Add assert_str_eq macro for comparing "raw" strings

[x3ro]

As discussed in #24, pretty_assertion's assert_eq will escape line
breaks in multi-line strings, making diffs for them much harder to
read.

This commit introduces another variant, assert_str_eq, and the
associated StrComparison, which compares the raw strings, where
the newlines are not escaped.

Fixes #24

[tommilligan]

Hi @x3ro - thanks for your contribution, and your well thought out discussion in #24 as well.

I actually really like your proposed solution. I think previously I've felt that proposed solutions (both others and my own) have been too generic/complex and would need to be split out to separate crates. This feels actually like you're simplifying the functionality - we're literally skipping a step (where we turn whatever type into &str before we diff it), and it totally feels natural to be able to do that.

I don't have any inline feedback on your implementation as is, it looks like it does the job for &str types. My general feedback would be:

• Docstrings need filling out still
• We can make this solution general for all types that implement AsRef<str>. This will allow users with custom string types to still use the diffing if they wish

I've pushed a commit that makes suggestions for both these points. Please take a look and let me know if you want to discuss anything further - otherwise I'm happy to get this merged in and released.

[x3ro]

Hey @tommilligan :) Thanks for the feedback and great suggestions. I had also played around with AsRef<str> during the implementation, but failed because I tried to use it in the constructor, where I then ran into issues with lifetime. I really like your solution.

The only potential issue I see is that one can instantiate StrComparison with arbitrary types and the compilation will only fail once trying to Display it, e.g.

let c = StrComparison::new(&6, &6); # works fine

# breaks
println!("{}", c);
               ^ the trait `AsRef<str>` is not implemented for `{integer}`

It's hard for me to judge whether this would actually be a problem, since this would only impact people who are using StrComparison directly, without the macros. In order to avoid this, we could apply the trait bound for StrComparison more generally:

pub struct StrComparison<'a, TLeft, TRight>
where
    TLeft: AsRef<str> + ?Sized,
    TRight: AsRef<str> + ?Sized,
{
    left: &'a TLeft,
    right: &'a TRight,
}

impl<'a, TLeft, TRight> StrComparison<'a, TLeft, TRight>
where
    TLeft: AsRef<str> + ?Sized,
    TRight: AsRef<str> + ?Sized,
{
    /// Store two values to be compared in future.
    ///
    /// Expensive diffing is deferred until calling `Debug::fmt`.
    pub fn new(left: &'a TLeft, right: &'a TRight) -> StrComparison<'a, TLeft, TRight> {
        StrComparison { left, right }
    }
}

Do you see any problems with that approach?

[tommilligan]

The only potential issue I see is that one can instantiate StrComparison with arbitrary types and the compilation will only fail once trying to Display it, e.g.

Ah good spot. Yes, I think adding the bound to the constructor is sensible, if for some reason we want to relax it later we can make it more permissive.

If you'd like to add that and then squash everything together, I think we're good to go

[x3ro]

Done!

[codecov-commenter]

Codecov Report

Merging #88 (e244a90) into main (88cd1be) will decrease coverage by 0.52%.
The diff coverage is 91.66%.

❗ Current head e244a90 differs from pull request most recent head 5271543. Consider uploading reports for the commit 5271543 to get more accurate results

@@            Coverage Diff             @@
##             main      #88      +/-   ##
==========================================
- Coverage   96.35%   95.83%   -0.53%     
==========================================
  Files           3        3              
  Lines         192      216      +24     
==========================================
+ Hits          185      207      +22     
- Misses          7        9       +2     

Impacted Files	Coverage Δ
pretty_assertions/tests/macros.rs	95.71% <90.00%> (-2.29%)	⬇️
pretty_assertions/src/lib.rs	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 88cd1be...5271543. Read the comment docs.