Struggles with image compares

My last resort was to ping this forum but I can’t imagine I am alone here. Does anyone out there struggle with the problem if the image recgonition method used in Eggplant?

Here’s my issue. I have my “baseline” image that I have range selected around an area of the screen". A new build comes out and now it fails (twice becuase we do the test again if we fail once). I have no way of detecting the difference becuase the image captured from Eggplant is the whole screen. When I examine the image vs. what is one screen it looks good. Nothing is wrong but eggplant says there is but can’t show me why.

There is no good documentation out there that I know of the describe how it works. Yes I know it is based on color values but that isn’t helpful in disecting an failure. I need to be able to see the difference so I can justify updating my “baseline” image or not. Like I said I can’t imagine I am alone on this one. I wish there was a better solution and if there is please post it here. As of right now I have found none and my frustation for this product builds.



As you probably already know you can’t compare images in Eggplant. There has been a lot of interest in such a feature, so there is a good chance that such a feature will make it into a future release of Eggplant.

In the meantime, you are going to have to compare your images with another application. Someone had already posted directions on comparing images with Photoshop Elements here:

Your request has been noted as a feature request, so you will be notified if any progress on this feature is made.



We understand your frustration and have a few features in the works to help deal with these issues.

For now, the first thing is to be able to compare images. In addition to that forum post that Jo mentioned you can also review the article Using Eggplant: Comparing Images Using Third-Party Tools.

Beyond just comparing the images to determine what is different you should consider keeping the old and new states of the image and using the “Any” image commands and functions. Also, make sure you have the correct matching type for your image - marking it as text if it contains text which is being anti-aliased. This is the most common form of image failure.

For other issues feel free to send them in to support and we should be able to help you develop some strategies based on what’s changing.

Thank you both for your replies. Let me try to respond to both here in the same thread.

Jo the link you sent was helpful but not the solution. Any of the functions that I can use to get the coords of an image only work if the image can be found. If the image can’t be found then I am stuck. For example I have an old image that has being causing a failure with new builds. I don’t know how far back it goes from when it worked. I can’t easily use of the functions such as ImageLocation() or FoundImageLocation().

Johnathan, point taken. Most of the time I use tolerant on all images unless it is strickly text I am verifying. What’s not clear here is what is considered “text” in the world of eggplant. Since I am testing Flash content it is generated text not just word processor or os type text.


I’m not sure if the Flash rendered text is antialiased in the same dynamic way as the rest of the MacOS text.

We generally go with the rule to use the TEXT type on any MacOS image that contains characters (including at times symbol glyphs used for buttons, like the X, +, - in the upper-left corner of windows as you mouse over them).