Designing browser-enabled forms for performance in InfoPath Forms Services (Part 2)
Part 2 – Postbacks and conditions that cause them
Welcome to the second article in our series on optimizing the performance of browser-enabled forms designed with Microsoft Office InfoPath 2007. In our first article, we defined performance in terms of responsiveness. In this article, we’ll focus on postbacks, one of the most common causes of reduced performance
Postbacks are probably the most important factor affecting form performance because the effect of other conditions is amplified with each postback. There are two significant aspects of postbacks which we’ll examine in this series. First, which form features cause postbacks? Those features are the subject of this article. Second, what kinds of postbacks are more expensive to process, thus imposing an additional burden on the performance and scalability of the form system? We’ll examine those factors in the next article.
What is a postback?
Certain form controls, actions, and features require the browser to communicate with the server during the form-filling session. This exchange of data during the session is called a postback, and usually occurs when a form feature needs to send data to the server for processing. After sending the postback request, the browser must then wait for a response before it can update and re-render the form with the new data. In the case of a full-page postback, the browser must stop rendering completely until the whole updated page is returned, blocking user interaction with the form until it has been re-rendered. Alternately, to improve the user experience in the browser, some features use a partial postback, where only a specific part of the form is updated. This allows the user to continue interacting with other parts of the form while the update happens in the background. In either case, however, unnecessary postbacks impose an additional load on both the browser and the server. If you can reduce or eliminate postbacks as you design, develop, and test your forms, you can improve the performance and scalability of your entire form system.
Browser-enabled form handling compared with InfoPath client forms
To better understand the effects of postbacks on performance, it is helpful to compare basic form handling of the InfoPath-rich client with InfoPath Forms Services and browser-enabled forms. Whether you are new to InfoPath, or you are familiar with rich client forms but new to InfoPath Forms Services and browser-enabled forms, understanding these basic differences will help you improve your form design and troubleshooting strategies.
The major difference between the InfoPath-rich client and browser-enabled form systems is that the InfoPath client doesn’t need to communicate with the form server as frequently as the browser does. This is because the InfoPath client processes most forms locally on the user’s computer. In contrast, the browser form relies on the server for a variety of form processing operations, requiring more communication to exchange data between the browser and the form server. Communication in the form of postbacks is a fundamental feature that allows InfoPath to support browser-enabled forms on the Web, but postbacks also make browser forms more sensitive to performance issues.
Let’s look at two form loading and editing scenarios that further illustrate the difference between form processing in the InfoPath client and browser-enabled form systems.
In the InfoPath-rich client scenario, when the user opens the form from a SharePoint form library, InfoPath first determines that the form should be opened in the client rather than the browser. After the initial download of a copy of the requested form template, InfoPath opens and renders an instance of the form on the user’s computer for editing. The form instance in the client is an XML document that the client renders in HTML for display. As the user interacts with the form, the InfoPath client processes and renders changes, and might run some custom form code, all within the context of the client running on the user’s computer. Generally, the only time the client needs to connect back to the server is to make additional data connection requests to query from or submit to a data source, to save a partially completed form, or to submit the final form.
Finally, the server sends the response through a Web Front End server over the network to the browser on the user’s computer, which renders the initial download into the form view that the user interacts with. As the user interacts with the form in the browser, some form processing happens in the browser, but more complex form elements and events require a postback to the server for processing.
During the transmission and processing of the form data, the browser instance of the form must wait for a response and then re-render the updated form. In this communication cycle, each postback request and response set constitutes a roundtrip. Each roundtrip affects the performance and scalability of the form system in terms of throughput, responsiveness, and resource utilization. This impacts both the browser side and the server side, as well as the network.
As you can see from these two scenarios, a form-filling session in the browser-enabled form system requires substantially more communication than a similar session in the InfoPath client. In addition, more of the browser-enabled form processing is concentrated on the server, especially when multiple form-filling sessions are being served. In contrast, with an InfoPath client form, the form processing is distributed to the user systems. Consequently, reducing the number and complexity of postbacks as you’re designing browser-enabled forms can substantially boost performance.
Ten common causes of postbacks
A calculation can cause a postback when there is a condition that cannot be evaluated in the browser and requires processing by the server. For example, if a calculation requires data that is not loaded in the DOM, then the browser must send a postback to the server and wait for the result. Another condition that will invoke a postback involves any calculation that is out of the context of the current node. For example, you could create an expression such as ...my:field1+. When the expression is invoked in the form, because it requires traversing up the tree several levels from the current node, it must be sent to the server for evaluation.
The function of a data adapter is to make a query connection to an external data source, such as an XML file, a database, a SharePoint list, or a Web service, so that a form can receive and be updated with some data from the data source. If a form invokes a data adapter when the form loads, the browser must wait for the data connection, increasing form load time. If the form invokes a data adapter after the form loads, a postback containing the query request is first sent to the form server, which executes the data connection request on behalf of the browser form, adding additional time to the roundtrip.
