Is This an Optimal Substring Solution???

[Rex Winn]

I'm new to Java but experienced as a developer. I'm working on a string parser that has to fire every second and scan a full page of text that is rendered as HTML. I'm wondering if I can make it run faster using better code. Here's what I'm using (see code) is there a better/faster way to do this? Would I get better performance using RegEx?

I think something helpful to experts is that HTML_STAT_SUMMARY gets refreshed about every 5 seconds. I query that page of HTML for about 30 substrings and I have 30 functions set up just like the one below. Is this a good approach or do I leverage my position in the document and move forward on each search? I do know where each value will tend to be in the HTML_STAT_SUMMARY so I think I could be more efficien in stepping through as well by saving my current position and doing forward lookups.

All comments or ideas are welcome.

 public String getCarrierInfo()
    {
        String carInfo = "";
        String strHTML = "";
 
        int intStringPos = 0;
 
        strHTML = HTML_STAT_SUMMARY;
 
        if(strHTML.length() < 20)
            carInfo = "??";
        else
        {
            intStringPos = strHTML.indexOf(TAG_CARRIER_INFO, 0);
            intStringPos += TAG_CARRIER_INFO.length();
 
            if(intStringPos < 10)
                carInfo = "??";
            else
                carInfo = strHTML.substring(intStringPos, intStringPos + 10).trim();
        }
        return carInfo;
    }

[Harald Kirsch]

A few remarks:

It is rather surprising that the string to parse is not passed in as a paramater. Getting it into the function scope by accessing the global variable HTML_STAT_SUMMARY is weird.

At one point you use TAG_CARRIER_INFO.length(), then again you use a 10 hardcoded.

If the 30 tags are in a certain order, it would definitively pay to search them in order and avoid reading the complete whole string 30 times.

Finally a shameless plug: Searching for 30 (or many more) strings (aka regular expressions) in a document to extract surrounding text is the perfect use case for monq.jfa (GPL software). It allows you to stick pattern/action pairs into a finite automaton that reads your text and calls the actions whenever a pattern is matched. Throughput is 1.5MB/s on a 2.6GHz Pentium. Setting up the automaton looks like this:

  Dfa dfa = new 
    Nfa("first pattern", new FirstAction())
    .or("[Ss]econd *[Pp]attern", new SecondAction())
    .or("[\r\n\t ]", new Drop()) /// ignore whitespace
    ...
    .compile(SKIP_UNMATCHED_TEXT)
    ;

Javadoc: http://www.ebi.ac.uk/~kirsch/monq-doc/
Download: ftp://ftp.ebi.ac.uk/pub/software/textmining/monq/
Tutorial: http://www.ebi.ac.uk/~kirsch/JfaWiki/

[Rex Winn]

OOPS... As far as the parameter goes... I kind of cobbled that whole thing together to simplify reading it. I missed a few things when I cobbled it. But you have the gist of it in your reply.

Shameless Plug? I had found a thing called CUP. Now I'll have to go check out what you are suggesting. A link is worth a 1000 googles...

Thanks for your reply though.

[Rex Winn]

Oops found it in your signature to. I'm at freshmeat right now. Will jump to your signature..