Add new tests for recent changes

[andreivasiliu]

Merging #290 as-is feels a bit scary, since some of the more subtle changes that were committed since then might be reverted.

I'd like to write some additional tests to make sure they won't be.

I went through the commit log, and these are all the changes that have happened since I made the original PR (2020-06):

These were already tested before this issue:

• Inline CData events when pretty printing
  • Only concerns the writer.
• $value rename should work the same for serialization too
• Parse & in cdata correctly
• Add serde tests, implement newtype and tuple serializations, fix some bugs
• Always capture "$value" key when text node is appeared in the map
• Add support for i128/u128
  • Unlikely to be related; tests are likely difficult, since i128/u128 are not always available
• Deserialize raw bytes
• Serialize basic type as attributes when using serde
• use write_all instead of write
  • Only concerns the writer.

These still need tests:

• Fix unescaping broken codepoints
  • This likely broke the html5 test, and needs some new tests
• Add '*_with_custom_entities' versions of unescape methods
  • Has some private tests, but needs tests on new public methods
  • Note to self: Probably a good place to check for the billion laughs attack vulnerability too
• Put escaping html entities behind a 'escape-html' feature
  • Has no tests; also kinda abuses features, as features should not change behavior of existing APIs
• Split text trim into start and end
  • trim_text option becomes two options: trim_text_start and trim_text_end
  • Definitely needs tests
• Accept html5 doctype
  • Allows case-insensitive matching of the doctype tag
  • Has tests, but they fail (likely broken by fix #270 #271)
• Unescape all existing HTML entities
  • This adds quite a lot of new entities.
  • It likely has nothing to do with the unbuffered code, but a test would be nice anyway.
• Add Clone for Reader
  • A test would be nice to make sure nobody removes that derive by mistake

Pull requests:

• Improve CI: test with more feature combinations, test on windows-latest too #292
• Add tests for escaping/encoding, trimming, and cloning #308

[andreivasiliu]

Just an update, still working on this, will likely have something in a week or two at most.

[tafia]

I have merged #290 even if it is scary :).

I'll bump the 0.x so it won't accidentally break anything. On one hand I'd like to wait for your tests before pushing a new version to crates.io, on the other there may be many people actually waiting for it. I could perhaps have a -alpha release?
Any thought?

PS: Thanks a lot for your hard work

[andreivasiliu]

Oh, cool. No, I don't mind that, although I personally would go with an -alpha for now for at least a week or two; I'm sure people can wait that long.

[tafia]

Just published v0.23.0-alpha1. there seems to be an issue with GitHub so I can't push the commit. I'll try later

[andreivasiliu]

Just wanted to chime in that I tried alpha2 on a work project where I was using borrowing on a really large and complex XML (~5Mb), and it worked with no issues. Although I'll still be continuing to work on this issue of course.

I noticed I was using #[serde(borrow)] everywhere, which is correct, but I was surprised (yes, despite it being my own code). I feel quick-xml should teach that to users, especially for Cows (see "this field is never borrowed" in serde's docs); I'll PR a README.md update for this.

[tafia]

Thanks for the report! (and for everything else you're doing)