eValid™ -- Automated Web Quality Solution Browser-Based, Client-Side, Functional Testing & Validation, Load & Performance Tuning, Page Timing, Website Analysis, and Rich Internet Application Monitoring.

Summary
This page provides some basic technical notes about eValid support for validating images of various kinds. This material is important because there is increasing need for functional testing (and performance analysis) of websites that deliver graphical information.

With some care, images can be validated very effectively so that the systems that rely on them can have automated regression tests that are effective at discovering errors. Examples of success in the mapping area include map-generated engines such as Google Maps, MapQuest, and include applications that are built on them such as NextBus.

Recommendations
Here are recommendations that concern applying eValid to validating images of all kinds.

• Image Synchronization
  The Validate and Synchronize on Screen Rectangle (VSSR) command [Ctrl+Q] meets most requirements for validating an image -- and for enhancing synchronization on it if required.

  VSSR saves a checksum extracted from the image at record time and saves the checksum in the script file. On playback eValid continually recalculates the checksum present within the saved image coordinates (in the part of the desktop you selected at record time) until a match against the saved checksum is found or until the synchronization timeout time is reached.

  Caution: In some cases the "blinking outline" feature causes the synchronization to fail, so the safest method is to leave this feature disabled. (Think of the blinking image as a crowd pleasing heuristic.)

• Dynamic Data Dependency
  Whenever the display you are seeing has a temporal component very likely you will NOT be able to get VSSR to work. The original image checksum values will "never" be repeated.

  However, if an images has a fixed number of repetitions you may achieve VSSR-type synchronization, but not necessarily on the first instance of the correct image being displayed (due to the client side time delays involved).

• Games
  Because modern PC-based games are largely based on simulations it will be unwise to try to do regression tests on them, even if you can assure that you have a precisely reproducible initial state. Game simulation activity and time variations will nearly always produce non-identical results from playback to playback.

• MapQuest Maps
  These images seem to be constant, but even being off by 1 pixel in the x-direction or y-direction will prevent reliable playback (will not sync).

• Google Maps
  These maps are DOM based sections of the image (you can see the delineations with the PageMap feedback feature), but with regions in a complex table individually filled in (using AJAX technology), based on user inputs and details of the state of the browser display.

  You should be able to get good reproducibility with Google Maps, assuming the database from which the images are taken has not been updated.

• MouseOver Popups
  If you can highlight the text in the popup then Validate and Synchronize on Test String (VSTS) or even Validate Selected Text (VST) will do the job.

  In the most common case, you need to VSSR, but one problem is: how do you get the popup to occur repeatedly. The answer is to do an absolute xyClick type command at the spot on the display that causes the popup. Then you can VSSR the popup without difficulty assuming it is persistent enough.

• Drag and Drop
  A recorded Drag and Drop is reliable, but not if the underlying picture is variable. You may drag and drop a picture that is in the wrong place a precise distance away from where it was only to THEN find out that it was not stable in the first place.

  What will appear to be a failure of the validation is in fact due to a problem with the initial state.

• Movies, Videos
  Not only will there by timing problems even if you do catch the same frame (possibly by single-framing up to a particular point) there is no guarantee that client-side processing has not changed the image so that the checksum with a prior version would be identical.

  The starting frame and the ending frame usually are stable/static and given that the control buttons on most players are objects in the DOM there is some hope for start/end movie confirmation. Sadly, however, there is almost no hope for any kind of "in the middle of the movie" validation of images or movie-frames within the sequence unless you have the capability to stop the move playback at a specific movie-frame number.