There is no one way that is best for every situation.

You can store a deliminated string, but it's then harder to match against with SQL:

my $field = join(',', @checked_values);

But you then can't easily get the values out through SQL into a list -- or even search for items reliably. -- but if you only care about what the whole set is, rather than who selected item1, but not item3, this is just fine.

Alternately, so that you can more easily do text-searching with SQL through the field, you can pad the beginning and end with a comma:

my $field = join(',', '',@checked_values,'');

So that you can then use SQL to get:

SELECT * FROM tablename WHERE checkboxes LIKE '%,item1,%'

Without worring about dealing with items that are first or last in the list, or the possibility that one item may be a sub-string of another item.

But the full values of the checkboxes can take up a lot of space, so you can also key them, and assign a single letter to each value, and so, if you only had 10 items, you could number them 0 to 9, and just make the string from that. (you can easily get more than 60 values from lower alpha, upper alpha, and numbers) ... so we might store:

my %keys = ( a => 'item1', b => 'item2', c => 'item3' );
my %encode = reverse @keys;
my $field = join('', map { $encode{$_} } @checked_values);
[download]

For space reduction, you can also use positional coding (eg, a string '011000' to represent that 2 of 6 choices were selected ... but this is a pain to search on, and causes problems if you add more options later), or you can use bitmasks (eg, from the same one above, we have items in position 3 and 4 (0 being the rightmost position), so 2**3 + 2**4 = 8 + 16 = 24, so we can store '24'. It's efficient storage for dealing with registers in a computer, but as a database, I find it to be annoying to deal with.

Another option is to include single bit (boolean) columns for each option (is _very_ efficient storage in a database that can allocate fields that small, but it's best when the set of checkboxes is never going to change. It becomes less efficient when the set is exceptionally large, but the number selected is significantly small in relation to the size of the list.

For these situations, we either use one of the earlier mentioned items, or we store your data in a normalized format -- you remove the 'checkboxes' field from your table, and make a 'checkboxes' table instead, which has a record for each checkbox-(whatever the original table was) pair. You'd store the (original table)'s primary key, and the value of the checkbox -- for size considerations, you might store the checkboxes' potential values in a seperate table, and just store a number that keys to the checkbox table's primary id.

Which way is best? I have no idea. There are too many factors that you just haven't mentioned -- how often do we write new records? modify records? read the records? How do we read the records? (grab everything, selectively get records?) if selective, what are the selection criteria?

If you think about these sorts of issues, it can help you lay out your data in a way that will be useful to you.