See #97 and recently #130 too.

This allows the user to pass in options such as "--ignore '*.pyc'" to not list any files ending in '.pyc' in the output. It uses the Rust glob crate and currently does a simple split on pipe, without any escaping, so it’s not really *complete*, but is at least something.

This has to be done for when ignore patterns get introduced and have to be stored in a Vec.

Fixes #123. The code assumes that every File that has its link_target() method called would first have been checked to make sure it’s actually a link first. Unfortunately it also assumed that the only thing that can go wrong while following a link is if the file wasn’t a link, meaning it crashes when given a link it doesn’t have permission to follow.

This makes the file_target() method able to return either a file or path for displaying, as before, but also an IO error for when things go wrong.

Fixes #129

This changes the way that views are used to display the actual lists of files. It used to pass empty vectors to the view methods, which most of the time would not print anything because there are no files to list — except when there’s a header row which gets printed for no files.

By not calling the view method at all when there’s nothing to print, exa won’t ever print extra things in the view unless it needs to for a file.

This fixes #106 “Don’t print the header if the result set is empty”

See the README section for more details. Basically, with this way, we can store a bunch of existing valid exa outputs, change a VM's environment to match our values, then check that exa still works by comparing outputs.

Add legal values to error messages

Change to put error messages into stderr

Change to ignore broken pipe error

Now when you do `--sort time` instead of saying "unknown option --sort
time" it will say "unknown options '--sort time' (choices: name...)"
with all legal options.

This also adds the legal values to the default help text.

Exa now recognizes pipes, devices, and sockets on unix systems

See https://github.com/rust-lang/rfcs/pull/1177

Fix building without default features

The tests, as I so clearly wrote a comment to indicate, can only be run after the script to generate the testcases has been run...

This commit removes the 'main' function present in main.rs, renames it to exa.rs, and puts the 'main' function in its own binary. This, I think, makes it more clear how the program works and where the main entry point is.

Librarification also means that we can start testing as a whole. Two tests have been added that test everything, passing in raw command-line arguments then comparing against the binary coloured text that gets produced.

Casualties include having to specifically mark some code blocks in documentation as 'tests', as rustdoc kept on trying to execute my ANSI art.

This will mean that we can test exa's output as a whole, without having to rely on process or IO or anything like that.

The original options was becoming a bit unwieldy, and would have been even more so if I added the same amount of comments. So this commit splits it up.

There's no extra hiding going on here, or rearranging things within the module: (almost) everything now has to be marked 'pub' to let other sub-modules in the new options module to see it.

The trait was only used internally to the options module, so it doesn't actually need to be exist or implemented on anything! We can just impl them directly on the types and have those methods be local to the module.

They were the same as the archive/compressed extensions, and I never noticed!

Currently these routines number two: file type checking based on a file's name, and source file checking, also based on the file's name.

This commit moves file, dir, and the feature modules into one parent 'fs' module. Now there are three main 'areas' of the code: main and options, the filesystem-touching code, and the output-displaying code.

It should be the case that nothing in 'output' touches 'std::fs'.