Thursday, December 17, 2009

Regular Expressions in Magik

It has been bothering me for years that Magik does not seem to support Regular Expressions. I tried to make do with the ro_charindex_mixin.matches?() method. But that method only uses wildcard characters %*, %? and %\.

So when l_santi asked a question in sw-gis about Magik support for Regular Expressions, it got me thinking: WWBD (What Would Bhimesh Do)? Bhimesh is the owner of the Magik Fun blog and many of his posts explore the interaction of Magik with various OLE objects. I began to wonder if there was an OLE object that could support regular expressions. It turns out that vbscript.regexp is exactly what I was looking for.

Here is an example of testing that a telephone number string matches the correct format for North America...


MagikSF> regexp <<  ole_client.createobject("vbscript.regexp")
$
MagikSF> regexp.pattern << "^[0-9]{3}-[0-9]{3}-[0-9]{4}$"
$
MagikSF> regexp.test("303-555")
$
False
MagikSF> regexp.test("303-555-5555")
$
True

... and another example testing if a string is in a valid e-mail format.

MagikSF> regexp.pattern << "\b[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,4}\b"
$
MagikSF> regexp.test("alfred@ifactorconsulting.com")
$
True
MagikSF> regexp.test("alfred!ifactorconsulting.com")
$
False
at

5 comments:

  1. AnonymousDecember 17, 2009 at 12:49 PM

    Thank you Alfred. Another missing piece of the puzzle created!

    I jogged down the road a step with your idea and came up with this...

    _pragma(classify_level=basic, topic={strings}, usage={subclassable})
    _method ro_charindex_mixin.matches_regexp?( a_reg_exp )
    ## Returns true if self matches the regular expression
    ## A_REG_EXP, else false. On error, returns unset and the
    ## condition that was raised
    ##
    ## Uses the ole object vbscript.regexp to perform the analysis

    # adapted from Alfred Sawatzky
    # http://sworldwatch.blogspot.com/2009/12/regular-expressions-in-magik.html

    matches? << _unset

    _try _with cond
    _protect
    regexp << ole_client.createobject( "vbscript.regexp" )
    regexp.pattern << a_reg_exp

    matches? << regexp.test( _self )
    _protection
    _if ( regexp _isnt _unset )
    _then
    regexp.release_object()
    _endif
    _endprotect
    _when error
    _return matches?, cond
    _endtry

    _return matches?
    _endmethod
    $

    Bruce Morehouse, InMaps

    ReplyDelete
  2. Alfred SawatzkyDecember 17, 2009 at 12:51 PM

    Hi Bruce,

    That's a great step! Wrapping it like that and writing it on ro_charindex_mixin as :matches_regexp?() will make it easy to find in the class browser.

    Alfred

    ReplyDelete
  3. Bhimu...December 17, 2009 at 11:34 PM

    Hi Alfred

    This is really cool..
    .Net has built in regular Expersssion validator.
    using similar validator in Magik session with OLE Automation. .. This is really good..

    I think finding the Ole components is real challange.. This post may help in finding available Ole components http://magikfun.blogspot.com/2009/12/how-do-i-ole.html

    What if the user Operating System is non-windows?

    Thanks,
    Bhimesh

    ReplyDelete
  4. Alfred SawatzkyDecember 18, 2009 at 6:41 AM

    Hi Bhimesh,

    Thanks for the information about the OLE Viewer tool. If the session is not running in Windows, I guess you are out of luck. It would be nice if regular expressions could be incorporated directly into Magik by the core product team.

    I always seem to have difficulty finding the exact download page at Microsoft. So I have found the OLE2 View download site and created a link for it here: http://tinyurl.com/ole2view-download

    Alfred

    ReplyDelete
  5. Mike BullerDecember 21, 2009 at 11:37 AM

    If you're looking for platform independence, how about using a Java ACP to leverage Java's built-in regular expression parsing? The class of relevance is...

    java.lang.String

    ...on which you'll find the method...

    matches(java.lang.String regex)

    I just gave this a shot, and it seems to work pretty nicely.


    Here's the main Java code.


    public RegExACP() throws Exception
    {
    String stringToMatch = this.getString8();
    String regEx = this.getString8();
    this.putBoolean(stringToMatch.matches(regEx));
    this.flush();
    }


    Here's the Magik call from ro_charindex_mixin...


    _method ro_charindex_mixin.matches?(a_regex)
    an_acp << regex_acp.new()
    matches? << _unset
    _protect
    matches? << an_acp.matches_regex?(_self, a_regex)
    _protection
    _if an_acp _isnt _unset
    _then
    an_acp.close()
    _endif
    _endprotect
    _return matches?
    _endmethod
    $


    And finally, here's the interesting ACP code.


    _method regex_acp.matches_regex?(a_string, a_regex)
    ok? << _false
    matches? << _unset
    _protect _locking _self
    _self.locked_start()
    _self.put_chars8(system.current_text_encoding, a_string)
    _self.put_chars8(system.current_text_encoding, a_regex)
    _self.flush()
    matches? << _self.get_boolean()
    ok? << _true
    _protection
    _self.locked_end(ok?)
    _endprotect
    _return matches?
    _endmethod
    $

    All that's left to write is the ACP plumbing in Java and Magik.

    ReplyDelete