Regular expressions

Tue Nov 3 22:12:08 EST 2015

On 2015年11月03日 19:04, Michael Torrie wrote:
> Grep can use regular expressions (and I do so with it regularly),
> but it's default mode is certainly not regular expressions, and it
> is still very powerful.

I suspect you're thinking of `fgrep` (AKA "grep -F") which uses fixed
strings rather than regular expressions. By default, `grep` certainly
does use regular expressions:
 tim at linux$ seq 5 | grep "1*"
 tim at bsd$ jot 5 | grep "1*"
will output the entire input, not just lines containing a "1"
followed by an asterisk.
> I've never used regular expressions in a database query language;
> until this moment I didn't know any supported such things in their
> queries. Good to know. How you would index for regular
> expressions in queries I don't know.

At least PostgreSQL allows for creating indexes on a particular
regular expression. E.g. (shooting from the hip so I might have
missed something):
 CREATE TABLE contacts (
 -- ...
 phonenumber VARCHAR(15),
 -- ...
 )
 CREATE INDEX contacts_just_phone_digits_idx
 ON contacts((regexp_replace(phonenumber, '[^0-9]', '')));
 INSERT INTO contacts(..., phonenumber, ...)
 VALUES (..., '800-555-1212', ...)
 SELECT *
 FROM contacts
 WHERE -- should use contacts_just_phone_digits_idx
 regexp_replace(phonenumber, '[^0-9]', '') = '8005551212';
It's not as helpful as one might hope because you're stuck using a
fixed regexp rather than an arbitrary regexp, but if you have a
particular regexp you search for frequently, you can index it.
Otherwise, you'd be doing full table-scans (or at least a full scan
of whatever subset the active non-regexp'ed index yields) which can
be pretty killer on performance.
You'd have to research on other DB engines.
-tkc