fulmicoton / fattable

Javascript Library to create scrollable table with infinite rows and columns.
MIT License
482 stars 39 forks source link

Fattable

Demo

Checkout the demo here.

What is it?

Fattable is a javascript Library to create table with infinite scroll, with infinite number of rows and number of columns.

Big table (more 10,000 cells) don't do well with DOM. Your scroll will start getting choppy.

Also big tables can rapidly grow in sizes. It is not always possible to have clients download or even retain all of the table data. Fattable includes everything required to load your data asynchronously.

This library is

API

var table = fattable({
  "painter": painter,    // your painter (see below)
  "model": model,          // model describing your data (see below)
  "nbRows": 1000000,     // overall number of rows
  "rowHeight": 35,       // constant row height (px)
  "headerHeight": 100,   // height of the header (px)
  "columnWidths": [300, 300, 300, 300] // array of column width (px) 
})  

Painter

painter is an object which role is to fill the content of your cells, and columnHeaders. It is expected to implement the following interface.

var painter = {

    "setupHeader": function(headerDiv) {
        /* Setup method are called at the creation
           of the column header. That is during
           initialization and for all window resize
           event. Columns are recycled. */
    }
,
    "setupCell": function(cellDiv) {
        /* The cell painter tells how 
           to fill, and style cells.
           Do not set height or width.
           in either fill and setup methods. */
    }
,
    "fillHeader": function(headerDiv, data) {
        /* Fills and style a column div.
           Data is whatever the datalayer
           is returning. A String, or a more
           elaborate object. */
        colHeaderDiv.textContent = data;
    }
,
    "fillCell": function(cellDiv, data) {
        /* Fills and style a cell div.
           Data is whatever the datalayer
           is returning. A String, or a more
           elaborate object. */
        cellDiv.textContent = data;
    }
,
    "fillHeaderPending": function(headerDiv) {
        /* Mark a column header as pending.
           When using an asynchronous.
           Its content is not in cache
           and needs to be fetched */
        cellDiv.textContent = "NA";
    }
,
    "fillCellPending": function(cellDiv) {
        /* Mark a cell content as pending
           Its content is not in cache and 
           needs to be fetched */
        cellDiv.textContent = "NA";
    }
};

Actually this very simple implementation is the default. And it is available as fattable.Painter, so that you can just override it.

DataLayer

Synchronous Data Layer

Demo

If your data is not too big, you probably can just fetch your data all at once, and then display the table. For this simple use case, the best is probably to extend the SyncTableData object.

You just need to extend fattable.SyncTableModel and implement the following methods

{
  "getCellSync": function(i,j) {
    return "cell " + i + "," + j;
  },
  "getHeaderSync": function(i,j) {
    return "col " + j;
  }
}

Asynchronous and paged async model

Demo

You probably don't want your backend to receive one request per cell displayed. A good solution to this problem is to partition your table into pages of cells.

Queries are only sent when the user stops scrolling.

To use such a system, you just have to extend the PagedAsyncTableModelclass with the following methods. In addition, it include a simple LRU cache.

{
  "cellPageName": function(i,j) {
      // returns a string which stands for the id of 
      // the page the cell (i,j) belongs to.
      var I = (i / 128) | 0;
      var J = (j / 29) | 0;
      return JSON.stringify([I,J]);
  },
  "fetchCellPage": function(pageName, cb) {
      // Async method to return the page of 
      var coords = JSON.parse(pageName);
      var I = coords[0];
      var J = coords[1];
      getJSON("data/page-" + I + "-" + J + ".json", function(data) {
          cb(function(i,j) {
              return {
                  rowId: i,
                  content: data[i-I*128][j-J*29]
              };
          });
      });
  },
  "headerCellPage" : function(j) {
   // Same as for cellPageName but for cells.
  },
  "fetchHeaderPage" : function(j) {
    // Same as for fetchCellPage but for headers
  }
}

Custom async model

If you want to go custom, you can implement your own data model, it just has to implement the following methods :

{
  hasCell: function(i,j) {
    // returns true if getting the data of the cell (i,j )
    // does not require an async call false if it does need it.
  },
  hasHeader: function(j) {
    // ... same thing for column header j
  },
  getCell: function(i,j,cb) {
      // fetch data associated to cell i,j 
      // and call the callback method cb with it
      // as argument
  },
  getHeader: function(j,cb) {
      // ... same thing for column header j
  }
}

Troubleshooting

Using <!DOCTYPE html> for using the library with IE >=9. IE < 9 is simply not supported.

The library currently use a huge container with overflow. Browser have a bit of limitation on the size of such a container. Expect some bug appearing at 40000 rows in IE, 500000 rows in other browsers.