Introduce the concept of an indexed vector tile

[chelm]

We want to contribute the code Esri's added / changed back to Mapbox. Let's use this issue to help spur that along.

[mourner]

That would be so awesome. Excited to see some contributions soon!

[chelm]

@mourner I'm sifting through the code we have internally and building a branch that has most of the changes that have been made. I hope to pull in @ycabon & @jcardonadcdev so they can review the branch (as they have a better understanding of the changes).

The first version will probably include the addition of a new data source type, and will ideally act as a conversation started between Esri and @mourner @tmcw & @jfirebaugh and the world.

[alexmcroberts]

🙌

[chelm]

I've got the first cut of a branch pushed

A few things to note

• I will do my best to explain what I've done here, but it must be pointed out that I was not the original author of the changes AND that I am in fact doing things differently in this branch than was the original intent.
• The current build of this code wont work with the current hosted Esri stylesheets as they use type: 'indexedVector' and here I've removed that type in favor of using the existing vector.
• There are changes that we need to like make in two other repos:
  • Required: https://github.com/mapbox/vector-tile-js
    • will add some work we've done on overzoomed features
  • Optional: https://github.com/mapbox/mapbox-gl-style-spec
    • add support for an "index" source property
• I'll next be working on pushing an html and stylesheet demo to show this working (and prove to my self that it works)

[chelm]

also I've gone ahead and checked in the original code, flagged as js/source/esri-*.js

Since I'm sorting of putting myself in the middle here (suggesting changes in 2 directions, internal to Esri and external) - I'd really love some comments from the JS team at Esri (@jcardonadcdev or @ycabon) and most def those from Mapbox

[chelm]

Changed the title to better reflect the actions needed in this issue. A PR will come out of it.

[chelm]

Okay. I've navigated a small minefield of changes between versions and have this branch working up to point where requests for missing tiles in the index are replaced with requests for parent tiles.

the next steps are to correctly scale and render geometries from parent tiles. This will potentially come in as new loadGeometry method loadClippedGeometry (in the vector-tile-js repo).

[tmcw]

👍 Cool: I've opened a pull request to incorporate these changes into the repo, and so that we can do inline code reviewing. Checklisted out next actions over there.

[tmcw]

cc @ycabon & @jcardonadcdev is the public GitHub repo where the work on this is happening? Just want to make sure we're all on the same page, since mapbox-gl-js master has been moving as of late with the introduction of the v8 style spec.

[ycabon]

hey @tmcw, the team is still compiling some numbers about the impact of the index client-side and cooking wise. It's close to completion. That will determine what direction we would finally take.

[tmcw]

@ycabon ping, any word from Redlands?

[ycabon]

@tmcw Getting there :). I will push as soon as I get write access later this week.

[ycabon]

@tmcw the latest commits have been pushed. I also pushed changes for vector-tile-js. https://github.com/Esri/vector-tile-js/tree/clipping-logic

Before moving forward we will give some context about what's the purpose of those PRs.

[odoe]

Here is a gist that discusses the concept of the Indexing of Vector Tiles and the justification for Clipping.

https://gist.github.com/odoe/ce6a150658526901ef27#file-vector-tile-pr-md

cc @tmcw @ycabon @jcardonadcdev