StringBuilder

[Alvin Bruney]

On the advice of a user, I've timed stringbuilder v string. Here are the
results.

Here are the numbers:
Total # queries 3747
Time in Milliseconds

StringBuilder: String
460.6624 320.4608
350.504 220.3168
240.3456 230.3312
----------- -----------
Ave. 350.504 Ave. 257.0362

Sample query:
"Update ds_active_rate SET (rate, surcharge, min_sec, increment_sec,
begin_date, new_rate, new_rate_date, location_desc, category_id,
country_code, customer_number , user_name)=('0' ,'0','0','0','2 150-01-01
12:00:00','','' ,'Afghanistan', 2,93,'1STAMERIC AN','') where rate_deck =
'COPYRATE' and location_id = 1001 and world_id = 1 and rate_level = 1"

Specs on my machine
1Gig SDRAM
P-IV 2Gig Intel
20Gig Hard Drive
Dell C-640

What I did: I let the routine run till it hit a breakpoint after a timespan
object. Noted the time, and pointed execution back to the first time object
and let it run to the break point. This gave me my three times. Obviously,
some sort of optimization was going on in the background for successive runs
which accounts for increasingly lower times. Then I converted the code to
string and repeated the procedure. I don't claim that these numbers are
scientific. But I did my best to control the process.

My conclusion is: I think these microsoft guys have been feeding us a bunch
of bullshit with this stringbuilder efficiency crap.

[Richard A. Lowe]

StringBuilder

Strings are expensive to concatenate as a (partial)
function of their size... While I can't be sure of what
kind of test you were doing, if you test concatenation of
increasingly large strings to one another, the
StringBuilder can be dozens, if not hundreds of times
faster than the + operator.

If you are just concatenating two small strings, then yes,
the + operator will be faster.

Richard
[color=blue]
>-----Original Message-----
>On the advice of a user, I've timed stringbuilder v[/color]
string. Here are the[color=blue]
>results.
>
>Here are the numbers:
>Total # queries 3747
>Time in Milliseconds
>
>StringBuilde r: String
>460.6624 320.4608
>350.504 220.3168
>240.3456 230.3312
>----------- -----------
>Ave. 350.504 Ave. 257.0362
>
>Sample query:
>"Update ds_active_rate SET (rate, surcharge, min_sec,[/color]
increment_sec,[color=blue]
>begin_date, new_rate, new_rate_date, location_desc,[/color]
category_id,[color=blue]
>country_code , customer_number , user_name)=[/color]
('0','0','0','0 ','2150-01-01[color=blue]
>12:00:00','',' ','Afghanistan' ,2,93,'1STAMERI CAN','')[/color]
where rate_deck =[color=blue]
>'COPYRATE' and location_id = 1001 and world_id = 1 and[/color]
rate_level = 1"[color=blue]
>
>Specs on my machine
>1Gig SDRAM
>P-IV 2Gig Intel
>20Gig Hard Drive
>Dell C-640
>
>What I did: I let the routine run till it hit a[/color]
breakpoint after a timespan[color=blue]
>object. Noted the time, and pointed execution back to the[/color]
first time object[color=blue]
>and let it run to the break point. This gave me my three[/color]
times. Obviously,[color=blue]
>some sort of optimization was going on in the background[/color]
for successive runs[color=blue]
>which accounts for increasingly lower times. Then I[/color]
converted the code to[color=blue]
>string and repeated the procedure. I don't claim that[/color]
these numbers are[color=blue]
>scientific. But I did my best to control the process.
>
>My conclusion is: I think these microsoft guys have been[/color]
feeding us a bunch[color=blue]
>of bullshit with this stringbuilder efficiency crap.
>
>
>
>.
>[/color]

[Jay B. Harlow [MVP - Outlook]]

Re: StringBuilder

Alvin,[color=blue]
> My conclusion is: I think these microsoft guys have been feeding us a[/color]
bunch[color=blue]
> of bullshit with this stringbuilder efficiency crap.[/color]
If you are simply building a single update statement then yes the
StringBuilder maybe like using a shot gun to kill the fly on your foot.

However if you were to build 1000 update statements as a single batch, then
the StringBuilder will really shine!

