Add has / !has operators

[lucaswoj]

ref mapbox/mapbox-gl-style-spec#364

This will require a minor version bump.

cc @mourner @jfirebaugh @ansis

Todo

• merge this PR
• release new feature-filter version https://www.npmjs.com/package/feature-filter
• add support to mapbox-gl-style-spec Add support for has / !has operators mapbox-gl-style-spec#451
• update feature-filter and mapbox-gl-style-spec in GL JS
• port changes to GL Native

[mourner]

👍

[jfirebaugh]

I'd expect this to be true.

[lucaswoj]

This PR is blocked on a decision about how the has operator should handle "literally undefined", undefined, and null. In vector tiles, the three are one-in-the-same. In GeoJSON, the three could be differentiated.

Open questions:

• Should has evaluate to true or false for null?
• Should has evaluate to true or false for undefined?
• Should has evaluate to true or false for "literally undefined"?
• Will these decisions be consistent with GL Native?
• Will these behaviours be predictable for a GeoJSON file converted to vector tiles serverside?
• Will these decisions be consistent with future implementations of geojson-vt?

[jfirebaugh]

Let's go with the definitions you've chosen:

• has is false for undefined
• has is false for "literally undefined"
• has is true for null

We may add explicit null to a future VT spec. I think it's unlikely we'll ever add explicit undefined.

[arunasank]

👋 @lucaswoj - this is going to be super useful to find stats with https://github.com/mapbox/osm-tag-stats (mapbox/osm-tag-stats#2) . Will this get merged soon?

[lucaswoj]

@arunasank I'll get this into the next release!

[lucaswoj]

With humility and hindsight, I think @jfirebaugh's proposed definitions for undefined are best. This is ready for 👀 . I also added a commit with some minor feature filter cleanup.

[jfirebaugh]

This doesn't work for strings.

[lucaswoj]

Fixed in latest commit

[jfirebaugh]

How so? I still see a - b which returns NaN for strings.

[lucaswoj]

😞 It got lost in the rebase shuffle. 😄 Fixed now!

[jfirebaugh]

With humility and hindsight, I think @jfirebaugh's proposed definitions for undefined are best.

Ah, ok, I'm fine with that too. 😄 For those following along, the behavior in question is "should has evaluate to true or false for a property that exists and whose value is undefined?" The behavior we're settling on is that has will have the same behavior as the Javascript in operator, evaluating to true in that case.

[lucaswoj]

Cleaned up the commit history and addressed #15 (comment). Ready to 🚢?

[jfirebaugh]

🚢