i) write instances of AbstractDict instead of Dict; ii) YAML.yaml returns a String.

[mirkobunse]

These are two minor changes which are combined in a single PR because they can nicely be tested together.

Writing AbstractDicts

This feature solves #82. As pointed out in this issue, writing instances of AbstractDict allows users to make use of ordered dictionaries like OrderedCollections.OrderedDict. Thus, the user can determine in which order the properties are written.

Note that none of the tests is broken (i.e. all test instances are written correctly) if the Dict requirement is dropped in favor of AbstractDict. Hurray! I conclude that my initial implementation of writing YAML Strings (#68) was unnecessarily picky with the Dict type.

YAML.yaml

YAML.yaml returns a YAML-formatted String representation of a dictionary. It is implemented as a simple alias for YAML.write, which could have done the same thing already. The name YAML.yaml, however, complies with the beautiful API of JSON.jl, which has JSON.json return a JSON-formatted String representation of a dictionary.

This feature takes a step towards #70 without making any breaking change (only an alias is provided). Also in #29, it has been argued in favor of a YAML.yaml method: #29 (comment)

[codecov]

Codecov Report

Merging #86 into master will increase coverage by 0.26%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #86      +/-   ##
==========================================
+ Coverage   74.30%   74.56%   +0.26%     
==========================================
  Files          12       12              
  Lines        1331     1333       +2     
==========================================
+ Hits          989      994       +5     
+ Misses        342      339       -3     

Impacted Files	Coverage Δ
src/writer.jl	95.12% <100.00%> (+14.16%)	⬆️
src/composer.jl	89.74% <0.00%> (-1.17%)	⬇️
src/constructor.jl	70.00% <0.00%> (-0.47%)	⬇️
src/parser.jl	70.55% <0.00%> (-0.23%)	⬇️

---

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 eb7af07...9b24a75. Read the comment docs.

[kescobo]

This all makes sense to me, but I'm not a big user of this package. I'll merge in a couple of days if there are no objections so that it doesn't languish, but can you maybe post on slack or zulip or discourse to see if there's any feedback from people that are actually using it?

[mirkobunse]

What exactly should I ask?

My motivation to open this PR was mostly the last point I mentioned: taking a step towards #70. What I did not yet mention is that, besides the trivial YAML.yaml, also the writing of OrderedDict instances is preparing a solution to #70. Why?

JSON.jl allows the user to pick the subtype of AbtractDict that should be produced by the parser. If a user picks an OrderedDict, it will reflect the order of properties in the JSON file. Since writing supports OrderedDict instances, this order is maintained further. Consequently, a round-trip read/write yields highly similar files.

Some users of YAML.jl have discussed this behaviour as a desireable feature since quite some time, in #29. With writing instances of AbstractDict, they are one step closer to this feature.

The PR we are discussing here is nothing fancy, for sure, but it'll help get some other things done.

[kescobo]

What exactly should I ask?

Basically just - "can people that use YAML.jl take a look at this PR and provide feedback or raise concerns if you have them?"

I'm sure it's fine, especially if you expect it to be non-breaking - I just don't have any experience with this package, and I'm worried that the people who do aren't watching the issues / PRs.

[mirkobunse]

https://discourse.julialang.org/t/looking-for-feedback-on-prs-in-yaml-jl/45116

[GunnarFarneback]

I have no opinion either way about the YAML.yaml function. The rest of the PR is uncontroversial and fine.

[mirkobunse]

@kescobo, how long do you plan to wait before merging?

I want to continue with #70 (comment) but not before this PR is done

[kescobo]

All set 👍.