There are two modules that can quickly help you manipulate CSV files. The first is Text::CSV, and the second is its faster but possibly harder to install sibling, Text::CSV_XS. There is also another module which is pretty flexible, allowing you to select any character as the delimiter (not just commas): Text::xSV.

The rest of your question is a little trickier, and very much dependant on the nature of your CSV data.

It may be helpful to build sets that can be compared using List::Compare. You will have to decide what constitutes "an added row", versus "a modified row". Is there some unique key in each row that you can use to watch for new keys being added as an indication of new rows? That would be the most ideal situation. But if not, you'll have to decide what combination of fields will come together to create unique keys so that you can determine whether you're lookig at a modified row or a new row.

DBD::CSV is a good choice, but it only supports very basic SQL statement, otherwise he can see what has been deleted just by saying:

select foo, bar from table1
minus
select foo, bar from table2
[download]

And see what has been added by saying:

select foo, bar from table2
minus
select foo, bar from table1
[download]

Not just minus is not supported, even sub-query is not well supported, so he has to do more coding than he should.

Fascinating - I've never seen the minus keyword for SQL yet. Up to day, I've always used left outer joins to find out the missing parts between two sets of keys:

-- in table2 but not in table1:
SELECT r.id
FROM table2 AS r
LEFT JOIN table1 AS l
  ON r.id = l.id
WHERE
       r.system = ?
   and l.id IS NULL
   and r.id IS NOT NULL

-- in table1 but not in table2:
SELECT l.id
FROM table2 AS l
LEFT JOIN table1 AS r
  ON l.id = r.id
WHERE
       l.system = ?
   and l.id IS NOT NULL
   and r.id IS NULL
SQL
[download]

SQLite doesn't seem to document the minus keyword, so I'll stay with my (left) outer join solution, as SQLite doesn't support the full outer join. Of course, I haven't checked whether DBD::CSV supports outer joins, and I guess it doesn't jZed tells me it does, so these two SQL statements will be put to more and more use. How convenient if your whole business logic can be expressed in two SQL statements.

This smells like homework ...

"How is what you need any different from 'diff'?"

For example, you have a.csv:

a,1
b,2
[download]

and b.csv:

b,2
a,1
[download]

When you diff, they are differnt, but if you consider them as database tables, they are really the same. That is the difference.