===================== Add user key bindings support. Recall $fisher_home/functions are always before user functions in $fish_function_path. This was an early design decision in order to prevent users from redefining core functions by mistake or by means other than using plugins (recommended). In other words, you are free to create a plugin that modifies a Fisherman core function, but you can't redefine a Fisherman function privately by saving it to your user config fish. If you found a bug in a Fisherman function, instead of creating a private patch send it upstream. If you created a function that overrides a Fisherman core feature, create a plugin. This way the community can benefit from your code whenever you are ready to publish it. By default, Fisherman provides no fish_user_key_bindings, so if the user has already defined their own fish_user_key_bindings that one will not be affected. Now, plugins can define their own key bindings inside a fish_user_key_bindings.fish or key_bindings.fish at the root of their repository or inside a functions directory. You can put your key bindings inside a function or not. If you put it inside a function, the function name must be the same as the file without the .fish extension. $fisher_config/bindings.fish When a plugin with key bindings is installed for the first time or the only one with bindings is uninstalled, Fisherman will modify ~/.config/functions/fish_user_key_bindings.fish (or create it for the first time) and add a single line at the top of the fish_user_key_bindings function to source the new $fisher_config/bindings.fish. All the key bindings defined by the enabled/installed plugins are concatenated and saved to this file. This mechanism has the following advantages: Does not slow down shell start. Does not require Fisherman to provide his own fish_user_key_bindings by default. Honors any previously existing user key bindings. Allows plugin to define their own key bindings and coexist with the user's key bindings. If the user updates his fish_user_key_bindings, re-running the function does update the key bindings. Mega Refactoring The entire source code of Fisherman received a major revision and refactoring. The validation and install/uninstall mechanisms were thoroughly revised and and broken down into smaller functions easier to test as well as several other sub parts of the system. Rewrite fisher search and remove features that are mostly already covered by fisher --list and remove the ability to generate information about plugins of unknown origin. The decision to remove this feature was based in performance concerns and the result of thinking about the usability and whether it was really worth the speed tradeoff. The conclusion is I would rather have better performance and if I need to query a plugins origin I can always use fisher --list or fisher --list=url or fisher --list=author. Add $fisher_update_interval to determine if the index should update or not when a search query is taking place. The default value is 10 seconds. This means the index will not be updated if less than 10 seconds have elapsed since the last action that triggered an update in the first place. See Improve Install/Uninstall/Update status output. If a plugin fails to install decrease the total. If any plugins are skipped because they are already installed in the case of fisher install or available in the cache, but disabled in the case of fisher uninstall they are collected into an array and displayed in a new section n plugin/s skipped (a, b, c) at the bottom of the report. Improve test coverage. Tightly coupled functions were making testing increasingly difficult. Most of the test effort was basically testing whether git clone or git pull. New separation of concerns makes tests run faster and the difficult install/uninstall algorithms has better coverage now. Other Now __fisher_list can list plugins from the cache, a fishfile/bundle and plugins that are installed/enabled or disabled. This removes __fisher_file and combines it with __fisher_list. This also removes fisher -f and replaces it with fisher -l <file> or fisher --list=<file>. Rename __fisher_parse_help to __fisher_complete and have the function create the completions automatically. This allows you to complete your commands with parseable usage help faster. The original design was fine, but this change improves auto-complete performance so it was preferred. Use __fisher_index_update when building file with Make. This helps prevent an error when using a fish version < 2.2.0. See #55 #50 #48. Add __fisher_index_update to update the index and remove previously undocumented fisher update --index. This function is designed to bypass GitHub's server network cache passing an arbitrary query string to curl like $fisher_index?RANDOM_NUMBER. This means index updates are immediately available now. Add fisher --list=url option to display local plugin url or path. Add fisher --list=bare option to display local plugins in the cache without the * enabled symbol. Prepend > to the currently enabled theme when using fisher --list[=cache]. Related #49. Prepend * to plugin names to indicate they are currently enabled when using fisher --list[=cache]. See #49.
