Editing Topic: RCR179 Project: RCR | RCRchive home

RCR 179: Create a 'NoMatchData' object on Regex match failing.

submitted by cyent on Mon Jan 05 2004 12:42:15 PM -0800

Status: pending

Abstract

Extend MatchData to have information on why a Regular Expression match failed.

Problem

How often have you written a Regex and expected it to match something and it just didn't? You got it wrong, but you have no idea how. Or worse, your user has entered input wrong, and you can't tell him what, just that it is wrong. Any sensible parser when choking on input will tell you, 'Syntax Error: Expecting one of "blah,bloo,foo"', so why can't Ruby Regex's? For example if I try match "digger" =~ /dig(by|raph)/ expecting to match either digby or digraph it would be nice if Ruby could tell me, "No match, longest match "dig", expecting a character in set [br] but found 'g' at position 3"

Proposal

There are several ways in which this could be achieved... 1. Rewrite a Regex engine in Ruby. 2. Change Regex::match to always return a MatchData that held the info. This would break some existing code. 3. Add a new method Regex::try_match(string) which would always return a MatchData, even if it didn't match. 4. Add a new modifier 'e' "digger" =~ /dig(by|raph)/e would return false, but set some global $whatever with the MatchData

Analysis

Option 1 would probably be too slow, option 2 would break too much existing code. I would favour option 3 or 4.

----
Unmatch is not an error, so that I'm not sure whether we need to report the detailed unmatch information.  Besides that, implementation is harder than you may expect.  Under the multibyte environment, to get "position 3" for example, is much more difficult.

Among choices, Option 1 is not realistic, option 2 is incompatible too much. and I hate option 3 to introduce new global variable.

- matz.
----
It seems that the use of the new Oniguramu regexp engine might make something like this more possible, but that it should be a specialized call on a regexp, and not a normal match. Perhaps the RCR author should ask the Oniguramu author(s) if something like this is feasible using Oniguramu, and if they would consider adding such a feature. Then a specialized method could be added to Regexp that attempts a match and returns an "error report" if it fails.

-- ntalbott
----
I have asked the author of Oniguruma if it is possible, I will wait on his reply.

I think an additional method regex.try_match( string) that returned an subclass of the MatchData object would be the best / least impacting extension.

The other think I noted whilst contemplating your replies, is that the failure information  should probably apply to the longest matching string.
----
Just my 2 cents: I know that I personally have spent many hours debugging complex regular expressions, and some return from a failed match that said where/why it failed would have been invaluable.  As for the implementations, I agree with matz in that adding a global variable is not a good solution.  I think the third option, which was to have an alternate method that would always return a MatchData.

-- JamisBuck

=== Add comments here

----

Back to RCRchive.

RCR Submission page and RCRchive powered by Ruby, Apache, RuWiki (modified), and RubLog