman file
|
|
scanner
|
|
- version 1.6 (December 2013) added parallelized search
|
|
- version 1.5 (September 2013) added memoization to the matching algorithm,
|
|
improving general performance on large hierarchies, but delivering
|
|
spectacular gains on hierarchies with "pathological" characteristics that
|
|
lead the algorithm to exhibit degenerate behavior
|
|
|
|
AUTHORS *command-t-authors*
|
|
|
|
Command-T is written and maintained by Greg Hurrell <greg@hurrell.net>.
|
|
Other contributors that have submitted patches include (in alphabetical
|
|
order):
|
|
|
|
Abhinav Gupta Lucas de Vries Roland Puntaier
|
|
Aleksandrs Ļedovskis Marcus Brito Ross Lagerwall
|
|
Andy Waite Marian Schubert Scott Bronson
|
|
Anthony Panozzo Matthew Todd Seth Fowler
|
|
Artem Nezvigin Mike Lundy Shlomi Fish
|
|
Ben Osheroff Nadav Samet Steven Moazami
|
|
Daniel Hahler Nate Kane Sung Pae
|
|
David Szotten Nicholas Alpi Thomas Pelletier
|
|
Emily Strickland Noon Silk Ton van den Heuvel
|
|
Felix Tjandrawibawa Ole Petter Bang Victor Hugo Borja
|
|
Gary Bernhardt Patrick Hayes Vít Ondruch
|
|
Ivan Ukhov Paul Jolly Woody Peterson
|
|
Jacek Wysocki Pavel Sergeev Yan Pritzker
|
|
Jeff Kreeftmeijer Rainux Luo Yiding Jia
|
|
Kevin Webster Richard Feldman Zak Johnson
|
|
|
|
As this was the first Vim plug-in I had ever written I was heavily influenced
|
|
by the design of the LustyExplorer plug-in by Stephen Bach, which I understand
|
|
was one of the largest Ruby-based Vim plug-ins at the time.
|
|
|
|
While the Command-T codebase doesn't contain any code directly copied from
|
|
LustyExplorer, I did use it as a reference for answers to basic questions (like
|
|
"How do you do 'X' in a Ruby-based Vim plug-in?"), and also copied some basic
|
|
architectural decisions (like the division of the code into Prompt, Settings
|
|
and MatchWindow classes).
|
|
|
|
LustyExplorer is available from:
|
|
|
|
http://www.vim.org/scripts/script.php?script_id=1890
|
|
|
|
|
|
DEVELOPMENT *command-t-development*
|
|
|
|
Development in progress can be inspected via the project's Git web-based
|
|
repository browser at:
|
|
|
|
https://wincent.com/repos/command-t
|
|
|
|
the clone URL for which is:
|
|
|
|
git://git.wincent.com/command-t.git
|
|
|
|
Mirrors exist on GitHub and Gitorious; these are automatically updated once
|
|
per hour from the authoritative repository:
|
|
|
|
https://github.com/wincent/command-t
|
|
https://gitorious.org/command-t/command-t
|
|
|
|
Patches are welcome via the usual mechanisms (pull requests, email, posting to
|
|
the project issue tracker etc).
|
|
|
|
As many users choose to track Command-T using Pathogen or similar, which often
|
|
means running a version later than the last official release, the intention is
|
|
that the "master" branch should be kept in a stable and reliable state as much
|
|
as possible.
|
|
|
|
Riskier changes are first cooked on the "next" branch for a period before
|
|
being merged into master. You can track this branch if you're feeling wild and
|
|
experimental, but note that the "next" branch may periodically be rewound
|
|
(force-updated) to keep it in sync with the "master" branch after each
|
|
official release.
|
|
|
|
|
|
WEBSITE *command-t-website*
|
|
|
|
The official website for Command-T is:
|
|
|
|
https://wincent.com/products/command-t
|
|
|
|
The latest release will always be available from there.
|
|
|
|
A copy of each release is also available from the official Vim scripts site
|
|
at:
|
|
|
|
http://www.vim.org/scripts/script.php?script_id=3025
|
|
|
|
Bug reports should be submitted to the issue tracker at:
|
|
|
|
https://wincent.com/issues
|
|
|
|
|
|
LICENSE *command-t-license*
|
|
|
|
Copyright 2010-2014 Greg Hurrell. All rights reserved.