9.3 KiB
fisher-tour(7) -- Fisherman Tour
DESCRIPTION
Fisherman is a plugin manager and CLI toolkit for Fish to help you build powerful utilities and share your code easily.
Fisherman uses a flat tree structure that adds no cruft to your shell, making it as fast as no Fisherman. The cache mechanism lets you query the index offline and enable or disable plugins as you wish.
Other features include dependency management, great plugin search capabilities and full compatibility with Tackle, Wahoo and oh-my-fish themes and packages.
This document describes Fisherman features and their implementation details. For usage and command help see fisher(1)
.
FLAT TREE
The configuration directory structure is optimized to help your shell start new sessions as quickly as possible, regardless of the numbers of plugins or prompts enabled at any given time. An old saying goes that Fisherman is as fast as no Fisherman.
To explain how this is possible, we need to make a digression and discuss function scope first. In fish, all functions share the same scope and you can use only one name per function.
In the following example:
function foo
echo $_
function bar
end
end
function bar
echo $_
end
foo and bar are available immediately at the command line prompt and both print their names. But there is a catch, calling foo at least once will create a new bar function, effectively erasing the previous bar definition. Subsequent calls to bar will print nothing.
By convention, functions that start with any number of underscores are intentionally private, but there is no mechanism that prevents you from calling them at any time once loaded.
With this in mind, it's possible to improve the slow shell start problem using a flat tree structure whose path is loaded only once.
The overhead of juggling multiple path hierarchies in a per-plugin basis yields no benefits as everything is shared in the same scope.
Loading a path simply means adding the desired location to the $fish_function_path
array. See also functions(1)
.
Here is a snapshot of a typical configuration path with a single plugin and prompt:
$fisher_config
|-- cache/
|-- conf.d/
|-- |-- my_plugin.config.fish
|-- functions/
| |-- my_plugin.fish
| |-- fish_prompt.fish
| |-- fish_right_prompt.fish
|-- completions/
| |-- my_plugin.fish
|-- man/
|-- man1/
|-- my_plugin.1
If you are already familiar in the way fish handles your user configuration, you will find the above structure similar to $XDG_CONFIG_HOME/fish
. See Initialization Files
in help fish
to learn more about fish configuration.
conf.d
, short for configuration directory, is used for initialization files, i.e., files that should run at the start of the shell. Files that follow the naming convention <name>.config.fish
are added there.
PLUGINS
Plugins are components that extend and add features to your shell. To see what plugins are available use fisher search
. You can also type fisher install
and hit tab once to get formatted plugin information. The same works for fisher update
and fisher uninstall
.
To learn how to create plugins, see fisher help plugins
.
You can install a plugin by their name, URL or path to a local project.
If you use a name, this must be listed in the index database. See Index
.
fisher install shark
You can use an URL too if you have one.
fisher install simnalamburt/shellder
If the domain or host is not provided, Fisherman will use https://github.com
by default.
In addition, all of the following owner/repo
variations are accepted:
- owner/repo
>
https://github.com/owner/repo - github/owner/repo
>
https://github.com/owner/repo - gh/owner/repo
>
https://github.com/owner/repo
Shortcuts to other common Git repository hosting services are also available:
- bb/owner/repo
>
https://bitbucket.org/owner/repo - gl/owner/repo
>
https://gitlab.com/owner/repo - omf/owner/repo
>
https://github.com/oh-my-fish/repo
Because of Fisherman's flat tree model, there is no technical distinction between plugins or prompts. Installing a prompt is equivalent to switching themes in other systems. The interface is always install, update or uninstall.
Throughout this document and other Fisherman manuals you will find the term prompt when referring to the concept of a theme, i.e., a plugin that defines a fish_prompt
and / or fish_right_prompt
functions.