In other words the first will not perform well (as I believe you
demonstrated), the second will fly!

for (I = 0; I < 1000; I++)
{
StringBuilder sb = new StringBuilder(" insert(");
sb.Append("x");
sb.Append(")");
String sql = sb.ToString();
}

StringBuilder sb = new StringBuilder() ;
for (i = 0; i < 1000; i++)
{
sb.Append("inse rt(");
sb.Append("x");
sb.Append(")");
sb.Append(';'); // end of one statement
}
String sql = sb.ToString();

As in the first I am building 1000 individual commands, presumable to
execute 1000 individual times. Where as in the second I am building a single
batch with 1000 commands.

Forgive the super simple insert statement, I'm demonstrating how to use the
StringBuilder, not SQL syntax ;-)

Also irregardless of the timing, you also have temporary string objects to
consider, in the first case there may be fewer temporary string objects
created, so although its slightly 'slower' the GC does not need to work as
hard.

Don't forget when you allocate the StringBuilder, that you should give the
capacity value if you know roughly how large the string may get, as this
improves performance also. (the StringBuilder does not need to reallocate
its buffer, reallocating the buffer takes time!)
[color=blue]
> This gave me my three times. Obviously,
> some sort of optimization was going on in the background for successive[/color]
runs[color=blue]
> which accounts for increasingly lower times.[/color]
Generally its the JIT compiler the first time through things get 'compiled'
the second time things are already compiled.

One last comment, as was discussed earlier (in a different thread), if you
are using string concatenation for the parameters to StringBuilder.A ppend,
then you may be negating the benefit of the StringBuilder, as you are
creating temporary string objects.

Depending on intended use of the class, I will actually allocate a single
StringBuilder in the constructor then continue to reuse the single
StringBuilder. Setting StringBuilder.L ength = 0 each time I started. This
causes the contents of the buffer to be 'cleared' without allocating a new
StringBuilder buffer.

Hope this helps
Jay

"Alvin Bruney" <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote in
message news:ebY1L7dgDH A.4024@TK2MSFTN GP11.phx.gbl...[color=blue]
> On the advice of a user, I've timed stringbuilder v string. Here are the
> results.
>
> Here are the numbers:
> Total # queries 3747
> Time in Milliseconds
>
> StringBuilder: String
> 460.6624 320.4608
> 350.504 220.3168
> 240.3456 230.3312
> ----------- -----------
> Ave. 350.504 Ave. 257.0362
>
> Sample query:
> "Update ds_active_rate SET (rate, surcharge, min_sec, increment_sec,
> begin_date, new_rate, new_rate_date, location_desc, category_id,
> country_code, customer_number , user_name)=('0' ,'0','0','0','2 150-01-01
> 12:00:00','','' ,'Afghanistan', 2,93,'1STAMERIC AN','') where rate_deck =
> 'COPYRATE' and location_id = 1001 and world_id = 1 and rate_level = 1"
>
> Specs on my machine
> 1Gig SDRAM
> P-IV 2Gig Intel
> 20Gig Hard Drive
> Dell C-640
>
> What I did: I let the routine run till it hit a breakpoint after a[/color]
timespan[color=blue]
> object. Noted the time, and pointed execution back to the first time[/color]
object[color=blue]
> and let it run to the break point. This gave me my three times. Obviously,
> some sort of optimization was going on in the background for successive[/color]
runs[color=blue]
> which accounts for increasingly lower times. Then I converted the code to
> string and repeated the procedure. I don't claim that these numbers are
> scientific. But I did my best to control the process.
>
> My conclusion is: I think these microsoft guys have been feeding us a[/color]
bunch[color=blue]
> of bullshit with this stringbuilder efficiency crap.
>
>
>[/color]

[Alvin Bruney]

Re: StringBuilder

I apologize for not providing the code. The batch command explanation passed
over my head. I'd like you to explain this a little more. I believe the code
follows your quidelines. This is a raw cut and paste by the way.

for(int index = GlobalRateMan.I NITVALUE; index < splitField2.Len gth;
index++)

{

splitField = GlobalRateMan.R egexComma.Split (splitField2[index]);

if(ForceCreate. Checked)

strTempString.A ppend(GlobalRat eMan.sqlUpdateA ctiveRate2);

else

//insert extra fields

strTempString.A ppend(GlobalRat eMan.sqlUpdateA ctiveRate1);

