Add use declaration re-ordering

[studoot]

This partially addresses #298, adding sorting of contiguous runs of use declarations depending on the value and type of path referenced. As an example (taken from the added integration test files), the following code

use std::str;
use std::cmp::{d, c, b, a};
use std::ddd::aaa;
use std::ddd::{d as p, c as g, b, a};
// This comment should stay with `use std::ddd:bbb;`
use std::ddd::bbb;

gets transformed to

use std::cmp::{a, b, c, d};
use std::ddd::{a, b, c as g, d as p};
use std::ddd::aaa;
// This comment should stay with `use std::ddd:bbb;`
use std::ddd::bbb;
use std::str;

The reorder_imports config item has changed from a boolean to {None, Lines, Items, LinesAndItems}, to allow combinations of use declaration lines and use declaration items to be selected.

I would have to say that this isn't a finished item - there are several areas that I'm not 100% happy with:

• Any 'missing spans' before the run of use declarations remain at the start of the run, even after the declarations are reordered. I can see situations where a comment is expected to stay attached to the first use in a run. I'm thinking situations like:

  // This is a file header comment
  
  // We need to use the filesystem
  use std::fs;
  use std::cmp;

  which will currently become

  // This is a file header comment
  
  // We need to use the filesystem
  use std::cmp;
  use std::fs;

• Similarly, any end of line comment after a use will stay with the code after that use:

  use std::fs; // We need to use the filesystem
  use std::cmp;

  which will currently become

  // We need to use the filesystem
  use std::cmp;
  use std::fs;

• Contents of import lists are currently ignored when comparing imports, meaning that use a::b::{c, d}; is considered equal to use a::b::{e, f}; (but that is relatively easy to implement).

• Comparison functions for AST items such as paths etc are clustered at the top of visitor.rs. Is there either a) a preferable place to put them, or b) a preferable way to implement them?!

• Looking at the code again, some refactoring would be beneficial so that the use declarations in functions were processed the same as those in modules (that would mean changing the code around the call to visit_item in visit_stmt)

Anyway - it's a start...

[nrc]

Also todo for later is that I would expect imports to be squished together, e.g.,

use std::ddd::aaa;
use std::ddd::{d as p, c as g, b, a};
// This comment should stay with `use std::ddd:bbb;`
use std::ddd::bbb;

becomes

use std::ddd::{aaa, a, b, c as g, d as p};
// This comment should stay with `use std::ddd:bbb;`
use std::ddd::bbb;

Note the second import stays separate due to the comment.

[nrc]

Contents of import lists are currently ignored when comparing imports, meaning that use a::b::{c, d}; is considered equal to use a::b::{e, f}; (but that is relatively easy to implement).

Could you file an issue for this and add a FIXME(#nn) to the code please?

[nrc]

Comparison functions for AST items such as paths etc are clustered at the top of visitor.rs. Is there either a) a preferable place to put them, or b) a preferable way to implement them?!

I would put them in imports.rs, I guess the more generic stuff (e.g., Paths) should be somewhere else, but we can deal with that if they become more widely used.

[nrc]

I would prefer to add another option, rather than have a single enum like this - this way lies combinatorial explosion!

[studoot]

Absolutely - I had a look to see if there was a way that a list of enumerations could be supplied rather than just one (something like reorder_imports = [ "Lines", "Imports" ] or reorder_imports = "Lines,Imports"), but that was looking a bit involved, so I'll change to separate options.

[nrc]

This looks great, thanks for the PR! A few very minor comments to address.

[studoot]

• Sorting on import item lists was simple enough, so I've added that to this PR...
• Moved format_imports and format_import to imports.rs
• Added comment to walk_mod_items
• Changed reorder_imports options to two separate boolean options

[nrc]

Thank you!