PathList
This is undergoing a major API redesign and when i'm done it'll be 1.0.0, please pin the exact version 0.whatever number in your gemfile
This started as a way to quickly and natively ruby-ly parse gitignore files and find matching files. It's now gained an equivalent includes file functionality, ARGV awareness, and some shebang matching, while still being extremely fast, to be a one-stop file-list for your linter.
Filter a directory tree using a .gitignore file. Recognises all of the gitignore rules
PathList.new.sort == `git ls-files`.split("\n").sort
Features
- Fast (faster than using
`git ls-files`.split("\n")
for small repos (because it avoids the overhead of``
)) - Supports ruby 2.4-3.0.0.preview1 & jruby
- supports all gitignore rule patterns
- doesn't require git to be installed
- supports a gitignore-esque "include" patterns. (
include_rules:
/include_files:
) - supports an expansion of include patterns, expanding and anchoring paths (
argv_rules:
) - supports matching by shebang rather than filename for extensionless files:
#!:
- reads .gitignore in all subdirectories
- reads .git/info/excludes
- reads the global gitignore file mentioned in your git config
Installation
Add this line to your application's Gemfile:
gem 'path_list'
And then execute:
$ bundle
Or install it yourself as:
$ gem install path_list
Usage
PathList.new.each { |file| puts "#{file} is not ignored by the .gitignore file" }
#each
, #map
etc
This yields paths that are not ignored by the gitignore, i.e. the paths that would be returned by git ls-files
.
A PathList instance is an Enumerable and responds to all Enumerable methods:
PathList.new.to_a
PathList.new.map { |file| file.upcase }
Like other enumerables, PathList#each
can return an enumerator:
PathList.new.each.with_index { |file, index| puts "#{file}#{index}" }
Warning: Do not change directory (e.g. Dir.chdir
) in the block.
#allowed?
To check if a single path is allowed, use
PathList.new.allowed?('relative/path')
PathList.new.allowed?('./relative/path')
PathList.new.allowed?('/absolute/path')
PathList.new.allowed?('~/home/path')
Relative paths will be considered relative to the root:
directory, not the current directory.
This is aliased as ===
so you can use a PathList instance in case statements.
case my_path
when PathList.new
puts(my_path)
end
It's recommended to save the PathList instance to a variable to avoid having to read and parse the gitignore file and gitconfig files repeatedly.
See Optimising allowed for ways to make this even faster
Note: A file must exist at that path and not be a directory for it to be considered allowed.
Essentially it can be thought of as `git ls-files`.include?(path)
but much faster.
This excludes all directories and all possible path names that don't exist.
root:
Default: Dir.pwd ($PWD, the current working directory)
This directory is used for:
- the location of
.git/core/exclude
- the ancestor of all non-global automatically loaded
.gitignore
files - the root directory for array rules (
ignore_rules:
,include_rules:
,argv_rules:
) containing/
- the path that yielded paths are relative to
- the ancestor of all paths yielded by
#each
- the path that
#allowed?
considers relative paths relative to - the ancestor of all
include_files:
andignore_files:
To use a different directory:
PathList.new(root: '/absolute/path/to/root').to_a
PathList.new(root: '../relative/path/to/root').to_a
A relative root will be found relative to the current working directory when the PathList instance is initialized, and that will be the last time the current working directory is relevant.
Note: Changes to the current working directory (e.g. with Dir.chdir
), after initialising a PathList instance, will not affect the PathList instance. root:
will always be what it was when the instance was initialized, even as a default value.
gitignore:
Default: true
When gitignore: true
: the .gitignore file in the root:
directory is loaded, plus any .gitignore files in its subdirectories, the global git ignore file as described in git config, and .git/info/exclude. .git
directories are also excluded to match the behaviour of git ls-files
.
When gitignore: false
: no ignore files or git config files are automatically read, and .git
will not be automatically excluded.