//check for valid record. replace some excel formatting

if(splitField.L ength == 37)

{

newrate = splitField[NEW_RATE];

newdate = GlobalRateMan.F ormatDate(split Field[NEW_RATE_DATE]);

//both fields must be entered or neither in the case of updates to field
like surcharges

if(newrate.Leng th > 0 && newdate.Length > 0)

{

strTempString.A ppend("'").Appe nd(splitField[RATE]).Append("',");

strTempString.A ppend("'").Appe nd(splitField[SURCHARGE]).Append("',");

strTempString.A ppend("'").Appe nd(splitField[MIN_SEC]).Append("',");

strTempString.A ppend("'").Appe nd(splitField[INCR_SEC]).Append("',");

strTempString.A ppend("'").Appe nd(GlobalRateMa n.FormatDate(sp litField[BEGIN_D
ATE])).Append("',") ;

strTempString.A ppend("'").Appe nd(newrate.Repl ace("$",GlobalR ateMan.EMPTY)). A
ppend("',");

strTempString.A ppend("'").Appe nd(newdate).App end("',");

strTempString.A ppend("'").Appe nd(InsertUserNa me().TrimEnd()) .Append("')");


//option which is checked for the time test

if(ForceCreate. Checked)

{

strTempString.A ppend("'").Appe nd(splitField[8] + "',");

strTempString.A ppend(splitFiel d[12]).Append(",");

strTempString.A ppend(splitFiel d[14]).Append(",");

strTempString.A ppend("'").Appe nd(splitField[2]).Append("',");

}

}


strTempString = strTempString.R eplace("',)","' )");

if(newrate.Leng th == 0 && newdate.Length == 0)

strTempString = strTempString.R eplace("new_rat e,new_rate_date ,","");

//usually 3 guys are in this listbox

for(int i = 0; i < RateDeckConfirm .Items.Count; i++)

{

strTempString2. Length = 0;

strTempString2. Append(" where rate_deck =
'").Append(Rate DeckConfirm.Ite ms[i].Text.TrimEnd() ).Append("'");

strTempString2. Append(" and location_id = ").Append(split Field[16]);

strTempString2. Append(" and world_id = ").Append(split Field[10]);

strTempString2. Append(" and rate_level = ").Append(split Field[34]);

//caught a live one

sql.Add(strTemp String + strTempString2. ToString());

}

strTempString.L ength = 0;

newrate = newdate = "";

