Leaflet 0.7 and plans for future

index.md

For the last and the ongoing week, apart from a few days speaking at a conference in Tallin, I've been focused on pushing Leaflet towards the next stable release, 0.7.

This is a bugfix-heavy release — as Leaflet becomes more and more stable feature-wise, the focus shifts towards stability, usability and API improvements over new features. I've also been holding back some of the planned deep refactorings (which I'll talk about later in the post) until 0.7 is released, so that the heavy risky stuff is done at the beginning of the release cycle, leaving plenty of room to catch bugs and incompatible changes that can unintentionally break existing apps.

0.7 changes

You can check out the detailed changelog of what's already done over the recent months for 0.7 (about 75 improvements and bugfixes), but I'd like to mention some highlights:

• Added support for IE11 touch devices. Damn MS unexpectedly broke their pointer API compatibility between Developer Preview and final IE11 release, and we eventually rewrote quite a bit of code to make everything work smoothly across all IE versions (both dekstop & mobile), fixing a bunch of IE10 bugs along the way as well.
• Officially dropped IE6 support (nobody cares anyway) and cleaned up/fixed IE7-8 styles.
• Dropped the need for IE conditional comment when including Leaflet, making the snippet much simpler — all IE7/8-specific styles got simplified and moved to the main leaflet.css file.
• Fixed an obscure iOS7 memory leak that crashed Safari when you tried to create several thousands of layers (e.g. markers for clustering). I still don't understand why it happens, but we managed to fix it with a bit of trickery.
• Fixed a critical Chrome for Android bug that made the tiles disappear after zooming on some devices.
• Added the ability to upscale tiles to higher zoom levels (e.g. have zoom 19-20 when the source has 18 max).
• Removed some Earth-related hardcode in TileLayer implementation to make it easier for plugins like Proj4Leaflet to handle complex projections without horrible hacks. Some other work in this direction to follow in 0.8.
• Improved panning performance on complex pages with significant number of elements — we found out that simple things like setting a different cursor to document (for a "grabbing" hand) caused noticeable performance hit on some browsers (Chrome in particular).
• Changed the way maxBounds works, not enforcing a derived minZoom from it but restricting panning across lower zoom levels, along with some tricks to make it play better with panning inertia or offset zooming, etc.

Plans for 0.8

There are several big undertakings in refactoring Leaflet that I'd want to switch to immediately after releasing 0.7 — I've been holding them off for too long, and they'll be extremely beneficial for plugin and Leaflet-based API authors. Some of them are already in progress.

• Refactoring the layers architecture. Currently there's a lot of duplication of logic across implementation of different layers (map, markers, vector layers, etc.), specifically event handling, zoom animation logic, zIndex and pane handling (what appears on top of what etc.). Making the code consistent, more universal and shared across different layers will make it much easier to customize layers and make your own (e.g. integrate d3, etc.)
• Splitting the huge TileLayer implementation into GridLayer and TileLayer, separating image tiles-related logic and grid-logic that will make other grid-like layer implementations (e.g. UTFGrid interaction or tiled GeoJSON) much simpler.
• Refactoring zoom animation logic to make the long-awaited Easey-style animations (mapbox.js 0.6-style) possible.
• Refactoring projections code to make it easier to set up flat maps and weird projections and customize how Leaflet handles them.
• Refactoring the vector layers code to make it possible to use different rendering backends (Canvas, SVG, etc.) for different layers on the same map and switch between them easily. This will also open it up for interesting extensions, like indexing layers with RBush for fast interaction features.

While it's an ambitious plan and it may take more than one stable release, finishing all those refactorings will mean that Leaflet is getting ready for a 1.0 release.

Another direction I'd like to focus on after releasing 0.7 is website and documentation improvements. First, Leaflet is begging for more step-by-step tutorials (with more advanced features like custom layers, custom controls, etc.), and I'd love to do a docs/tutorials sprint some time in future. Second, the presentation could be significantly improved — adding a prominent visual showcase or app gallery, making Leaflet users more prominent with some logos and quotes/testimonials, and updating the layout/design for a more stylish, clean look, etc.

Hope that gives a good glimpse of the stuff to expect from Leafet in near future, and don't hesitate to ask any questions in comments — I'll be happy to answer!