INDEX
You can install, update and uninstall plugins by name, querying the Fisherman index, or by URL using several of the variations described in Plugins
. The index is a plain text flat database independent from Fisherman. You can use a custom index file by setting $fisher_index
to your own file or URL. Redirection urls are not supported due to security and performance concerns. See fisher help config
.
A copy of the index is downloaded each time a query happens. This keeps the index up to date and allows you to search the database offline.
The index is a list of records, each consisting of the following fields:
name
,url
,info
,author
and one or moretags
.
Fields are separated by a new line '\n'
. Tags are separated by one space. Here is a sample record:
shark
https://github.com/bucaran/shark
Sparklines for your Fish
graph spark data
bucaran
To submit a new plugin for registration install the submit
plugin:
fisher install submit
For usage see the bundled documentation fisher help submit
.
You can also submit a new plugin manually and create a pull request.
git clone https://github.com/fisherman/fisher-index
cd index
echo "$name\n$URL\n$info\n$author\n$tags\n\n" >> index
git push origin master
open http://github.com
Now you can create a new pull request in the upstream repository.
CACHE
Downloaded plugins are tracked as Git repositories under $fisher_cache
. See fisher help config
to find out about other Fisherman configuration variables.
When you install or uninstall a plugin, Fisherman downloads the repository to the cache and copies only the relevant files from the cache to the loaded function and / or completion path. In addition, man pages are added to the corresponding man directory and if a Makefile is detected, the command make
is run.
The cache also provides a location for a local copy of the Index.
FISHFILES
Dependency manifest file, or fishfiles for short, let you share plugin configurations across multiple installations, allow plugins to declare dependencies, and prevent information loss in case of system failure. See fisher help fishfile
.
Here is an example fishfile inside $fisher_config
:
# my plugins
gitio
fishtape
# my links
github/bucaran/shark
The fishfile updates as you install / uninstall plugins. See also fisher help install
or fisher help uninstall
.
Plugins may list any number of dependencies to other plugins in a fishfile at the root of each project. By default, when Fisherman installs a plugin, it will also fetch and install its dependencies. If a dependency is already installed, it will not be updated as this could potentially break other plugins using an older version. For the same reasons, uninstalling a plugin does not remove its dependencies. See fisher help update
.
CONFIGURATION
Fisherman allows a high level of configuration using $fisher_*
variables. You can customize the home and configuration directories, debug log file, cache location, index source URL, command aliases, etc. See fisher help config
.
You can also extend Fisherman by adding new commands and ship them as plugins as well. Fisherman automatically adds completions to commands based in the function description and usage help if provided. See fisher help help
and fisher help commands
.
To add completions to standalone utility plugins, use complete(1)
.
CLI
If you are already familiar with other UNIX tools, you'll find Fisherman commands behave intuitively.
Most commands read the standard input by default when no options are given and produce easy to parse output, making Fisherman commands ideal for plumbing and building upon each other.
Fisherman also ships with a CLI options parser and a background job wait spinner that you can use to implement your own commands CLI. See getopts(1)
and wait(1)
.
COMPATIBILITY
Fisherman supports oh-my-fish (Wahoo) themes and plugins by default, but some features are turned off due to performance considerations.
oh-my-fish evaluates every .fish file inside the root directory of every plugin during initialization. This is necessary in order to register any existing init
events and invoke them using fish emit(1)
.
Since it is not possible to determine whether a file defines an initialization event without evaluating its contents first, oh-my-fish sources all *.fish
files and then emits events for each plugin.
Not all plugins opt in the initialization mechanism, therefore support for this behavior is turned off by default. If you would like Fisherman to behave like oh-my-fish at the start of every session, install the omf
compatibility plugin.
fisher install omf
This plugin also adds definitions for some of oh-my-fish Core Library functions.
SEE ALSO
fisher(1)
fisher help
fisher help config
fisher help plugins
fisher help commands
wait(1)
getopts(1)