We announced the Grid project back in February. Since then, we finished the first three stages and are now getting started on the fourth. In this post we’ll take a look at what we’ve built so far.
Grid
This is what we designed as the “zero feature grid”. All it does is enhance HTML tables, but it does that pretty well while also providing the right hooks for all kinds of other features that we can add on top of that. And that turns out to be really useful, as there are a bunch of things that regular HTML tables can’t do. With Grid, you get:
- styling with the CSS framework, making it ThemeRoller-ready
- a proper titlebar, based on the table’s caption
- markup and styles necessary for scrolling of the table body, while keeping the header fixed
The API for the Grid controls what content to render, and how to present it:
- The columns option specifies which columns to render. If not specified, it picks up existing table header elements.
- The rowTemplate option allows the grid to render each row from a custom template. If this option is not provided, each row is generated based on the columns option.
- The source option specifies the content to render, in the form of a plain array of objects. When not specified, existing table rows are used.
- By default, the grid’s body grows with the number of rows. With heightStyle: ‘fill’, it stays at a fixed height and the body starts scrolling if there are more rows.
- When picking up column information from the existing table headers, certain data-attributes are read. The grid itself uses only the data-property attribute (and the actual text of each header cell), but the dataFields option specifies a few more attributes that other components can use, such as “type”, “culture” and “format” to configure local sorting and filtering. If a grid add-on handles other data attributes, adding those to the dataFields option will make them automatically available as part of the columns option.
So far we only have an enhanced table, but often enough, a grid needs to be fed data from some remote resource. That’s where Dataview comes in.
Dataview
Dataview is a low-level abstraction for retrieving content. It has an API for specifying what content to retrieve, and an SPI (service provider interface), implemented by components that provide that content. There are built-in options for filtering, sorting and paging, and the design makes it easy to add more options, like grouping. The API is asynchronous by default, even for local data, so that all components relying on Dataview can work with both local and remote data.
Dataview only depends on Widget and Observable (we’ll get to that below), but not on Grid. This allows Dataview to be used in a variety of contexts. For example it could power a product listing, such as those on eBay or Amazon, where a table presentation is not the right format.
The Dataview SPI makes it easy to write implementations from scratch as well as to create reusable extensions.
As an example, we can use a custom dataview as the input for the Autocomplete widget. The same data is also displayed in the grid below the input field. This demonstrates how a Dataview can be used in multiple representations, where each decides on its own what data to show.
As for reusable extensions, we currently have three implemented:
- localDataview takes an input array and does sorting, paging and filtering on that array. It uses Globalize (see below) to implement filtering and sorting of localized numbers and dates. When combined with the grid on a table with existing content (without specifying the source option), you get a complete tablesorter.
- odataDataview takes a resource URL, pointing at a webservice that understands OData, the Open Data Protocol. While our implementation doesn’t yet cover all of the OData options, you can use it to sort, filter and page, without having to implement any custom request/response mapping. We have an example of a grid using the OData based Netflix API.
- preloaderDataview wraps a Dataview and adjusts the paging behaviour to load more data than rendered, paging locally before preloading another batch. In this Flickr API slideshow example we preload both the API responses and the actual images. This pushes all the loading in the background, allowing the user to page through images without interruptions.
All three need testing in actual projects before we can consider them stable. We’re also looking for other use cases for dataview that we might be missing.
Observable
Data binding is currently a common theme among newer JavaScript frameworks, with various competing solutions available. We don’t yet have a full-featured alternative, but we’ve developed a low-level abstraction that might power a data binding component in the future. We call this abstraction Observable. It provides an API for making changes on plain JavaScript objects and arrays that can be observed by listening to events that each change triggers. We’ve designed Observable to have a very small number of methods and events, with the purpose of making it easy to implement the same API in other contexts. These events are:
- change: Triggered on objects after one or more properties have changed.
- insert: Triggered on arrays after inserting one or more new items.
- remove: Triggered on arrays after removing one or more items.
- replaceAll: Triggered on arrays after replacing all items in an array.
That last event may seem odd at first, but this makes it possible to create a dataview, pass it to a grid, have the grid subscribe to events on that dataview’s output array, and then update itself when the dataview updates.
For all four events, Observable provides method counterparts:
- property: to work on objects
- insert, remove and replaceAll: to work on arrays.
Usage is consistent in both cases:
- $.observable( object ).property( “name”, “Fred” );
- $.observable( array ).insert({ name: “Peter” });
We’re still working on Observable. The next step is to find an abstraction for the array bindings, which would remove some overhead both from the Grid as well as from the todo-app demo.
When combining Observable with the Grid, we can easily add editing capabilities. In our grid editing demo, you can add, edit and remove developers from the table. Results are persisted via localStorage.
Instead of adding custom columns that have “Edit” and “Remove” buttons, we can use the selectable interaction to select one or more rows and then interact with the selection. In that demo, the selected array is also an observable array, which the second grid displays. Note that you can select rows on one page, go to the next page (you may need to add a few developers for a second page to show up), select more rows with command/ctrl-click (to extend the selection) and it will display all those rows in the second grid. The selected state is separated from the visual representation, making selection over multiple pages pretty easy.
Globalize
Globalize originally started as the jquery-global plugin. We rewrote it as part of the Grid project to be independent of jQuery, allowing usage on both client and server. Our localDataview implementation (mentioned above) uses it, and its also behind custom inputs like Spinner and Timepicker.
Mask, Timepicker, Selectmenu
These three custom inputs are being developed as part of the Grid project. They are intended to be used within the Grid for inline editing, as well as standalone in regular forms or websites. We’re getting close to landing all three in master, to release them as part of 1.9.
And more…
Work is still in progress on Template, Datepicker and other Custom Inputs. We’ll report back on those once we have more stable solutions.
Up next
We’re finishing existing components and are starting on Stage 4. For details, check out the main Grid wiki page.
So, when is it done?
Our Roadmap has the Grid on the 2.1 release, while the main focus for jQuery UI is on 1.9. In other words, there is a lot that needs to get done before we can put the Grid into a stable release and support it along with all the other components. At the same time, we encourage you to start testing the features outlined above now. Grid, Observable and Dataview are all pretty stable and working well. Globalize is a separate project that you should start using now if you want to support localized number and date handling on the clientside.
We value your input and help. Please use the Developing jQuery UI forum, comment on individual wiki pages or visit us on IRC (#jqueryui-dev on Freenode). If you found a solution to an issue, send a pull request.
When designing the Autocomplete widget, we started with 
There’s also a set of 
Going a step further, you can use custom data formats and define custom displays. For example, you can