Silverlight and Azure Table Data Paging

[Cross posted from here]

I’m playing around with a data visualization app using Silverlight and data hosted in Azure Tables and have been learning quite a bit in the process. Firstly, Azure tables only allows you to return 1000 records in a given query. If you issue a query that has a larger matching result set, Azure will return some extra headers indicating as such (x-ms-continuation-NextPartitionKey and x-ms-continuation-NextRowKey). It wasn’t hard to find an example of data paging using Azure table data, however it used the Execute() method of the DataServiceQuery object. Unfortunately, this isn’t available in Silverlight as you have to use the asynchronous methods (BeginQuery and EndQuery). I’m a bit slow, and for whatever reason translating the MSDN sample for synchronous to the asynchronous model took me longer than it should have. I’m posting this so that maybe the next person will find this, get the answer they need, and move on and not waste the same amount of time I did.

My button event handler looks quite a bit different from the MSDN sample but is pretty easy to figure out:

I instantiate the context, create a query based on that context, cast that query as a DataServiceQuery<t> and then call the BeginExecute method passing my callback method and the query as my state object. (Note: in case you are wondering about the Where clause in the query above, I know that all of the data that matches the first conditional is located in a given partition within the Azure table and have found that specifying the partition greatly increases the performance).

My callback method (ProcessDataRequest) does a bit of recursion to support the unknown number of subsequent calls needed to retrieve all of the matching records.

The contents of the ProcessDataRequest method are listed below:

Note that unlike some samples that are simply focusing on async data calls and don’t handle paging via headers, I cast the output of EndExecute as a QueryOperationResponse object which allows me to subsequently access the headers and interrogate them for the continuation keys. If I find the continuation keys, I create a new query object, set the additional query options, and execute it in the same fashion as the original call.  The AddPointsToScreen method simply processes the values and renders them as polygons to the screen. I’ve included it here not because there is anything special in it, but rather for completeness.