image comparison fails

I just done simple test on https://launchpad.net/sikuli with to see how recorded image is recognized. it works fine with exception that it is relatively slow on my 21" widescreen.

But in general i think the approach of validating if web site is loaded is not correct. I would suggest following

-- On the site homepage take 2 - 3 major elements that should be present. They may be ones that you will use afterwards anyway. Search for them

-- Copy the whole content of the page by clicking in the middle and doing ctrl-a-c. The using the power of python check that some expected texts are there and some unexpected are not there(like "exception" and "error")

I believe that above approach is more reliable since it will tolerate significant changes in website layout.

Hope this helps

does blank_page.png mean a totally white (or uni-color) area?

If yes: this is problematic (but should work with latest version r930).

Another thing:

Look into screen region and test if blank_page.png (1021x672 pixel) exists.

    #screen region // screen size ipad
    screenRegion = Region(1,43,1020,804)

Since the image is 1 pixel wider than the region, this will never be found.

Thanks for your replies,

@dinev
I thought about taking visual elements of the website and validate against them, but I'm looking for a more general approach so I can easily change the tested websites without touching the sikuli script.

@raiman
You're right, by blank page I mean a totally white (uni-color) are, e.g. you type about:blank in your browser's address bar and hit enter. I'm going to give r930 a try.

What do you think? Which approach should I take? Which is one is more stable in terms of correct image comparision results?

thanks in advance

regards

--- but I'm looking for a more general approach so I can easily change the tested websites without touching the sikuli script.
Generally this is a good idea, but my (and other people's) experiences tell me, that this is usually combined with a huge scripting effort.
I think each Sikuli script will have some specifics about the images used (different browsers, resolutions, systems, ... ) that lead to different versions of the same script.
The only thing one should do: identify often used workflows and turn them in classes/methods/functions, that can be reused.

--- Which is one is more stable in terms of correct image comparision results?
I agree (and usually use) dinev's approach: looking for some key visuals, that should be there, to be sure, the workflow might proceed.
You might put this check in a function or functions, that accept some parameters (list of images, expected regions, combination logic, ...)

examples:

checkAll(images, reg = SCREEN):
    ret = True
    for image in images:
        if not reg.exists(image, 0):
            ret = False
            break
    return ret

checkSome(images, reg = SCREEN):
    ret = False
    for image in images:
        if reg.exists(image, 0):
            ret = True
            break
    return ret

usage:

browser = App("your browser")
if not checkAll( ("image1.png", "image2.png", "image3.png" ), browserApp.window() ):
     print "makes no sense to go on"; exit(1)

You could even use this approach, to check wether some white area is contained in the browser window.