Codecov Report

Merging #103 into master will decrease coverage by 0.18%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master     #103      +/-   ##
==========================================
- Coverage   84.49%   84.31%   -0.19%     
==========================================
  Files           8        8              
  Lines         458      459       +1     
==========================================
  Hits          387      387              
- Misses         71       72       +1

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 861a222...ac912d9. Read the comment docs.

Codecov Report

Merging #103 into master will decrease coverage by 0.63%.
The diff coverage is 73.68%.

@@            Coverage Diff             @@
##           master     #103      +/-   ##
==========================================
- Coverage   84.71%   84.07%   -0.64%     
==========================================
  Files           8        8              
  Lines         458      471      +13     
==========================================
+ Hits          388      396       +8     
- Misses         70       75       +5

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 1558072...a1e0344. Read the comment docs.

Is there a reason why we can't just use Dict instead of defining a custom Schema type?

Not really, although having a wrapper type allows for more convenience constructors (e.g., if you use symbols as keys in your constructor instead of Terms). But really I just added it for dispatch so any type would work for that.

Not really, although having a wrapper type allows for more convenience constructors (e.g., if you use symbols as keys in your constructor instead of Terms).

But aren't symbols accepted already in dicts? That can be handled when applying the schema.

At the moment you can pass a dict with symbol keys and it'll choke since apply_schema tries to use Terms to index. So the way that happens will have to be changed in every single apply_schema method (and worse, anyone who writes their own method will have to be careful to include whatever support code is necessary). That sounds like too much boilerplate to me, which can be avoided by using a wrapper type that provides a single place for enforcing a particular standard and accessing the elements.

Anyway, I'm not really wedded to using a wrapper type (and in fact I tried to avoid it originally, using a vanilla dict), but having something concrete that's easy to dispatch on makes it much easier to avoid method ambiguities which were becoming a big headache for apply_schema.

OK. One solution would be to have terms and symbols compare and hash equal (so that they can be used interchangeably in dicts), but I guess you wouldn't want that.