CODEWITHSUNDEEP
delphisearchpattern-matchingdelphi-xe2
Jerry Dodge
Jerry Dodge

Reputation: 27276

Is there something in Delphi to make it easier to perform keyword searches?

Part of my current project is comparing search results (filenames only) with a search string (multiple words). I have a very basic mechanism I use at the moment to identify the relevance of a result, all handled with one function.

When the search begins, I split the search string into a string list of keywords...

procedure TSearcherThread.ParseKeywords;
var
  S, T: String;
  P: Integer;
begin
  //Clear current list of keywords
  FKeywords.Clear;
  S:= LowerCase(Trim(FSearchString));
  //Remove all excess spaces
  while Pos('  ', S) > 1 do
    S:= StringReplace(S, '  ', ' ', [rfReplaceAll]);
  if Copy(S, Length(S)-1, 1) <> ';' then
    S:= S + ';';
  //Parse out keywords
  while Length(S) > 0 do begin
    P:= Pos(';', S);
    T:= Copy(S, 1, P-1);
    Delete(S, 1, P);
    FKeywords.Append(T);
  end;
end;

Now when I'm iterating through the master list of files to be searched, I pass each filename into this function...

function TSearcherThread.MatchKeywords(const Filename: String): Single;
var
  S: String; //Temp keywords
  FN: String; //Filename
  X: Integer; //Iterator
  C: Integer; //Match counter
begin
  Result:= 0; //Default no match
  S:= Trim(LowerCase(FSearchString)); //Lowercase Keywords, trim outside spaces
  FN:= LowerCase(ExtractFileName(Filename)); //Get lowercase filename
  Delete(FN, Pos('.', FN), MAXINT); //Strip off extension leaving only the name

  //Check if exact match
  if FN = S then Result:= 2;

  //If nothing matches yet, then look for individual keywords...
  if Result < 2 then begin
    C:= 0;
    if FKeywords.Count > 0 then begin
      //Iterate through keywords
      for X := 0 to FKeywords.Count - 1 do begin
        //If keyword is found in filename
        if Pos(FKeywords[X], FN) > 0 then begin
          Inc(C);
        end;
      end;
      //Return how often keywords showed up
      Result:= C / FKeywords.Count;
    end;
  end;
end;

How this works is the function passes back a decimal number of relevance. A result of 0 means no match, between 0 and 1 means partial match, where the higher the number, the better of a match, 1 means all keywords were found, and 2 means it's an exact match. I can also do a comparison to only include results which have a certain percentage, like this:

M:= MatchKeywords(Filename);
if M >= 0.2 then AddResult(Filename);

The problem is that my method above considers only AND operation, meaning it expects all the keywords, and compares how many keywords were found. However, I would like to also implement combinations of both AND and OR operations together, which my structure doesn't support. So I need to re-write the guts of this function to make this possible.

What I would like to know is not how to write this, but is there something in Delphi which can make this possible? Someone mentioned to me TDictionary as a Hash Table is what I would need, but I have no clue how it relates to what I'm doing, as I've never used them. I just don't want to re-invent the wheel of pattern matching if it already exists in Delphi XE2.

Upvotes: 1

Views: 633

Answers (1)

julienf2logicells
julienf2logicells

Reputation: 34

the simplest way finding some string patterns is using Regular Expressions Engine. You could find some free units and packages on the web site of FPC.

http://wiki.freepascal.org/Regexpr

And read more about Regular Expression.

Upvotes: 1

Related Questions