Last week I needed to convince some folks that the World Wide Web can act as a deployment platform for enterprise applications. They had some bad experiences with Web applications a couple of years ago and, since that time, felt convinced that desktop applications could only serve that niche of data-intense applications. I believe in the power of the browser and worked on a small demonstration Web application that would corroborate my beliefs.
Data-rich applications traditionally have grids. Users have grown comfortable scanning hundreds of rows of text and numbers to perform some kind of analysis. Trying to convince them that alternatives exist that could present the data in more comprehensible ways can pose challenges because these good people have adopted the Religion of the Grid.
They can have my Excel when they pry it from my cold, dead fingers.
- gracefully handle thousands of cells
allow keyboard-only input to:
- activate cells with arrow keys
- activate cells with TAB and ENTER keys
start editing cell content by
- pressing F2 to edit at the end of the cell
- start typing to replace the entire cell's content
- comply with the jQuery UI CSS framework
- have read-only functionality for cells, rows, and tables
- have column resizing
- handle fixed widths and heights
- have fixed headers
- have a pluggable editor architecture
That last requirement emerges from a vested interest in using knockout.js for binding data in the browser. I really want an interactivity widget that rests atop knockout‘s HTML rendering. I don’t want the grid to render, just handle interaction.
For one reason or another, the grids that I’ve used in the past failed to meet one or another of these needs: jqGrid, ExtJs Grid v3 and v4, SlickGrid, dojox.grid, and the DataTables from YUI 2 and YUI 3.
That’s pronounced “gridge.” The “q” is silent.
That’s the mantra of software developers everywhere. And, I believe it. Last year I built a fixed header grid for my former employer which garnered a lot of praise. It didn’t have most of the requirements listed above; however, I figured I could do it without too much pain. With jQuery and jQuery UI on my side, how could I not get it done quickly?
I “finished” it in three fifteen work hour days. As it goes with a lot of projects, I wrote the majority of the code during the first third and spent the next two-thirds of the project time tweaking the script and style for it to meet my expectations.
This week I will dive into the details of creating grijq. I’ll write about the main challenges of creating it. The menu reads (and I hope it makes your mouth water)
Tuesday : Designing the scrolling of a fixed-header grid
Wednesday : Get column resizing working on big tables
Thursday : Navigating the grid with the keybaord
Friday : Creating a pluggable editor system
Remember, grijq does not promise feature parity with the grids listed above. I designed grijq as an eventing and interaction layer. It should tie into a rendering system. I like the cleanliness of separating those two concerns: one thing renders HTML and another handles the interactivity with the user. In the Taligent MVP pattern, grijq acts as an interactor between the view (the browser) and the presenter (in my case, knockout.js).