Fasder

A descendant of Wei Dai's fasd project offering improved performance on Cygwin/Windows and (eventually) more...

Fasder (approximate pronunciation "faster") is a command-line productivity booster that offers quick jumping and naming access to files and directories at the POSIX shell prompt. Fasder tracks your command history to give priority access to the files you've accessed most often or most recently. Fasder is a direct descendant of fasd which was inspired by tools like autojump, z and v.

fasder (and its ancestor fasd) takes its name from the convenient default aliases that provide most of fasder's functionality:

• f(files),
• a(all, i.e. files/directories)
• s(show/search/select interactively)
• d(directories).

Fasder ranks files and directories by "frecency," that is, by a combination of "frequency" and "recency." The term "frecency" was coined by Mozilla and is used in Firefox (link).

Introduction

If you use your shell to navigate and launch applications, fasder can help you do so more efficiently. With fasder, you can open files from any directory. Fasder can find a "frecent" file or directory and open it with the command you specify. Here are some scenarios where the fasder command on the left will expand to the shell command on the right. Pretty magical, huh?

  v def conf       =>     vim /some/awkward/path/to/type/default.conf
  j abc            =>     cd /hell/of/a/awkward/path/to/get/to/abcdef
  m movie          =>     mplayer /whatever/whatever/whatever/awesome_movie.mp4
  o eng paper      =>     xdg-open /you/dont/remember/where/english_paper.pdf
  vim `f rc lo`    =>     vim /etc/rc.local
  vim `f rc conf`  =>     vim /etc/rc.conf

Fasder comes with some useful aliases by default:

alias a='fasder -a'        # any
alias s='fasder -si'       # show / search / select
alias d='fasder -d'        # directory
alias f='fasder -f'        # file
alias sd='fasder -sid'     # interactive directory selection
alias sf='fasder -sif'     # interactive file selection
alias z='fasder_cd -d'     # cd, same functionality as j in autojump
alias zz='fasder_cd -d -i' # cd with interactive selection

Fasder will smartly detect when to display a list of files or just the best match. For instance, when you call fasder in a subshell with some search parameters, fasder will only return the best match. This enables you to do:

mv update.html `d www`
cp `f mov` .

Install

Fasder is a self-contained POSIX shell script that can be either sourced or executed. A Makefile is provided to install fasder and fasder.1 to the desired places.

System-wide install:

make install

Install to $HOME:

PREFIX=$HOME make install

Or alternatively you can just copy fasder anywhere you like (preferably under some directory in your $PATH).

To get fasder working in a shell, some initialization code must be run. Put the following line in your shell rc:

eval "$(fasder --init auto)"

This will setup a command hook that executes on every command as well as advanced tab completion for zsh and bash.

If you want more control over what gets into your shell environment, you can pass a customized set of arguments to fasder --init.

zsh-hook             # define _fasder_preexec and add it to zsh preexec array
zsh-ccomp            # zsh command mode completion definitions
zsh-ccomp-install    # setup command mode completion for zsh
zsh-wcomp            # zsh word mode completion definitions
zsh-wcomp-install    # setup word mode completion for zsh
bash-hook            # add hook code to bash $PROMPT_COMMAND
bash-ccomp           # bash command mode completion definitions
bash-ccomp-install   # setup command mode completion for bash
posix-alias          # define aliases that applies to all posix shells
posix-hook           # setup $PS1 hook for shells that's posix compatible
tcsh-alias           # define aliases for tcsh
tcsh-hook            # setup tcsh precmd alias

Example for a minimal zsh setup (no tab completion):

eval "$(fasder --init posix-alias zsh-hook)"

Note that this method will slightly increase your shell start-up time, since calling binaries has overhead. You can cache fasder init code if you want minimal overhead. Example code for bash (to be put into .bashrc):

fasd_cache="$HOME/.fasder-init-bash"
if [ "$(command -v fasder)" -nt "$fasd_cache" -o ! -s "$fasd_cache" ]; then
  fasder --init posix-alias bash-hook bash-ccomp bash-ccomp-install >| "$fasd_cache"
fi
source "$fasd_cache"
unset fasd_cache

Optionally, you can also source fasder if you want fasder to be a shell function instead of an executable.

