9.1 Validation and Error Reporting Principles

There is nothing worse for a user than annoying, overly persistent, inaccurate, or uninformative validation. For example, error messages that describe an error but don't specify which field contains the error are difficult to correct. However, there is no recipe for balancing validation with system requirements: what is pleasing or mandated by requirements in one application might be annoying or useless in another. In this section, we consider practical validation models for web database applications.

Validation is actually two processes: finding errors and presenting error messages. Finding errors can be interactive, where data is checked as it's entered, or post-validation, where the data is checked after entry. Presenting errors can be field-by-field?where a new error message is presented to the user for each error found?or it can be batched, where all errors are presented as a single message. There are other dimensions to validation and error processing, such as the degree of error that is tolerated and the experience level of the user. However, considering only the basic processes, the choice of when to error-check and when to notify the user, leads to four common approaches:

Interactive validation with field-by-field errors

The data in each field is validated when the user exits or changes the field. If there is an error, the user is alerted to that error and may be required to fix the error before proceeding.

Interactive validation with batched errors

The data in all fields is validated when the user leaves one field. If there are one or more errors, the user is alerted to these, and can't proceed beyond the current page without fixing all errors.

Post-validation with field-by-field errors

The user first enters all data with no validation. The data is then checked and errors are reported for each field, one by one. The user fixes each error in turn and resubmits the data for revalidation.

Post-validation with batched errors

The user first enters all data with no validation. The data is then checked, and all errors in the data are reported in one message to the user. The user then fixes all errors and resubmits the data for revalidation.

In Chapter 8?without discussing the details?we covered several simple post-validation techniques to check whether mandatory form data was entered before inserting or updating data in the database. In addition, we used a batch reporting method, where errors were reported as a list by constructing an error page using a template.

In the examples in this chapter, we discuss additional validation techniques to inspect both mandatory and optional fields. We use these techniques to create a batch error report in Chapter 10. Examples of complete validation code for a customer details form are listed in Chapter 17.

9.1.1 Models That Don't Work

Interactive models are difficult to implement in the web environment. Server-side scripts are impractical for this task, because an HTTP request and response is required to validate each field that's entered. This is usually unacceptable, because the user is required to submit the data after entering each field. The result is that response times are likely to be slow and the server load high.

Client-side scripts can implement an interactive model. However, validation on the client side should not be the only method of validation because the user can passively or actively bypass the client-side processes. We discuss the partially interactive solution of including client-side scripts with an HTML form later in this chapter.

9.1.2 Models That Do Work

Post-validation models are practical in web database applications. Both client- and server-side scripts can validate all form data during the submission process.

In many applications, reasonably comprehensive validation is performed on the client side when the user clicks the form submit button. Client-side validation reduces server and network load, because the user's browser ensures the data is valid prior to the HTTP request. Client-side validation is also usually faster for the user.

If client-side validation succeeds, data is submitted to the server and the same (or often more comprehensive) validation is performed. Duplicating client validation on the server is essential because of the unreliability of client-side scripts and lack of control over the client environment.

The post-validation model can be combined with either field-by-field or batch error reporting. For server-side validation, the batch model is preferable to a field-by-field implementation, as the latter approach has more overhead and is usually slower because each form error requires an additional HTTP request and response.

For client-side post-validation, either error-reporting model can be used. The advantage of the field-by-field model is that it leads the user through the process of correcting the data and the cursor can be directed to the field containing the error, making error correction easier. The disadvantage is that several errors require several error messages, and this can be frustrating for the user. The advantage of the batch approach is that all errors are presented in one message but the disadvantage is that the cursor can't easily be directed to the field requiring correction and its sometimes unclear to the user how to correct the data.

Server-side validation is essential to secure a web database and to ensure that system and DBMS constraints are met.

Client-side validation may be implemented in addition to server-side validation, but all client-side functionality should be duplicated at the server side. Never trust the user or the client browser.

The choice of which reporting model to use depends on the size and complexity of the form and on the system requirements.