Release 0.13.3

[github-actions]

• #274: Include a dependency graph in the generated documentation
• #275: Revert Include benches/ again #271 and exclude benches/ again
• #276: Extend display_width to handle emojis when unicode-width is disabled
• #277: Collect coverage data with and without default features
• #279: Make indent preserve existing newlines in the input string
• #281: Ensure all Options fields have examples
• #282: Add a wrap_columns function
• #283: Rename width to width_or_options
• #284: Fix typo in fill_inplace docstring
• #285: Make LineNumbers::get generic over penalty type
• #286: Test handling of whitespace-only lines
• #287: Ignore left-over merge conflict files
• #288: Run files through rustfmt to fix formatting
• #291: Setup weekly build to catch errors on nightly Rust
• #292: Forbid warnings when building in CI
• #293: Remove unused [badges] section from Cargo.toml
• #294: Add new unfill and refill functions

[ma2bd]

Thanks @mgeisler for supporting the library and for publishing this release. It turns out #279 was breaking for us. The fix was easy but it could have been a bit messy given that the default behavior of cargo allows to pick any package version with a higher or equal minor number. May I suggest using major versions (e.g. 0.14.0) for semantic changes in the future?

[mgeisler]

Hi @MatBD, thanks a bunch for letting me know!

I'm sorry that this broke you and I'm glad you could fix up the tests so quickly. I considered the change in #279 a bugfix since indent wasn't supposed to change the number of newlines in the input string.

You're right that Cargo considers 0.x.y and 0.x.(y+1) to be compatible and the plan is definitely to follow semantic versioning. I guess you could say that this case was bad because indent is such a simple function, so the change in behavior affected not just a few corner cases... I'll try to take this into consideration in the future!