|
|
|
|
Redistribution and use in source and binary forms, with or without
|
|
modification, are permitted provided that the following conditions are met:
|
|
|
|
1. Redistributions of source code must retain the above copyright notice,
|
|
this list of conditions and the following disclaimer.
|
|
2. Redistributions in binary form must reproduce the above copyright notice,
|
|
this list of conditions and the following disclaimer in the documentation
|
|
and/or other materials provided with the distribution.
|
|
|
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
|
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
|
|
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
POSSIBILITY OF SUCH DAMAGE.
|
|
|
|
HISTORY *command-t-history*
|
|
|
|
1.11.2 (2 September 2014)
|
|
|
|
- fix error while using Command-T outside of an SCM repo (bug present since
|
|
1.11.1)
|
|
|
|
1.11.1 (29 August 2014)
|
|
|
|
- compatibility fixes with Ruby 1.8.6 (patch from Emily Strickland)
|
|
- compatibility fixes with Ruby 1.8.5
|
|
- fix 'wildignore' being ignored (bug present since 1.11)
|
|
- fix current working directory being ignored when |g:CommandTTraverseSCM| is
|
|
set to "pwd" (bug present since 1.11)
|
|
- performance improvements
|
|
|
|
1.11 (15 August 2014)
|
|
|
|
- improve edge-case handling in match results window code (patches from
|
|
Richard Feldman)
|
|
- add "git" file scanner (patch from Patrick Hayes)
|
|
- speed-up when 'wildignore' is unset (patch from Patrick Hayes)
|
|
- add |g:CommandTTraverseSCM| setting which anchors Command-T's file finder to
|
|
the nearest SCM directory (based on patches from David Szotten and Ben
|
|
Osheroff)
|
|
- add AppStream metadata (patch from Vít Ondruch)
|
|
|
|
1.10 (15 July 2014)
|
|
|
|
- improve tag finder performance by caching tag lists (patch from Artem
|
|
Nezvigin)
|
|
- consider the |'autowriteall'| option when deciding whether to open a file in
|
|
a split
|
|
- make selection acceptance commands configurable (patch from Ole Petter Bang)
|
|
- add <C-w> mapping to delete previous word of the match prompt (patch from
|
|
Kevin Webster)
|
|
- try harder to always clear status line after closing the match listing
|
|
(patch from Ton van den Heuvel)
|
|
- don't allow MRU autocommands to produce errors when the extension has not
|
|
been compiled
|
|
- add |g:CommandTIgnoreCase| and |g:CommandTSmartCase| options, providing
|
|
support for case-sensitive matching (based on patch from Jacek Wysocki)
|
|
|
|
1.9.1 (30 May 2014)
|
|
|
|
- include the file in the release vimball archive that was missing from the
|
|
1.9 release
|
|
|
|
1.9 (25 May 2014)
|
|
|
|
- improved startup time using Vim's autload mechanism (patch from Ross
|
|
Lagerwall)
|
|
- added MRU (most-recently-used) buffer finder (patch from Ton van den Heuvel)
|
|
- fixed edge case in matching algorithm which could cause spurious matches
|
|
with queries containing repeated characters
|
|
- fixed slight positive bias in the match scoring algorithm's weighting of
|
|
matching characters based on distance from last match
|
|
- tune memoization in match scoring algorithm, yielding a more than 10% speed
|
|
boost
|
|
|
|
1.8 (31 March 2014)
|
|
|
|
- taught Watchman file scanner to use the binary protocol instead of JSON,
|
|
roughly doubling its speed
|
|
- build changes to accommodate MinGW (patch from Roland Puntaier)
|
|
|
|
1.7 (9 March 2014)
|
|
|
|
- added |g:CommandTInputDebounce|, which can be used to improve responsiveness
|
|
in large file hierarchies (based on patch from Yiding Jia)
|
|
- added a potentially faster file scanner which uses the `find` executable
|
|
(based on patch from Yiding Jia)
|
|
- added a file scanner that knows how to talk to Watchman
|
|
(https://github.com/facebook/watchman)
|
|
