Comments on: (How|Where) does one find Google's (guide|help) on (searching|querying) using regular expressions?

Question: (How|Where) does one find Google's (guide|help) on (searching|querying) using regular expressions?

Gyan — Thu, 28 Jul 2005 04:34:41 -0800

Well, what do you know, err.. what do I know, Google does support some sort of regular expressions. However, I can't find the usage guide. Anyone?

By: jessamyn

jessamyn — Thu, 28 Jul 2005 07:28:35 -0800

Here's a previous thread on a similar but not identical topic.

By: Gyan

Gyan — Thu, 28 Jul 2005 07:48:35 -0800

I know, I was going to post a comment there, but it's locked.

By: mendel

mendel — Thu, 28 Jul 2005 07:59:04 -0800

It's a big jump from "supports permutations" to "supports regular expressions" -- just like it supporting "*" doesn't mean it supports shell globbing. The "(a|b|c)" notation seems to be identical to the documented "(a OR b OR c)" notation.

By: majick

majick — Thu, 28 Jul 2005 08:04:51 -0800

I've played quite a bit with search syntax but, no, Google does not support some sort of regular expressions. Perhaps you've been misled as to what a regular expression is?

By: Netzapper

Netzapper — Thu, 28 Jul 2005 08:09:45 -0800

Check out GoogleTutor.

By: Gyan

Gyan — Thu, 28 Jul 2005 08:24:13 -0800

mendel : "It's a big jump from 'supports permutations' to 'supports regular expressions'"

I know, that's why I said 'some sort'. Clearly, escape sequences don't work. I know Google can support regex, because my friend who's starting work there in August, has seen it on-campus.

By: robhuddles

robhuddles — Thu, 28 Jul 2005 08:24:48 -0800

Also see GoogleGuide

By: Rhomboid

Rhomboid — Thu, 28 Jul 2005 09:36:07 -0800

Just because it's something their internal interface supports doesn't mean you'll ever see it available on the external interface. The fact is that plain "keyword" searching is computationally about a million times easier to implement on the google scale than a regular expression DFA / NFA engine. The only possible way to make keyword searching efficient over hundreds of terabytes (or whatever their index is up to these days) is to precompute an index of words.

In fact a full regex engine is turing-complete, and you can write arbitrary regexps that will gobble up near infinite amounts of CPU time and memory. For all these reasons it would be technical insanity for them to offer regex searching to the general public.

By: grouse

grouse — Thu, 28 Jul 2005 12:15:58 -0800

Since you guys want to be so nitpicky about what a regular expression is, it's an expression that defines a regular language. Even a single character is a regular expression. You don't have to have wildcards to have a regular expression. So in some sense, Google has always supported "some sort of" regular expressions. In another, they do not support the full expressiveness of regular expressions.

In fact a full regex engine is turing-complete

Perl-compatible "regexes" have features that mean a Turing machine is necessary to accept them, but properly speaking the "regular" in regular expressions means that you can accept them with only a finite state automaton.

By: andrew cooke

andrew cooke — Thu, 28 Jul 2005 14:00:30 -0800

ok, it's late, but it seems to me that the best description of google is that it "supports regular expressions over the alphabet consisting of words in web pages except for the implementation of a kleene star".

gyan's example shows both concatenation and alternation, as long as you remember that the alphabet is words, not letters.

wikipedia

By: andrew cooke

andrew cooke — Thu, 28 Jul 2005 14:02:37 -0800

(although the lack of ordering in most queries - which don't use quotes or plus - can't be described in regular expressions without a kleene star!)