Faster way to split text in Delphi TStringList - delphi

Question

I have an app that needs to do heavy text manipulation in a TStringList. Basically i need to split text by a delimiter ; for instance, if i have a singe line with 1000 chars and this delimiter occurs 3 times in this line, then i need to split it in 3 lines. The delimiter can contain more than one char, it can be a tag like '[test]' for example.
I've wrote two functions to do this task with 2 different approaches, but both are slow in big amounts of text (more then 2mbytes usually).
How can i achieve this goal in a faster way ?
Here are both functions, both receive 2 paramaters : 'lines' which is the original tstringlist and 'q' which is the delimiter.
function splitlines(lines : tstringlist; q: string) : integer;
var
s, aux, ant : string;
i,j : integer;
flag : boolean;
m2 : tstringlist;
begin
try
m2 := tstringlist.create;
m2.BeginUpdate;
result := 0;
for i := 0 to lines.count-1 do
begin
s := lines[i];
for j := 1 to length(s) do
begin
flag := lowercase(copy(s,j,length(q))) = lowercase(q);
if flag then
begin
inc(result);
m2.add(aux);
aux := s[j];
end
else
aux := aux + s[j];
end;
m2.add(aux);
aux := '';
end;
m2.EndUpdate;
lines.text := m2.text;
finally
m2.free;
end;
end;
function splitLines2(lines : tstringlist; q: string) : integer;
var
aux, p : string;
i : integer;
flag : boolean;
begin
//maux1 and maux2 are already instanced in the parent class
try
maux2.text := lines.text;
p := '';
i := 0;
flag := false;
maux1.BeginUpdate;
maux2.BeginUpdate;
while (pos(lowercase(q),lowercase(maux2.text)) > 0) and (i < 5000) do
begin
flag := true;
aux := p+copy(maux2.text,1,pos(lowercase(q),lowercase(maux2.text))-1);
maux1.add(aux);
maux2.text := copy(maux2.text,pos(lowercase(q),lowercase(maux2.text)),length(maux2.text));
p := copy(maux2.text,1,1);
maux2.text := copy(maux2.text,2,length(maux2.text));
inc(i);
end;
finally
result := i;
maux1.EndUpdate;
maux2.EndUpdate;
if flag then
begin
maux1.add(p+maux2.text);
lines.text := maux1.text;
end;
end;
end;

Answer 1

I've not tested the speed, but for academic purposes, here's an easy way to split the strings:
myStringList.Text :=
StringReplace(myStringList.Text, myDelimiter, #13#10, [rfReplaceAll]);
// Use [rfReplaceAll, rfIgnoreCase] if you want to ignore case
When you set the Text property of TStringList, it parses on new lines and splits there, so converting to a string, replacing the delimiter with new lines, then assigning it back to the Text property works.

Answer 2

The problems with your code (at least second approach) are
You are constantly using lowecase which is slow if called so many times
If I saw correctly you are copying the whole remaining text back to the original source. This is sure to be extra slow for large strings (eg files)
I have a tokenizer in my library. Its not the fastest or best but it should do (you can get it from Cromis Library, just use the units Cromis.StringUtils and Cromis.Unicode):
type
TTokens = array of ustring;
TTextTokenizer = class
private
FTokens: TTokens;
FDelimiters: array of ustring;
public
constructor Create;
procedure Tokenize(const Text: ustring);
procedure AddDelimiters(const Delimiters: array of ustring);
property Tokens: TTokens read FTokens;
end;
{ TTextTokenizer }
procedure TTextTokenizer.AddDelimiters(const Delimiters: array of ustring);
var
I: Integer;
begin
if Length(Delimiters) > 0 then
begin
SetLength(FDelimiters, Length(Delimiters));
for I := 0 to Length(Delimiters) - 1 do
FDelimiters[I] := Delimiters[I];
end;
end;
constructor TTextTokenizer.Create;
begin
SetLength(FTokens, 0);
SetLength(FDelimiters, 0);
end;
procedure TTextTokenizer.Tokenize(const Text: ustring);
var
I, K: Integer;
Counter: Integer;
NewToken: ustring;
Position: Integer;
CurrToken: ustring;
begin
SetLength(FTokens, 100);
CurrToken := '';
Counter := 0;
for I := 1 to Length(Text) do
begin
CurrToken := CurrToken + Text[I];
for K := 0 to Length(FDelimiters) - 1 do
begin
Position := Pos(FDelimiters[K], CurrToken);
if Position > 0 then
begin
NewToken := Copy(CurrToken, 1, Position - 1);
if NewToken <> '' then
begin
if Counter > Length(FTokens) then
SetLength(FTokens, Length(FTokens) * 2);
FTokens[Counter] := Trim(NewToken);
Inc(Counter)
end;
CurrToken := '';
end;
end;
end;
if CurrToken <> '' then
begin
if Counter > Length(FTokens) then
SetLength(FTokens, Length(FTokens) * 2);
FTokens[Counter] := Trim(CurrToken);
Inc(Counter)
end;
SetLength(FTokens, Counter);
end;

Answer 3

How about just using StrTokens from the JCL library
procedure StrTokens(const S: string; const List: TStrings);
It's open source
http://sourceforge.net/projects/jcl/

Answer 4

As an additional option, you can use regular expressions. Recent versions of Delphi (XE4 and XE5) come with built in regular expression support; older versions can find a free regex library download (zip file) at Regular-Expressions.info.
For the built-in regex support (uses the generic TArray<string>):
var
RegexObj: TRegEx;
SplitArray: TArray<string>;
begin
SplitArray := nil;
try
RegexObj := TRegEx.Create('\[test\]'); // Your sample expression. Replace with q
SplitArray := RegexObj.Split(Lines, 0);
except
on E: ERegularExpressionError do begin
// Syntax error in the regular expression
end;
end;
// Use SplitArray
end;
For using TPerlRegEx in earlier Delphi versions:
var
Regex: TPerlRegEx;
m2: TStringList;
begin
m2 := TStringList.Create;
try
Regex := TPerlRegEx.Create;
try
Regex.RegEx := '\[test\]'; // Using your sample expression - replace with q
Regex.Options := [];
Regex.State := [preNotEmpty];
Regex.Subject := Lines.Text;
Regex.SplitCapture(m2, 0);
finally
Regex.Free;
end;
// Work with m2
finally
m2.Free;
end;
end;
(For those unaware, the \ in the sample expression used are because the [] characters are meaningful in regular expressions and need to be escaped to be used in the regular expression text. Typically, they're not required in the text.)