PathList.new(gitignore: false).to_a
ignore_files:
This is a list of files in the gitignore format to parse and match paths against, not a list of files to ignore If you want an array of files use ignore_rules:
Additional gitignore-style files, either as a path or an array of paths.
You can specify other gitignore-style files to ignore as well.
Missing files will raise an Errno::ENOENT
error.
Relative paths are relative to the root:
directory.
Absolute paths also need to be within the root:
directory.
PathList.new(ignore_files: 'relative/path/to/my/ignore/file').to_a
PathList.new(ignore_files: ['/absolute/path/to/my/ignore/file', '/and/another']).to_a
Note: the location of the files will affect rules beginning with or containing /
.
To avoid raising Errno::ENOENT
when the file doesn't exist:
PathList.new(ignore_files: ['/ignore/file'].select { |f| File.exist?(f) }).to_a
ignore_rules:
This can be a string, or an array of strings, and multiline strings can be used with one rule per line.
PathList.new(ignore_rules: '.DS_Store').to_a
PathList.new(ignore_rules: ['.git', '.gitkeep']).to_a
PathList.new(ignore_rules: ".git\n.gitkeep").to_a
These rules use the root:
argument to resolve rules containing /
.
include_files:
This is an array of files in the gitignore format to parse and match paths against, not a list of files to include. If you want an array of files use include_rules:
.
Building on the gitignore format, PathList also accepts rules to include matching paths (rather than ignoring them). A rule matching a directory will include all descendants of that directory.
These rules can be provided in files either as absolute or relative paths, or an array of paths.
Relative paths are relative to the root:
directory.
Absolute paths also need to be within the root:
directory.
PathList.new(include_files: 'my_include_file').to_a
PathList.new(include_files: ['/absolute/include/file', './relative/include/file']).to_a
Missing files will raise an Errno::ENOENT
error.
To avoid raising Errno::ENOENT
when the file doesn't exist:
PathList.new(include_files: ['include/file'].select { |f| File.exist?(f) }).to_a
Note: All paths checked must not be excluded by any ignore files AND each included by include file separately AND the include_rules:
AND the argv_rules:
. see Combinations for solutions to using OR.
include_rules:
Building on the gitignore format, PathList also accepts rules to include matching paths (rather than ignoring them). A rule matching a directory will include all descendants of that directory.
This can be a string, or an array of strings, and multiline strings can be used with one rule per line.
PathList.new(include_rules: %w{my*rule /and/another !rule}, gitignore: false).to_a
Rules use the root:
argument to resolve rules containing /
.
Note: All paths checked must not be excluded by any ignore files AND each included by include file separately AND the include_rules:
AND the argv_rules:
. see Combinations for solutions to using OR.
argv_rules:
This is like include_rules:
with additional features meant for dealing with humans and ARGV
values.
It expands rules that are absolute paths, and paths beginning with ~
, ../
and ./
(with and without !
).
This means rules beginning with /
are absolute. Not relative to root:
.
Additionally it assumes all rules are relative to the root:
directory (after resolving absolute paths) unless they begin with *
(or !*
).
This can be a string, or an array of strings, and multiline strings can be used with one rule per line.
PathList.new(argv_rules: ['./a/pasted/path', '/or/a/path/from/stdin', 'an/argument', '*.txt']).to_a
Warning: it will not expand e.g. /../
in the middle of a rule that doesn't begin with any of ~
,../
,./
,/
.
Note: All paths checked must not be excluded by any ignore files AND each included by include file separately AND the include_rules:
AND the argv_rules:
. see Combinations for solutions to using OR.
shebang rules
Sometimes you need to match files by their shebang/hashbang/etc rather than their path or filename
Rules beginning with #!:
will match whole words in the shebang line of extensionless files.
e.g.
#!:ruby
will match shebang lines: #!/usr/bin/env ruby
or #!/usr/bin/ruby
or #!/usr/bin/ruby -w
e.g.
