Can Tools Reduce Problem-Solving Time?

A Review of ParaSoft's CodeWizard for Java™ 1.0 and 1.1

By Benjamin Long

Time. We all have the same amount, yet no one ever seems to have enough. We start a task that we think is simple-one that shouldn't take much time to do-and hours and hours or days later we are frustrated because it is still not finished.

This scenario plagues professional developers every day. Given an algorithm or patch of code that seems simple to implement, they go at it with gusto and confidence, believing that they will be done in no time. Yet somehow they manage to get stuck on simple, hard-to-find problems, such as overlooking an uninitialized variable or hidden inherited instance variables. Some might criticize these developers, arguing that careful, meticulous work will prevent these difficulties; however, in practice, these things can happen to anyone.

ParaSoft has developed a line of tools to help programmers find and diagnose coding errors quickly. I will review CodeWizard for Java, versions 1.0 and 1.1. Version 1.1 was just released "hot off the compiler," so to speak. I will touch on some of the main similarities and differences between the two.

CodeWizard for Java™ (CWJ) is not a compiler, but a source-code analysis tool. Some well-known books on source-code analysis are Effective C++ and More Effective C++, by Scott Meyers. Philosophically speaking, CWJ is essentially Scott Meyers in a tool for Java. It parses the source code identifying syntax errors, possible logic errors, and violations of the best current programming practices by using a predefined set of stylistic, design, and coding heuristics (see Figure 3.)

Code Wizard for Java™ can be installed on Windows 95 or Windows NT, or Unix platforms. It can be run from the DOS command line, Unix command line, or from the Microsoft Visual Studio environment. It integrates very nicely into the "Tools" pull-down menu in Developer Studio. CWJ works in conjunction with a control panel and Insra, a GUI display.

The control panel in 1.0 is a set of four-tabbed sections: Analysis, Suppressions, Password, and About. Version 1.1 introduced a control panel that is also composed of four tabbed sections. In place of the Suppressions tab, a new tabbed page called Naming Conventions is shown. This tab allows project developers and managers the flexibility to standardize coding styles and conventions for things like classes, instance variables, and methods. The logic of the Suppressions tab from 1.0 is accessed via a button entitled "Suppressions," on 1.1's Analysis tab.

Conventions Tab

Insra displays error messages and feedback from CWJ-analyzed source file. With Insra, the programmer is presented with several levels of functionality. It provides displaying and sorting options for error messages, error suppression capabilities, report saving and loading capabilities, and application editing options.

CWJ's primary purpose is to act like a microscopic filter that catches all manner or errors and even minor inaccuracies that have been overlooked by the compiler. It can sit on the front end of the compiler or function stand-alone, giving developers immediate, meaningful feedback.

The error messages are composed of the class, the specific item that is violated, the severity level or error category of the message, the source file name, and the line number where the error occurred. The messages also include the code responsible for the violation.

Messages are placed in the following error categories: Informational (I), Possible Violation (PV), Possible Severe Violation (PSV), or Violation (V). Examples of these are shown in Figure 3.

With regard to the item violated, CWJ is designed to detect errors resulting from one of 23 various items, or heuristics, of good programming practice. Version 1.1 upped the ante to 41 items that are checked (see Fig. 3). The reason 1.1 has more rules is that Java™ is still being developed and new rules are discovered as this development progresses. Items that are violated in CWJ may be viewed by double-clicking on the error message header listed in Insra

Insra also allows the developer the functionality of double-clicking on the error message to open the source file where the error occurred. The editor that displays the file may be specified easily in the setup configuration of CWJ. The error may be fixed and the file may be resaved on the spot. The developer can then reanalyze the file immediately to observe the impact of the changes.

Because CWJ parses for many types of errors, it sometimes works almost too well for developers; identifying too many "informational" errors that are more stylistic in nature than application-critical. Insra, in version 1.0, allows the developer to suppress error messages according to error category. However, the developer must go to the control panel to suppress error messages according to the other parts of an error message: item, class, file.

ParaSoft fixed this inconvenience in version 1.1. Developers can now access the CWJ Suppressions Control Panel either directly from Insra, or from the main control panel. The newest implementation of the Suppressions section is easier to use as well. ParaSoft added several drop-down boxes and buttons that weren't available in 1.0. This is an improvement over 1.0, where the programmer needed to use more keystrokes to utilize all the available functions.

Insra sorts messages according to the error category, by file name, or chronologically by the time when each message was generated. Neither version displays the time, even though both can sort "by time," so this function may not be as useful as it should be. The different sorting options can be very nice when the programmer has many error messages to deal with. However, this sorting is limited only to the error message display.

Error messages can be saved in an ASCII report file for further evaluation, processing, and storage. Report files may also be loaded back into Insra's active display for evaluation and editing. Even if the messages were sorted while they were in the display, they will not be sorted in the report file. This is true for both versions of the product.

In addition, the report file contains a lot of redundant information and is hard to read at a glance. Developers at ParaSoft tell me they will consider the possibility of making the report files sortable and configurable to a user's preferences in future releases.

The transition from 1.0 to 1.1 is very easy to make. The interface is more functional and user-friendly. The product is also more functional. The main selling point of CodeWizard for Java™ is that it will save developers time because it catches errors more quickly than developers had previously experienced.

This review appeared in the Product Review section of the July 1998 edition of Component Strategies Magazine.