}

}

"Jay B. Harlow [MVP - Outlook]" <Jay_Harlow@ema il.msn.com> wrote in message
news:exT3aaegDH A.2260@TK2MSFTN GP10.phx.gbl...[color=blue]
> Alvin,[color=green]
> > My conclusion is: I think these microsoft guys have been feeding us a[/color]
> bunch[color=green]
> > of bullshit with this stringbuilder efficiency crap.[/color]
> If you are simply building a single update statement then yes the
> StringBuilder maybe like using a shot gun to kill the fly on your foot.
>
> However if you were to build 1000 update statements as a single batch,[/color]
then[color=blue]
> the StringBuilder will really shine!
>
> In other words the first will not perform well (as I believe you
> demonstrated), the second will fly!
>
> for (I = 0; I < 1000; I++)
> {
> StringBuilder sb = new StringBuilder(" insert(");
> sb.Append("x");
> sb.Append(")");
> String sql = sb.ToString();
> }
>
> StringBuilder sb = new StringBuilder() ;
> for (i = 0; i < 1000; i++)
> {
> sb.Append("inse rt(");
> sb.Append("x");
> sb.Append(")");
> sb.Append(';'); // end of one statement
> }
> String sql = sb.ToString();
>
> As in the first I am building 1000 individual commands, presumable to
> execute 1000 individual times. Where as in the second I am building a[/color]
single[color=blue]
> batch with 1000 commands.
>
> Forgive the super simple insert statement, I'm demonstrating how to use[/color]
the[color=blue]
> StringBuilder, not SQL syntax ;-)
>
> Also irregardless of the timing, you also have temporary string objects to
> consider, in the first case there may be fewer temporary string objects
> created, so although its slightly 'slower' the GC does not need to work as
> hard.
>
> Don't forget when you allocate the StringBuilder, that you should give the
> capacity value if you know roughly how large the string may get, as this
> improves performance also. (the StringBuilder does not need to reallocate
> its buffer, reallocating the buffer takes time!)
>[color=green]
> > This gave me my three times. Obviously,
> > some sort of optimization was going on in the background for successive[/color]
> runs[color=green]
> > which accounts for increasingly lower times.[/color]
> Generally its the JIT compiler the first time through things get[/color]
'compiled'[color=blue]
> the second time things are already compiled.
>
> One last comment, as was discussed earlier (in a different thread), if you
> are using string concatenation for the parameters to StringBuilder.A ppend,
> then you may be negating the benefit of the StringBuilder, as you are
> creating temporary string objects.
>
> Depending on intended use of the class, I will actually allocate a single
> StringBuilder in the constructor then continue to reuse the single
> StringBuilder. Setting StringBuilder.L ength = 0 each time I started. This
> causes the contents of the buffer to be 'cleared' without allocating a new
> StringBuilder buffer.
>
> Hope this helps
> Jay
>
> "Alvin Bruney" <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote in
> message news:ebY1L7dgDH A.4024@TK2MSFTN GP11.phx.gbl...[color=green]
> > On the advice of a user, I've timed stringbuilder v string. Here are the
> > results.
> >
> > Here are the numbers:
> > Total # queries 3747
> > Time in Milliseconds
> >
> > StringBuilder: String
> > 460.6624 320.4608
> > 350.504 220.3168
> > 240.3456 230.3312
> > ----------- -----------
> > Ave. 350.504 Ave. 257.0362
> >
> > Sample query:
> > "Update ds_active_rate SET (rate, surcharge, min_sec, increment_sec,
> > begin_date, new_rate, new_rate_date, location_desc, category_id,
> > country_code, customer_number , user_name)=('0' ,'0','0','0','2 150-01-01
> > 12:00:00','','' ,'Afghanistan', 2,93,'1STAMERIC AN','') where rate_deck =
> > 'COPYRATE' and location_id = 1001 and world_id = 1 and rate_level =[/color][/color]
1"[color=blue][color=green]
> >
> > Specs on my machine
> > 1Gig SDRAM
> > P-IV 2Gig Intel
> > 20Gig Hard Drive
> > Dell C-640
> >
> > What I did: I let the routine run till it hit a breakpoint after a[/color]
> timespan[color=green]
> > object. Noted the time, and pointed execution back to the first time[/color]
> object[color=green]
> > and let it run to the break point. This gave me my three times.[/color][/color]
Obviously,[color=blue][color=green]
> > some sort of optimization was going on in the background for successive[/color]
> runs[color=green]
> > which accounts for increasingly lower times. Then I converted the code[/color][/color]
to[color=blue][color=green]
> > string and repeated the procedure. I don't claim that these numbers are
> > scientific. But I did my best to control the process.
> >
> > My conclusion is: I think these microsoft guys have been feeding us a[/color]
> bunch[color=green]
> > of bullshit with this stringbuilder efficiency crap.
> >
> >
> >[/color]
>
>[/color]

[Jon Skeet]

Re: StringBuilder

Alvin Bruney <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote:[color=blue]
> On the advice of a user, I've timed stringbuilder v string.[/color]

Doing what with them? You haven't actually said what operations you're
performing on them in the first place.

Note that for doing simple benchmarks like this you might like to use
my benchmarking framework:

http://www.pobox.com/~skeet/csharp/benchmark.html

That's much easier than using breakpoints, and it means you can run
without the debugger attached (which is going to be much more like
real-world performance).
[color=blue]
> My conclusion is: I think these microsoft guys have been feeding us a bunch
> of bullshit with this stringbuilder efficiency crap.[/color]

StringBuilder *is* more efficient for building strings - but as you
haven't said what you did in your tests, it's rather hard to tell what
you've actually measured.