You can tweak the initialization code. For instance, if you want to use "c" instead of "z" to do directory jumping, you can use the alias below:

alias c='fasder_cd -d'
# `-d` option present for bash completion
# function fasder_cd is defined in posix-alias

After you install fasder, warm up the fasder database by opening some files (with any program) as you cd around your filesystem. Then try some of the examples below.

Examples

f foo           # list frecent files matching foo
a foo bar       # list frecent files and directories matching foo and bar
f js$           # list frecent files that ends in js
f -e vim foo    # run vim on the most frecent file matching foo
mplayer `f bar` # run mplayer on the most frecent file matching bar
z foo           # cd into the most frecent directory matching foo
open `sf pdf`   # interactively select a file matching pdf and launch `open`

You can add your own aliases to fully utilize the power of fasder. Here are some examples to get you started:

alias v='f -e vim' # quick opening files with vim
alias m='f -e mplayer' # quick opening files with mplayer
alias o='a -e xdg-open' # quick opening files with xdg-open

If you're using bash, you have to call _fasder_bash_hook_cmd_complete to make completion work. For instance:

_fasder_bash_hook_cmd_complete v m j o

You can select an entry in the list of matching files.

Matching

Fasder has three matching modes: default, case-insensitive, and fuzzy.

For a given set of queries (groups of command-line arguments passed to fasder), a pathname is a match if and only if:

1. Queries match the path in order.
2. The last query matches the last segment of the path.

If no match is found, fasder will try the same process ignoring case. Failing this, fasder will allow extra characters to be placed between query characters for fuzzy matching.

Tips:

• If you want your last query not to match the last segment of the path, append / to the last query.
• If you want your last query to match the end of the filename, append $ to the last query.

How It Works

When you initialize the fasder system (typically through your shell config scripts), fasder creates a hook function which will be executed after every shell command. The hook will scan your commands for file and directory names and add them to its database.

Compatibility

Fasder's basic functionalities are POSIX compliant, meaning that you should be able to use fasder in all POSIX compliant shells. Your shell needs to support command substitution in $PS1 in order to automatically track your commands and files. This feature is not specified by the POSIX standard, but is present in many POSIX-compliant shells. In shells without prompt command or prompt command substitution (csh for instance), you can add entries manually with fasder -A. You are most welcome to contribute shell initialization code for shells not yet supported.

Fasder has been tested on the following shells: bash, zsh, mksh, pdksh, dash, busybox ash, FreeBSD 9 /bin/sh and OpenBSD /bin/sh.

Fasder is written in bash-4.1; you need a bash interpreter at least as new. Basically, we've traded in the POSIX compatibility of fasd for Bash's greater range of built-in features to "squeeze" the script for faster performance.

Synopsis

fasder [options] [query ...]
[f|a|s|d|z] [options] [query ...]
  options:
    -s         list paths with scores
    -l         list paths without scores
    -i         interactive mode
    -e <cmd>   set command to execute on the result file
    -b <name>  only use <name> backend
    -B <name>  add additional backend <name>
    -a         match files and directories
    -d         match directories only
    -f         match files only
    -r         match by rank only
    -t         match by recent access only
    -R         reverse listing order
    -h         show a brief help message
    -[0-9]     select the nth entry

fasder [-A|-D] [paths ...]
    -A    add paths
    -D    delete paths

Tab Completion

Fasder offers two completion modes: command completion and word completion. Command completion works in bash and zsh. Word completion only works in zsh.

Command completion is just like completion for any other command. It is triggered when you hit tab on a fasd command or its aliases. Under this mode your queries can be separated by a space. Tip: if you find that the completion result overwrites your queries, type an extra space before you hit tab.

Word completion can be triggered on any command. Word completion is triggered by any command line argument that starts with , (all), f, (files), or d, (directories), or that ends with ,, (all), ,,f (files), or ,,d (directories). Examples:

$ vim ,rc,lo<Tab>
$ vim /etc/rc.local

$ mv index.html d,www<Tab>
$ mv index.html /var/www/

There are also three zle widgets: fasder-complete, fasder-complete-f, fasder-complete-d. You can bind them to any keybindings you like:

bindkey '^X^A' fasder-complete    # C-x C-a to do fasder-complete (files and directories)
bindkey '^X^F' fasder-complete-f  # C-x C-f to do fasder-complete-f (only files)
bindkey '^X^D' fasder-complete-d  # C-x C-d to do fasder-complete-d (only directories)

Backends

Fasder can take advantage of different sources of recent / frequent files. Most desktop environments (such as OS X and Gtk) and some editors (such as Vim) keep a list of accessed files. Fasder can use them as additional backends if the data can be converted into fasder's native format. Below is a list of available backends.

