Results from SQL query using subquery is blank

[kokitabayashi]

I'm currently running SQL command using Excel ODBC Driver and trying to run layers of JOIN statements using subquery but facing an issue.

I ran the same SQL query on My SQL DB via MySQL Workbench and that one worked as expected but, the same query with the same data running on PAD with Excel ODBC Driver gives me different (and unexpected) results.

What I'm trying to do overall: Join two tables trying multiple columns one by one using subqueries.

High level steps (my question is a subset of the step 1 and 2):

1. Join TBL_1 and TBL_2 with column 'A' => Keep the matched result (will call Res_X here) and unmatched result (will call Res_D here)

2. Join TBL_1 and Res_D (unmatched result from step 1) with column 'B' => Keep the matched result (will call Res_Z here) and unmatched result

3. Union ALL Res_X (matched result from step 1) and Res_Z (matched result from step 2)

4. Do the same operation using column 'C' and Union All the matched results (continue total four times)

Following is the test tables I'm using:

[TBL_1]

COL_A	COL_B
100	1
101	2
102	3
103	5
104	6
105	7
106	8

[TBL_2]

COL_C	COL_D
100	A
101	B
102	C
5	D
601	E
701	F
10AB	XXX801YYY
11ABC	XXX801YYY

[Expected result (run query with unmatched result which is a subset of what I'm trying)]

c.COL_A	c.COL_B	COL_C	COL_D	d.COL_A	d.COL_A
		5	D	103	5
		601	E
		701	F
		10AB	XXX801YYY
		11ABC	XXX801YYY

[Connection String]

As some columns have mixed data, I put "IMEX=1" in the connection string.

[SQL Query]

As some columns have mixed data, ON c.[COL_C] = d.[COL_B] returned an error. Because of it, I used LIKE without any wildcard to join two tables.

[Result from the SQL Query]

As you can see, the first JOIN statement is working and returning 5 rows but the second JOIN statement which is connecting the results from the 1st JOIN and the original table 1 is returning blank.

When I run this on MySQL via MySQL Workbench, it returns value for d.COL_A and d.COL_B as expected (103 and 5).

Following are some experiments I also did to specify where the issue is coming from.

If I run the subquery separately, the result looks like this (looks fine).

[Sub query only]

[Result when running sub query only]

If I just join the two tables using the soma column without using subquery, the result looks like this (looks fine).

[Query which is just joining two tables]

[Result from running the query which just join two tables]

If I flip the order of the original query it gave me different result (which is also confusing)

[Flip the order]

[Result from flipped SQL]

Based on the original result, result from MySQL and additional experiments, I'm assuming the issue is coming from how the subquery works on PAD & Excel ODBC combination.

Would be great if anyone could provide me an insight on this!

Thanks!

[ZS440 · Accepted solution]

Hi.

@kokitabayashi 

Sorry I can't help you, but I have verified this.

In my environment, your Query works fine.

If there is any difference from my environment, you may have a clue.

Windows 10 Pro 21H2

Microsoft Excel for Microsoft 365 MSO (Version2207 Build 16.0.15427.20182) 64 bit 

Power Automate for desktop Version 2.22.263.222.02 Installer Edition

Underscores in column names are omitted, but it works fine.

In my opinion, it is better not to use underscores(_) in column names.

Because when using the Microsoft Office Access Database Engine OLE DB Provider, underscores act as ANSI SQL wildcards.

https://docs.microsoft.com/en-us/office/client-developer/access/desktop-database-reference/compariso...

I can use "=" instead of "LIKE" if I do not use IMEX=1 for Extend Properties and the Excel data source is explicitly formatted as text type.

I hope it works out for you.

[kokitabayashi]

Thank you so much for your reply, @ZS440 

This is indeed, really helpful!

The following was the environment I was using when I originally posted.

** Original Test (Test-A) **

[Test Environment-A]

Windows 10 Enterprise 20H2

Microsoft Excel for Microsoft 365 MSO (Version 2108 Build 16.0.14326.21052) 64-bit

PAD Version 2.16.215.22020

[SQL & Connector]

(Refer to the original post)

[Result in the Test Environment-A (as posted in the original post)]

Based on your reply, I did the following additional experiments and got interesting results.

