PNG image embedded inside xlsx file

Question

Hello,

I am experiencing DocViewer behavior where an XLSX file containing  a PNG image of a table is previewed as a selectable, copyable table. The table does not contain native Excel cells, and there isn't any other content in the file other than the image. It's a single embedded PNG image.

So I am wondering if the DocViewer performs OCR or text extraction on images embedded inside Office files (XLSX) during a preview render?

dbrace · Accepted Answer

Hello @Jerome,

Thank you for providing the Excel file. There is actually more to this than you initially said based on the design of this file. Let me briefly explain how I came to that conclusion in an outline format.

* I downloaded the Excel file that you shared with us and uploaded it into my Canvas Free for Teachers account. I do that because of the testing I do and it is a basic version of Canvas and not potentially altered by any customizations.
* When I did that, after zooming out so that I could see as much as possible, I saw what is attached in the PDF file of two-screenshots as "01 -- Canvas view of Excel file.pdf".
* I then opened the file in Excel on Mac and noticed that there was only one worksheet at the bottom. I confirmed that what was showing was a picture because I clicked on the table and was able to move it around. You can see this in the two-screenshots that are in the "02 -- Excel view of Excel file.pdf" PDF file that is attached.
* After doing that, I closed the file and did not save any of the changes and then re-opened it.
* I started to suspect that maybe the Excel file had a worksheet that was hidden and followed the Microsoft documentation in their "Hide or Unhide worksheets" support article at https://support.microsoft.com/en-us/office/hide-or-unhide-worksheets-69f2701a-21f5-4186-87d7-341a8cf53344 to see if that was the case and my suspicions were correct. After unhiding the worksheet, this worksheet is "normal" content that someone would enter into Excel and you can see that by looking at the formula bar when viewing cell A1 because it is text instead of empty. You can see that in the three-screenshots that are in the

Based on all of that, the experience in Canvas for this particular Excel file is because the LMS is ignoring the hidden worksheet and not because the LMS is doing any type of OCR.

-Doug

01 -- Canvas view of Excel file.pdf

02 -- Excel view of Excel file.pdf

03 -- Excel view of Excel file with unhidden worksheet.pdf

Jerome · Accepted Answer

Thanks a lot!

I totally missed the hidden sheet indeed.

It's interesting though that the hidden date column in the original file is written in Dutch while the DocViewer preview renders this in English (and in a slightly different format). Also in our Dutch language Canvas accounts!

Anyway, thanks for your help. Much appreciated!

dbrace · Accepted Answer

You are welcome, @Jerome.

Without know the "unpinnings" for how the Canvas DocViewer works, I believe that the DocViewer converted the raw data into a United States regional readable format.

For example using cell B3:

* when I "work some magic" and open the Excel file in an XML viewer, cell B3 contains only "45586"
* when type "45586" into a new Excel file and in an empty cell and then format the cell so that it appears as something like "Day of the Week, Month of the Year Number Day of the Month, Calendar Year", the cell now shows "Monday, October 21, 2024"
* however in Excel's formula bar, the previously entered 45586 now appears as (at least for me in the United States and for my - I believe mostly default - date settings) "10/21/2024"

Circling back, it seems that the DocViewer is taking that raw data in the Excel file and converting it to a United States regional format and ignoring any Canvas user account settings or localization/regionalization settings of a user's web browser or operating system or device. I do not know why the DocViewer decided on that particular format for a date+time format.

-Doug