`spotlight`
OSX spotlight, provides entries that were changed today or opened within the
past month

`recently-used`
GTK's recently-used file (Usually available on Linux)

`current`
Provides everything in $PWD (wherever you are executing `fasder`)

`viminfo`
Vim's editing history, useful if you want to define an alias just for editing
things in vim

You can define your own backend by declaring a function by that name in your .fasdrc. You can set default backend with _FASD_BACKENDS variable in your .fasdrc.

Fasder can mimic v's behavior by this alias:

alias v='f -t -e vim -b viminfo'

Configuration settings

The following shell variables can be set before sourcing fasder. You can set them in $HOME/.fasdrc

$_FASD_DATA
Path to the fasder data file, default "$HOME/.fasd".

$_FASD_BLACKLIST
List of "blacklisted" strings. Commands matching them will not be processed.
Default is "--help".

$_FASD_SHIFT
List of all command names that need to be shifted; defaults to "sudo busybox".

$_FASD_IGNORE
List of all commands that will be ignored; defaults to "fasder ls echo".

$_FASD_TRACK_PWD
Fasder tracks your "$PWD" by default. Set this to 0 to disable this behavior.

$_FASD_AWK
The awk program to use. Fasder can detect and use a compatible awk.

$_FASD_SINK
Log file to capture the standard error; defaults to "/dev/null".

$_FASD_MAX
Maximum total score / weight; defaults to 2000.

$_FASD_SHELL
The shell to execute. Some shells will run faster than others. fasder
runs faster with dash and ksh variants.

$_FASD_BACKENDS
Default backends. (See the "backends" section above.)

$_FASD_RO
If set to any non-empty string, fasder will not add or delete entries from
the database. You can set and export this variable from the command line.

$_FASD_FUZZY
Level of "fuzziness" when doing fuzzy matching. More precisely, the number of
characters that can be skipped to generate a match. Set to empty or 0 to
disable fuzzy matching. Default value is 2.

$_FASD_VIMINFO
Path to .viminfo file for viminfo backend; defaults to "$HOME/.viminfo"

$_FASD_RECENTLY_USED_XBEL
Path to XDG recently-used.xbel file for recently-used backend, defaults to
"$HOME/.local/share/recently-used.xbel"

Debugging

If fasder does not work as expected, please file a bug report describing the unexpected behavior along with your OS version, shell version, awk version, sed version, and a log file.

You can set _FASD_SINK in your .fasdrc to obtain a log.

_FASD_SINK="$HOME/.fasd.log"

COPYING

Fasder is adapted from Wei Dai's fasd project. Fasd is based on code from z by rupa deadwyler under the WTFPL license. Most if not all of the code has been rewritten. Fasder is licensed under the "MIT/X11" license.

This document is adapted from the README for the original fasd project.

Please see also:

Thoughts about a faster "fasder"

TODO

• Discuss the optimization strategies employed.
• Discuss the remaining bottlenecks in the prompt hook: the call to mapfile, the "heaviness" of forking a process under Cygwin, and the extra burden when context-switching into a Cygwin shell (especially the DLLs) from the outside.
• Optimization work thus far has focused on the prompt hook and the --add and --query options. For the other options, there is still plenty of low-hanging fruit.
• There are many style issues: Comments and variable names can be made better, for example. Sub-functions can be extracted. Low-hanging fruit here too.
• Implement a --clean option to delete duplicate entries and entries no longer present in the filesystem
• Inventory the open issues and PRs of the original fasd project for anything we might wish to address here.
• Develop a more thorough, systematic testing process. There are two classes of use cases: (1) the hidden action of the _fasder_prompt_func (the prompt hook) and (2) the explicit invocation of fasder or its short aliases. Regarding the former, the script's dependency on a PROMPT_COMMAND whose exact behavior changes whenever we edit the script makes testing challenging; the situation is akin to the observer effect familiar to physicists. As a workaround, we could structure our test rig to invoke "fasder --proc " directly for each test case while we monitor the effects this has on a suitably mocked-up data file. OTOH, the "type 2" use cases wherein fasder is explicitly called should be easier to test.
• The original fasd contains notes about downloading it through package managers or the project website. Consider pursuing either or both of these for fasder.