02-01-2022 23:02 PM - last edited 01-18-2023 17:33 PM
I ran into some issues when pulling CSV data in from Power Automate Desktop because of commas in the actual data. I wanted to change the delimiter so I could more easily parse the data in a single Select action without commas in the actual data messing things up. I also may be parsing CSV files with hundreds or thousands of rows, so I didn’t want to use all my daily actions on this in a slow Apply to each loop.
Attached is the scope/flow I built so anyone can easily select their CSV data that has quotes around the comma-containing records, enter a new delimiter, and get the new delimiter separated data from the final compose action without the usual errors. And it only takes a few actions to do this, even on very large files.
I've found that many CSV files don't put quotes around their records with in-data commas, and this only works when there are quotes around those records. But if the file is saved as a text file, then it often puts quotes around the right records.
If you are using Power Automate Desktop, program the file to be saved as .txt and read that into your output variable.
It’s currently set to handle up to 50 comma-containing columns, but you can expand that to as many columns as needed by adding extra lines & expressions to the 1st Select action. Just follow the pattern by replacing some of the array numbers, like  with .
Also if your data has more unique values like an array with mixed quoted string data, Ex: ["string1", 2, 03/05/2022, "string3"], then this will create errors in the output.
The template for parsing CSV to JSON & entering it into a dataset uses the same change delimiter set-up: https://powerusers.microsoft.com/t5/Power-Automate-Cookbook/CSV-to-Dataset/m-p/1508191#M584
*Copying the template scope into another flow may not work as it may mess up the expressions. You may need to start with a copy of this template and copy & paste the rest of your flow into this template flow.
Version 3 Uploaded 03/26/2022 (Adjusted the 1st Select input so it can now also deal with in-data commas in the 1st column. Added more lines to the 1st Select so it can now handle up to 50 columns with commas in them.)
Version 4 Uploaded 04/09/2022
(More minor fixes & additions.
I adjusted several expressions so it can now handle a few more scenarios with arrays in the CSV data. It should handle any array that doesn't include double quotes and any array that is all strings with double quotes, so ["String1", "String2", "String3"], but it will have issues if it is a mixed array with some double-quoted strings and some other values, for example ["String", 4, 03/05/2022, "String2"] won't work.
I also adjusted how the LineBreak setting is set-up so it now uses the /r/n for the LineBreak. I also provided this link in the flow so anyone can look up the right string for the decodeUriComponent expression(s) if they happen to have different LineBreak characters. This change also made it possible to differentiate between in-data line-breaks and CSV row line-breaks on the files I tested, so it should now replace the in-data line-breaks, like the multiple-choice fields some sites use, with semi-colons. That should make those records much easier to deal with & parse in later actions.
I also looked over a problem with in-data trailing commas. I added a line in the settings where anyone can toggle whether they want it to adjust for trailing OR leading commas in the data, it just can't handle both in one dataset. So if one column in one row has ",String1 String2" and another column in another row has "String 3 String4," then it will have errors.)
More adjustments for comma edge cases and for instances with one comma-containing value following another in the file data.
Microsoft Power Platform & Paul Murana recently did a video demonstration of how to handle CSV & other files in dataflows: https://youtu.be/8IvHxRnwJ7Q
But it currently only outputs to a dataverse or Dataverse for teams table.
I figured out how to better format things in the Select statements & simplified the expressions in a second version.
It should help with some of those errors.
Thank you for the quick reply! Unfortunately I'm still getting the same issues but also the code looks exactly the same so perhaps you uploaded the old version? I checked all the actions and expressions and they look the same (from what I could tell) from your original version. Thanks again!
I just downloaded it from a different account and there should be obvious differences. Maybe there is some bug with updating download files on the forum.
Here is a copy to the new version in my Google Drive: https://drive.google.com/file/d/1zDJ2qKowBpQN2_BwB-VloumQnVnk8oUF/view?usp=sharing
Maybe that will work.
You can also download a different flow that will contain the same change delimiter set-up here: https://powerusers.microsoft.com/t5/Power-Automate-Cookbook/CSV-to-Dataset/m-p/1508191
This solution here is fantastic and saving me a ton of grief with importing CSV file reports from Ivanti.
Second version is also a huge help as I was also seeing the "/" issues.
My big question here is on guidance for tactics on how to scale it up.
Dealing with reports that have more than 10 columns and its not 100% clear to me on how the select commands in one row, feed in to the next.
I figured 10 <columns with quotes around them> was enough as any tests I did with CSV files only had the quotes around the comma-containing column values. So any individual row could have like 30 columns, and if commas only appeared in 9 or 10 of them, then it should still be okay.
But I definitely get the issue if your CSV files are putting quotes around every column value regardless of in-data commas.
Frankly, I could just copy and paste the 1 line of expressions in one of the Select statements 50 times, only changing the array item number in each, then no one should ever have an issue with it. I just didn't really know if that was necessary.