Any suggestions

Use a real database. Track changes to parts and stock levels in an audit table and generate reports by dumping the current period's changes from the audit table.

... without making this program any longer and hard to trace.

We can't see "this program", or even the bits relevant to the question. As described you should only need half a dozen lines of code to compare the two files. Maybe there is important stuff you forgot to describe that makes the job more interesting?

G'day edge99off,

Welcome to the Monastery.

Unfortunately, a prosaic description, such as you have here, doesn't provide us with sufficient information to offer much help.

What we really need to see is: a minimal piece of code that's shows what you're currently doing; a small, yet representative, sample of your data; what your current results are and how they differ from what you want; and, possibly any messages (e.g. warnings or errors) that you're currently receiving. You'll find details of this in "How do I post a question effectively?".

On the basis of what you've posted, here's some potential help.

Use the Text::CSV module to parse your CSV files.

"REGEX or pattern matching" doesn't make a lot of sense (at least, not in the context you've used here). You use a regex for pattern matching; not as an alternative. Maybe, when you post some code, your meaning will become clearer. Perhaps take a look at "perlintro: Regular expressions" and follow the links therein for more information.

Reading a file will not damage it. Parsing its data will not damage the file. The way you open your files could be an issue. Again, this is another example where seeing your code will clear this up for us.

As the others already stated, do not parse CSV with regexes. Ever. If your data is extremely simple (and guaranteed to stay simple), a split might be the way to go, but usually simple CSV doesn't stay simple and embedded newlines, separation characters or quotation will lead into unmaintainable code. Stick with Text::CSV_XS (or Text::CSV if you don't care about the speed) for parsing CSV.

The Text::CSV_XS distribution has an examples section that hosts the csvdiff script, which might be exactly what you are looking for. Feel free to get that and amend it to your needs.

Otherwise, we don't have enough details about your procedure and your data, but, in general, comparing 20,000 lines in less than 2 minutes seems to be a very realistic aim (if coded reasonably efficiently). I am quite often comparing 30 million lines in 10 to 15 minutes or even less if the comparison to be performed is simple or the lines relatively short, on a platform which is far from being a racing horse.

Finally, as already pointed out, if you open your files in read mode, there is no danger to alter them. But show your code to confirm this as well as my previous (quite general) comments.

If you only need to know what changed, why not use first Comm -23 NEW.csv OLD.csv.
This way you know which identifiers to check, and skip the ones that are identical.