- added |g:CommandTFileScanner|, which can be used to switch file scanners
|
|
- fix processor count detection on some platforms (patch from Pavel Sergeev)
|
|
|
|
1.6.1 (22 December 2013)
|
|
|
|
- defer processor count detection until runtime (makes it possible to sensibly
|
|
build Command-T on one machine and use it on another)
|
|
|
|
1.6 (16 December 2013)
|
|
|
|
- on systems with POSIX threads (such as OS X and Linux), Command-T will use
|
|
threads to compute match results in parallel, resulting in a large speed
|
|
boost that is especially noticeable when navigating large projects
|
|
|
|
1.5.1 (23 September 2013)
|
|
|
|
- exclude large benchmark fixture file from source exports (patch from Vít
|
|
Ondruch)
|
|
|
|
1.5 (18 September 2013)
|
|
|
|
- don't scan "pathological" filesystem structures (ie. circular or
|
|
self-referential symlinks; patch from Marcus Brito)
|
|
- gracefully handle files starting with "+" (patch from Ivan Ukhov)
|
|
- switch default selection highlight color for better readability (suggestion
|
|
from André Arko), but make it possible to configure via the
|
|
|g:CommandTHighlightColor| setting
|
|
- added a mapping to take the current matches and put then in the quickfix
|
|
window
|
|
- performance improvements, particularly noticeable with large file
|
|
hierarchies
|
|
- added |g:CommandTWildIgnore| setting (patch from Paul Jolly)
|
|
|
|
1.4 (20 June 2012)
|
|
|
|
- added |:CommandTTag| command (patches from Noon Silk)
|
|
- turn off |'colorcolumn'| and |'relativenumber'| in the match window (patch
|
|
from Jeff Kreeftmeijer)
|
|
- documentation update (patch from Nicholas Alpi)
|
|
- added |:CommandTMinHeight| option (patch from Nate Kane)
|
|
- highlight (by underlining) matched characters in the match listing (requires
|
|
Vim to have been compiled with the +conceal feature, which is available in
|
|
Vim 7.3 or later; patch from Steven Moazami)
|
|
- added the ability to flush the cache while the match window is open using
|
|
<C-f>
|
|
|
|
1.3.1 (18 December 2011)
|
|
|
|
- fix jumplist navigation under Ruby 1.9.x (patch from Woody Peterson)
|
|
|
|
1.3 (27 November 2011)
|
|
|
|
- added the option to maintain multiple caches when changing among
|
|
directories; see the accompanying |g:CommandTMaxCachedDirectories| setting
|
|
- added the ability to navigate using the Vim jumplist (patch from Marian
|
|
Schubert)
|
|
|
|
1.2.1 (30 April 2011)
|
|
|
|
- Remove duplicate copy of the documentation that was causing "Duplicate tag"
|
|
errors
|
|
- Mitigate issue with distracting blinking cursor in non-GUI versions of Vim
|
|
(patch from Steven Moazami)
|
|
|
|
1.2 (30 April 2011)
|
|
|
|
- added |g:CommandTMatchWindowReverse| option, to reverse the order of items
|
|
in the match listing (patch from Steven Moazami)
|
|
|
|
1.1b2 (26 March 2011)
|
|
|
|
- fix a glitch in the release process; the plugin itself is unchanged since
|
|
1.1b
|
|
|
|
1.1b (26 March 2011)
|
|
|
|
- add |:CommandTBuffer| command for quickly selecting among open buffers
|
|
|
|
1.0.1 (5 January 2011)
|
|
|
|
- work around bug when mapping |:CommandTFlush|, wherein the default mapping
|
|
for |:CommandT| would not be set up
|
|
- clean up when leaving the Command-T buffer via unexpected means (such as
|
|
with <C-W k> or similar)
|
|
|
|
1.0 (26 November 2010)
|
|
|
|
- make relative path simplification work on Windows
|
|
|
|
1.0b (5 November 2010)
|
|
|
|
- work around platform-specific Vim 7.3 bug seen by some users (wherein
|
|
Vim always falsely reports to Ruby that the buffer numbers is 0)
|
|
- re-use the buffer that is used to show the match listing, rather than
|
|
throwing it away and recreating it each time Command-T is shown; this
|
|
stops the buffer numbers from creeping up needlessly
|
|
|
|
0.9 (8 October 2010)
|
|
|
|