** Additional Test (Test-B) **

[Test Environment-B]

Windows 10 Pro 21H2

Microsoft Excel for Microsoft 365 MSO (Version 2207 Build 16.0.15427.20182) 64-bit

PAD Version 2.22.263.22202

[SQL & Connector]

(Same with the original one: Refer to the original post)

[Result in the Test Environment-B]

** Additional Test (Test-C) **

[Test Environment-C]

Windows 10 Pro 21H2
Microsoft Excel for Microsoft 365 MSO (Version 2207 Build 16.0.15427.20182) 64-bit
PAD Version 2.21.244.22174

[SQL & Connector]

(Same with the original one: Refer to the original post)

[Result in the Test Environment-C]

As you can see, in the two additional environments (Test-B and C), I got the same results as yours (which was what I was originally expecting).

For these additional tests, I created a new bot and new excel file from scratch in the original environment (Test-A) as well, just in case, but the result was the same ("d.COL_A" and "d.COL_B" are blank).

It seems like not only PAD version is different but also Windows 10 Pro version and Excel versions are also different; therefore, I cannot conclude which exactly software version difference is causing the issue, but at this moment, considering the maturity of the software and the number of users of it, I'm guessing it may be a bug in PAD version 2.16.215.22020.

Also, thank you for sharing your opinion on using the underscore in column names and also sharing the results of trying = without IMEX=1 instead of LIKE with IMEX=1.

I didn't know about what you mentioned about the underscore with ODBC and it sounds like I should stop using it whenever I have an option. It is great to know that.

For the use of = instead of LIKE without IMEX=1, I tried as the following but, it seemed like some data were dropped.

** Experiment of = without IMEX=1 **

[Connector]

[SQL replaced LIKE with =]

[Result (from Environment-A)]

[Result (from Environment-B)]

As you can see, in the both environments, some data from COL_C were dropped. As I still need these data in the later steps, I tried to figure out the solution and came up with using LIKE with IMEX=1, but if you see any risk doing this way (or any other better way to do it), it would be great if I could hear it!

Thank you!

[ZS440 · Accepted solution]

Hi @kokitabayashi 

Thank you for sharing your various verifications and results with us.
Thank you too.

However, I believe we need to be careful about IMEX=1.
IMEX=1 is telling the driver to read the data as text, but we are having problems.
I think we can confirm this by trying the following mixed data.

If there are more than 8 numerical values in a row and then text values are mixed, reading will fail and result in null, even though IMEX=1.

Therefore, I think it would be safe to pre-cleanse the data source side (EXCEL in this case) to text type data.

No use the IMEX=1 option.

Similarly, by data formatting all columns to text type, your query can be executed using "=" without the IMEX option.

If you do not want EXCEL to format your data, you can also explicitly enable IMEX=1 by setting HDR=NO. This will cause the first row of data to be the original column name and the entire column to be text type as well. (If the column name is a text type)

I hope this helps you or anyone else.

[kokitabayashi]

Hi @ZS440 ,

Thank you for sharing your insights and pointing out the potential risk relying on IMEX=1!

I tried what you explained about IMEX=1 on my end and was able to recreate the issue ("1hundred8" and "ten" disappear), as well. This is actually, an interesting and also a big problem.

Also, the solution you suggested (convert the columns to Text format, stop using IMEX=1, and Join on '=') worked great, as well.

I added a "Run VBScript" action for changing the columns' format to Text before running the SQL statement and everything is working great right now.

'Opens the Excel file'
Set objExcel = CreateObject("Excel.Application")
Set objWorkbook = objExcel.Workbooks.Open("%ExcelFile%")
objExcel.Application.Visible = False

'Selects the sheet 1'
Set objSheet = objWorkbook.Sheets("%SheetName%")

'Change the number format of the row to Text'
for col=1 to 2
objSheet.columns(col).NumberFormat = "@"
next

'Selects the sheet 2'
Set objSheet = objWorkbook.Sheets("%SheetName2%")

'Change the number format of the row to Text'
for col=1 to 2
objSheet.columns(col).NumberFormat = "@"
next

'Saves and closes the Excel file'
objWorkbook.Save
objWorkbook.Close SaveChanges = True
objExcel.Application.Quit
Set objExcel = Nothing

Thanks!

[ZS440]

Hi. @kokitabayashi 

I'm glad it worked out for you.

I have learned a lot from you as well.
Thanks for sharing your great idea of using VBScript to change the formatting.