Multiple binding: data bound to multiple controls
Multiple binding happens when one piece of data in the form is bound to two or more controls in the same view. This is usually done so that the same data can be displayed in multiple places. For example, you might design a report with several sections to include a single non-repeating field, such as "Report Date", inside a repeating section so that the same date is displayed in each report section. If you design a form this way, then changing the "Report Date" field in one location will update it everywhere because it is multiply bound. Unfortunately, the cost of this multiple binding might be a roundtrip to the server if the binding is more complex than the browser can resolve. In this case, server analysis is required to reliably identify all the controls bound to the same piece of data. In addition, if something is both multiply bound and out of context, then user edits to form fields will always require roundtrips. Other form actions that will cause roundtrips to the server when there is multiple binding include insertion or removal of repeating tables and repeating or optional sections.
Out-of-context binding describes a situation where a control inside a repeating or optional section is bound to a node that is in a different part of the XML data source tree, outside the group node that contains the repeating or optional section control. When an out-of-context situation occurs in a form, the correct binding may be too complex to determine within the browser. If either the validity or the kind of action—such as a calculation—cannot be determined in the browser, the out-of context binding will trigger a roundtrip to the server to resolve the binding.
You are most likely to encounter this problem as you’re designing a form in the InfoPath Designer if you first insert a group or control, and then later decide to move that group or control into an optional section. To avoid creating an out-of-context binding, after you drag and drop the control into the optional section, in the data source pane, drag the node the control is bound to into the group for the new optional section. The group or control will then be bound to the correct context.
Multiple views are often used to present a specific view to a specific user audience in a workflow. Multiple views are also commonly used to improve the performance of a form; by careful grouping of related items and limiting the number of controls in each view you can improve the form loading and rendering performance compared to placing everything in a single view. However, each time the view is switched, there is a full postback to send the user’s changes to the server, get the new view, and reload it in the browser. If possible, try to group the items on a view so that any controls that might postback do so when the view is switched. Then only one postback occurs when switching the view. A good example of this is using a button to invoke a query connection that returns data for a list, an operation that requires a postback. If the view is also switched when the button is clicked, you improve the efficiency of the postback. Another benefit of this design is that it makes the view switch a user-invoked action, which improves the user’s perception of the form response. By coordinating the postback of controls in the view with a view switch, you not only improve the responsiveness of the form, but also reduce stress on the server and network by eliminating extra postbacks. When designing your form there is no formula for when to use multiple views, so you will need to balance your use of multiple views with the potential design and performance advantages and costs in your user scenarios.
Conditional formatting is commonly used to hide part of a form from view until some condition is met. However, conditional formatting expressions fire every time the form is rendered whether hidden or not. Additionally, all of the data, calculations, and expressions used by fields that are conditionally hidden are still downloaded and can cause extra performance overhead. Thus if anything else invokes a postback, your form will incur this additional overhead. Similarly, if the conditional formatting requires a postback, then there is a postback each time the form is rendered. Although the primary reason for using conditional formatting has more to do with your form design and presentation, there may also be a slight increase in the responsiveness of the form if it uses conditional formatting. This is possible because it can be slower to render something in the browser than to execute the conditional formatting. However, if you need to substantially limit the presentation, consider using multiple views instead of conditional formatting because multiple views generally perform better.
For forms that are filled out in Internet Explorer, signing a form uses an ActiveX control that communicates with the server using out-of-band postbacks. The signature operation involves 3 main steps:
- Gather client-specific information to be used by the signature data
- Server preprocessing to prepare the stored XML document for signing
- Apply the final signature
Each step causes a postback, the first and the last being relatively expensive. This is because the first postback sends, among other things, the data corresponding to the non-repudiation image to be stored in the signature. The last is a full postback that refreshes the browser with the new signed document that has been prepared on the server.
Business logic event handling
Custom hosting pages and external page events
If you are using advanced form techniques such as developing your own ASP.NET page to host the InfoPath Form control, then there are a couple of things you should be aware of. Submitting to a custom hosting page requires full page postback because the XmlFormView control resides on the server and it involves rebuilding the page controls by ASP.NET. In addition, when submitting the form, before the submit is finalized, data is sent back to the browser with a pending submit action added. This pending submit action must then be sent back to the server in another postback to actually complete the submission. If the page is hosted on a SharePoint site, it also must initialize the SharePoint page infrastructure as well. Also, if an XmlFormView control is hosted on a custom ASPX page and the page listens to events from the control, then every operation triggering an event in the XmlFormView control will require full page postback.
In some advanced form design scenarios, you might build your own complex custom expressions or calculations using the "Edit XPath (advanced)" option. Although most common expressions or calculations can run in the browser, those that require more complex XPaths to evaluate may need to postback to the server. The following complex XPath conditions will cause extra postbacks:
More than one possible context for a filter
Complex XPath that is not used as a filter
XPath where the user input is not a filter
XPath filter or rule that references an element that is not in the current view
XPath filter that is position-based
XPath with inline functions
XPath that includes an unrecognized function
XPath that includes an extension function that is not supported by the browser
XPath has an inner expression
Expressions containing inner nodes
In our next article we’ll examine some of the conditions that make postbacks more expensive and that affect the performance and scalability of a server system running InfoPath Forms Services.