Question

I have two tables in Sql Server, one containing IDs for files and the slides contained in those original files, and another for “sections” that can contain slides from one or more of the files, potentially in arbitrary order, duplicated, and/or with some slides eliminated.

Sample data looks like this:

FileSlide

FileID       SlideID
214          716
214          717
214          718
223          770
223          771
223          772
223          773
223          774
223          775

SectionSlide

SectionID    SlideID
527          716
527          718
527          717
527          770
527          773
527          774
527          775
527          774

I originally didn’t need a “SectionFile” relation, but now I do need that information to see which files were chosen for a particular section, regardless of slide details. My problem is examining the slide IDs between the SectionSlide and FileSlide tables to see whether there’s an overlap between the slides in any given File-Section pair. I would like to find all File-Section pairs that share slides.

For the sample data above, output would look like this:

SectionFileCandidates

SectionID    FileID
527          214
527          223

What is the query to produce this output?

Is it possible to calculate a metric that indicates what proportion of the original file’s slides exists in the section?

For the sample data above, output would look like this:

SectionFileCandidates

SectionID    FileID    Overlap
527          214       1.00
527          223       0.67

…that is, 3 out of 3 slides from file 214 are in section 527, and 4 out of 6 slides from file 223 are in section 527.

I was originally trying to compare groups of rows using the OVER (PARTITION BY ...) clause, but could not figure it out.

How can I do these two queries?

Answer 1

Both queries are possible!

---

First query:

SELECT s.SectionID,
       f.FileID
FROM SectionSlide s
INNER JOIN FileSlide f ON s.SlideID = f.SlideID
GROUP BY s.SectionID, f.FileID

or

SELECT DISTINCT s.SectionID,
                f.FileID
FROM SectionSlide s
INNER JOIN FileSlide f ON s.SlideID = f.SlideID

---

Second query:

select s.SectionID, f.FileID,
       round(((count(distinct f.SlideID)*1.0) / aux.total), 2) as 'Overlap'
from SectionSlide s
inner join FileSlide f on f.SlideID = s.SlideID
inner join (select f.FileID, count(f.SlideID) as 'total'
            from FileSlide f
            group by f.FileID) aux on aux.FileID = f.FileID
group by f.FileID, s.SectionID, aux.total