#!:bin/ruby
will match #!/bin/ruby
or #!/usr/bin/ruby
or #!/usr/bin/ruby -w
Only exact substring matches are available, There's no special handling of * or / or etc.
These rules can be supplied any way regular rules are, whether in a .gitignore file or files mentioned in include_files:
or ignore_files:
or include_rules:
or ignore_rules:
or argv_rules:
PathList.new(include_rules: ['*.rb', '#!:ruby']).to_a
PathList.new(ignore_rules: ['*.sh', '#!:sh', '#!:bash', '#!:zsh']).to_a
Note: git considers rules like this as a comment and will ignore them.
Combinations
In the simplest case a file must be allowed by each ignore file, each include file, and each array of rules. That is, they are combined using AND
.
To combine files using OR
, that is, a file may be matched by either file it doesn't have to be referred to in both:
provide the files as strings to include_rules:
or ignore_rules:
PathList.new(include_rules: [File.read('/my/path'), File.read('/another/path')])).to_a
This does unfortunately lose the file path as the root for rules containing /
.
If that's important, combine the files in the file system and use include_files:
or ignore_files:
as normal.
To use the additional ARGV
handling of argv_rules:
on a file, read the file into the array.
PathList.new(argv_rules: ["my/rule", File.read('/my/path')]).to_a
This does unfortunately lose the file path as the root /
and there is no workaround except setting the root:
for the whole PathList instance.
optimising #allowed?
To avoid unnecessary calls to the filesystem, if your code already knows whether or not it's a directory, or if you're checking shebangs and you have already read the content of the file: use
PathList.new.allowed?('relative/path', directory: false, content: "#!/usr/bin/ruby\n\nputs 'ok'\n")
This is not required, and if PathList does have to go to the filesystem for this information it's well optimised to only read what is necessary.
Limitations
-
Doesn't know what to do if you change the current working directory inside the
PathList#each
block. So don't do that.(It does handle changing the current working directory between
PathList#allowed?
calls) -
PathList always matches patterns case-insensitively. (git varies by filesystem).
-
PathList always outputs paths as literal UTF-8 characters. (git depends on your core.quotepath setting but by default outputs non ascii paths with octal escapes surrounded by quotes).
-
Because git looks at its own index objects and PathList looks at the file system there may be some differences between PathList and
git ls-files
. To avoid these differences you may want to use thegit_ls
gem instead- Tracked files that were committed before the matching ignore rule was committed will be returned by
git ls-files
, but not by PathList. - Untracked files will be returned by PathList, but not by
git ls-files
- Deleted files whose deletions haven't been committed will be returned by
git ls-files
, but not by PathList - On a case insensitive file system, with files that differ only by case,
git ls-files
will include all case variations, while PathList will only include whichever variation git placed in the file system. - PathList is unaware of submodules and just treats them like regular directories. For example:
git ls-files --recurse-submodules
won't use the parent repo's gitignore on a submodule, while PathList doesn't know it's a submodule and will. - PathList will only return the files actually on the file system when using
git sparse-checkout
.
- Tracked files that were committed before the matching ignore rule was committed will be returned by
Contributing
Bug reports and pull requests are welcome on GitHub at https://github.com/robotdana/path_list.
Some tools that may help:
-
bin/setup
: install development dependencies -
bundle exec rspec
: run all tests -
bundle exec rake
: run all tests and linters -
bin/console
: open apry
console with everything required for experimenting -
bin/ls [argv_rules]
: the equivalent ofgit ls-files
-
bin/prof/ls [argv_rules]
: ruby-prof report forbin/ls
-
bin/prof/parse [argv_rules]
: ruby-prof report for parsing root and global gitignore files and any arguments. -
bin/time [argv_rules]
: the average time for 30 runs ofbin/ls
This repo is too small to stress bin/time more than 0.01s, switch to a large repo and find the average time before and after changes. -
bin/compare
: compare the speed and output of PathList andgit ls-files
. (suppressing differences that are because of known limitations)
License
The gem is available as open source under the terms of the MIT License.