OCRExtractRelative selects wrong area

[Michelp76]

Hello everyone, I’m pretty new to UI.Vision although I did my homework and searched the forum thoroughly already.

OCRExtractRelative always fails to scrap any values in our PDF
Here is the green/pink mask :

(The two bold values are the ones I’m interested in)
Here is the csv output :

Nothing is scraped as you can see

So, obviously I tried to make the green box more or less big
I also tried to tweak the size of the pink box

But it always results in an empty output no matter what.
I also played around the variables !OCRScale and isTable activated, or !OCREngine set to 2, but still no luck

Here is the __lastscreenshot.png :

Any recommendation ?
Thanks a lot in advance

Now that I think about it, is __lastscreenshot.png the area that OCRExtractRelative captures ?
Well it very much seems like it’s the wrong lower part portion of the page.
So the green box would be the culprit ?

[Michelp76]

Also tried narrowing down both boxes like so :

(Also switched the OCR language to “French”)

Here is the __lastscreenshot.png :

I’m pretty sure the spot it tries to capture is way below the target (pink box)
I wonder what’s wrong

[Michelp76]

It could be helpful to add that my PDF is actually full text, so if I understand this right :

Create Searchable PDF - OCR text appended instead of replacing

This sounds like the PDF document that you want to OCR has already the text as text inside, or? In this case no OCR is required, as the PDF is already searchable and you can get text from pdf directly. OCR is helpful if the text is provided as image inside a PDF, as it is typically for PDF from scans.

I might not need OCR commands, as my PDF is already searchable.

That being said I have no clue how to make a search inside my PDF.
Something like storeText with an xpath say, “xpath=(//*[text()[contains(.,'My Keyword… ')]])” ?

I’ll try that

EDIT : doesn’t seem to work but I forgot to mention that I’m using an embedded Acrobat Reader inside an iframe :

That don’t seem to work well with Web scraping obviously.
So I guess OCR is the only way to go…

Provided it works !

[ulrich]

@Michelp76

It could be helpful to add that my PDF is actually full text

In this case I would recommend you use XTYPE and simulate the “copy & paste” commands with CTRL+A (select all text in PDF) and then CTRL+C. Then, inside the macro, the data is in the !clipboard variable.

You can also try to use a triple-click to just select the line you need:

• XCLICK | | #tripleclick
• XType | ${!KEY_CTRL+KEY_C}

As mentioned below, you might need to switch to desktop automation, if the coordinates are wrong in the web automation mode.

PS: storeText works only on websites, not PDF.

PPS: This answer is a solution without OCR. The next answer is an idea to fix the OCR capture. This way you have 2 suggestions to try

[ulrich]

Now that I think about it, is __lastscreenshot.png the area that OCRExtractRelative captures ?

Yes!

…(then) I’m pretty sure the spot it tries to capture is way below the target (pink box)

I agree. I assume the iframe and the embedded PDF confuse the BROWSER coordinates calculation of UI Vision.

=> For OCR to find the correct area, switch to desktop automation mode, then it will work! You can either switch in the UI Vision IDE or use this macro command: XDesktopAutomation | true

[Michelp76]

In this case I would recommend you use XTYPE and simulate the “copy & paste” commands with CTRL+A (select all text in PDF) and then CTRL+C. Then, inside the macro, the data is in the !clipboard variable.

Thanks, that did the trick !
Now I’m on my way to parse the !clipboard content