- use relative paths when opening files inside the current working directory
|
|
in order to keep buffer listings as brief as possible (patch from Matthew
|
|
Todd)
|
|
|
|
0.8.1 (14 September 2010)
|
|
|
|
- fix mapping issues for users who have set |'notimeout'| (patch from Sung
|
|
Pae)
|
|
|
|
0.8 (19 August 2010)
|
|
|
|
- overrides for the default mappings can now be lists of strings, allowing
|
|
multiple mappings to be defined for any given action
|
|
- <Leader>t mapping only set up if no other map for |:CommandT| exists
|
|
(patch from Scott Bronson)
|
|
- prevent folds from appearing in the match listing
|
|
- tweaks to avoid the likelihood of "Not enough room" errors when trying to
|
|
open files
|
|
- watch out for "nil" windows when restoring window dimensions
|
|
- optimizations (avoid some repeated downcasing)
|
|
- move all Ruby files under the "command-t" subdirectory and avoid polluting
|
|
the "Vim" module namespace
|
|
|
|
0.8b (11 July 2010)
|
|
|
|
- large overhaul of the scoring algorithm to make the ordering of returned
|
|
results more intuitive; given the scope of the changes and room for
|
|
optimization of the new algorithm, this release is labelled as "beta"
|
|
|
|
0.7 (10 June 2010)
|
|
|
|
- handle more |'wildignore'| patterns by delegating to Vim's own |expand()|
|
|
function; with this change it is now viable to exclude patterns such as
|
|
'vendor/rails/**' in addition to filename-only patterns like '*.o' and
|
|
'.git' (patch from Mike Lundy)
|
|
- always sort results alphabetically for empty search strings; this eliminates
|
|
filesystem-specific variations (patch from Mike Lundy)
|
|
|
|
0.6 (28 April 2010)
|
|
|
|
- |:CommandT| now accepts an optional parameter to specify the starting
|
|
directory, temporarily overriding the usual default of Vim's |:pwd|
|
|
- fix truncated paths when operating from root directory
|
|
|
|
0.5.1 (11 April 2010)
|
|
|
|
- fix for Ruby 1.9 compatibility regression introduced in 0.5
|
|
- documentation enhancements, specifically targetted at Windows users
|
|
|
|
0.5 (3 April 2010)
|
|
|
|
- |:CommandTFlush| now re-evaluates settings, allowing changes made via |let|
|
|
to be picked up without having to restart Vim
|
|
- fix premature abort when scanning very deep directory hierarchies
|
|
- remove broken |<Esc>| key mapping on vt100 and xterm terminals
|
|
- provide settings for overriding default mappings
|
|
- minor performance optimization
|
|
|
|
0.4 (27 March 2010)
|
|
|
|
- add |g:CommandTMatchWindowAtTop| setting (patch from Zak Johnson)
|
|
- documentation fixes and enhancements
|
|
- internal refactoring and simplification
|
|
|
|
0.3 (24 March 2010)
|
|
|
|
- add |g:CommandTMaxHeight| setting for controlling the maximum height of the
|
|
match window (patch from Lucas de Vries)
|
|
- fix bug where |'list'| setting might be inappropriately set after dismissing
|
|
Command-T
|
|
- compatibility fix for different behaviour of "autoload" under Ruby 1.9.1
|
|
- avoid "highlight group not found" warning when run under a version of Vim
|
|
that does not have syntax highlighting support
|
|
- open in split when opening normally would fail due to |'hidden'| and
|
|
|'modified'| values
|
|
|
|
0.2 (23 March 2010)
|
|
|
|
- compatibility fixes for compilation under Ruby 1.9 series
|
|
- compatibility fixes for compilation under Ruby 1.8.5
|
|
- compatibility fixes for Windows and other non-UNIX platforms
|
|
- suppress "mapping already exists" message if <Leader>t mapping is already
|
|
defined when plug-in is loaded
|
|
- exclude paths based on |'wildignore'| setting rather than a hardcoded
|
|
regular expression
|
|
|
|
0.1 (22 March 2010)
|
|
|
|
- initial public release
|
|
|
|
------------------------------------------------------------------------------
|
|
vim:tw=78:ft=help:
|