Question

I am working on a parsing program that parses info from two files into a
sqlite3 table.

Lets say the table has the following values:
filename, value2, value3, value4

There are never more than two filename values in the table,
I would like to write a sql query that will join two rows when the following
conditions are true:

• row X:filename != row Y:filename
• row X:value2 == row Y:value2
• row X:value3 == row Y:value3
• row X:value4 == row Y:value4

---

Ok the actual program I am working on is a little more complicated so maybe this will be a little clearer

file1.txt contents

abcd, 1234, efgh
klmn, 5678, opqr
stuv, 9abc, wxyz

file2.txt contents

abcd, 1234, efgh
klmn, 9ffx, opqr
stuv, 9abc, wxyz

Desired Output:

• file1.txt, abcd, 1234, efgh file2.txt, abcd, 1234, efgh
• file1.txt, klmn, 5678, opqr, –, –, –, —
• –,–,–,–, file2.txt klmn, fffx, opqr
• file1.txt, stuv, 9abc, wxyz, file2.txt, stuv, 9abc, wxyz

Answer 1

x below is the table name

select a.*, b.*
from x as a
inner join x as b
  on a.filename<b.filename
    and a.value2 = b.value2
        and a.value3 = b.value3
        and a.value4 = b.value4
union all
select a.*, b.*
from x as a
left join x as b
  on a.filename!=b.filename
        and a.value2 = b.value2
        and a.value3 = b.value3
        and a.value4 = b.value4
where b.filename is null
  and a.filename = (select min(filename) from x)
union all
select b.*, a.*
from x as a
left join x as b
  on a.filename!=b.filename
        and a.value2 = b.value2
        and a.value3 = b.value3
        and a.value4 = b.value4
where b.filename is null
  and a.filename = (select max(filename) from x)

There are three parts

1. Where all values match between the two files. The lesser file name is listed on the left side
2. Where the values cannot be matched to the right side. min(filename) filters only for unmatched left-file rows.
3. Where the values cannot be matched to the left side. max(filename) filters only for unmatched right-file rows.

There is no way to produce your output order exactly (some inherent sort order) unless there exists some line number column, which can be spliced into the query.