If you really doubt that using StringBuilder is a good idea for
building up strings, I suggest you try these two programs:

using System;
using System.Text;

public class TestStringBuild er
{
public static void Main(string[] args)
{
DateTime start = DateTime.Now;
StringBuilder sb = new StringBuilder() ;
for (int i=0; i < 100000; i++)
{
sb.Append ('x');
}
DateTime end = DateTime.Now;
Console.WriteLi ne (sb.ToString(). Length);
Console.WriteLi ne (end-start);
}
}



using System;
using System.Text;

public class TestString
{
public static void Main(string[] args)
{
DateTime start = DateTime.Now;
string s = "";
for (int i=0; i < 100000; i++)
{
s = s+'x';
}
DateTime end = DateTime.Now;
Console.WriteLi ne (s.Length);
Console.WriteLi ne (end-start);
}
}

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too

[Hasani]

Re: StringBuilder

You didn't track the memory usage. which is one of the points of using the
stringbuilder over string
"Alvin Bruney" <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote in
message news:ebY1L7dgDH A.4024@TK2MSFTN GP11.phx.gbl...[color=blue]
> On the advice of a user, I've timed stringbuilder v string. Here are the
> results.
>
> Here are the numbers:
> Total # queries 3747
> Time in Milliseconds
>
> StringBuilder: String
> 460.6624 320.4608
> 350.504 220.3168
> 240.3456 230.3312
> ----------- -----------
> Ave. 350.504 Ave. 257.0362
>
> Sample query:
> "Update ds_active_rate SET (rate, surcharge, min_sec, increment_sec,
> begin_date, new_rate, new_rate_date, location_desc, category_id,
> country_code, customer_number , user_name)=('0' ,'0','0','0','2 150-01-01
> 12:00:00','','' ,'Afghanistan', 2,93,'1STAMERIC AN','') where rate_deck =
> 'COPYRATE' and location_id = 1001 and world_id = 1 and rate_level = 1"
>
> Specs on my machine
> 1Gig SDRAM
> P-IV 2Gig Intel
> 20Gig Hard Drive
> Dell C-640
>
> What I did: I let the routine run till it hit a breakpoint after a[/color]
timespan[color=blue]
> object. Noted the time, and pointed execution back to the first time[/color]
object[color=blue]
> and let it run to the break point. This gave me my three times. Obviously,
> some sort of optimization was going on in the background for successive[/color]
runs[color=blue]
> which accounts for increasingly lower times. Then I converted the code to
> string and repeated the procedure. I don't claim that these numbers are
> scientific. But I did my best to control the process.
>
> My conclusion is: I think these microsoft guys have been feeding us a[/color]
bunch[color=blue]
> of bullshit with this stringbuilder efficiency crap.
>
>
>[/color]

[Jon Skeet]

Re: StringBuilder

Jay B. Harlow [MVP - Outlook] <Jay_Harlow@ema il.msn.com> wrote:[color=blue]
> Depending on intended use of the class, I will actually allocate a single
> StringBuilder in the constructor then continue to reuse the single
> StringBuilder. Setting StringBuilder.L ength = 0 each time I started. This
> causes the contents of the buffer to be 'cleared' without allocating a new
> StringBuilder buffer.[/color]

I don't believe this is a good idea, personally. Creating a single
empty StringBuilder is cheap - and the code is more natural. When you
want to build a string, you create a StringBuilder and use it. When you
want to build the next string, you create a new StringBuilder. No
threading worries, no extra member which has no logical reason for
existing, etc.

Also, things are likely to stay in generation 0 - why get stuff pushed
up to higher generations (where write barriers might become an issue)
unnecessarily?

Unless I had some reason to think that the StringBuilder code was the
bottleneck *and* I had profiled both ways of working, I'd use the more
natural code.

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too

[Jon Skeet]

Re: StringBuilder

Alvin Bruney <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote:[color=blue]
> I apologize for not providing the code. The batch command explanation passed
> over my head. I'd like you to explain this a little more. I believe the code
> follows your quidelines. This is a raw cut and paste by the way.[/color]

Yes - if you could change the formatting so it came out looking
slightly more readable in future, that would be appreciated :)

You've shown the StringBuilder code, but not the String code.

I strongly suggest you refactor this code in a way that lets you do a
straight, repeatable benchmark (i.e. it's a complete program which
doesn't depend on any external factors, it has all the data constructed
before the loop is entered, etc) and post both cases of that. Certainly
based on that code, I would expect it to be significantly faster at
constructing the queries than straight string concatenations.

<snip>

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too

[Jay B. Harlow [MVP - Outlook]]

Re: StringBuilder

Jon,
Doh! ;-)

That's why the statement is prefixed with 'depending on intended use of the
class'. Read as 'The StringBuilder is an integral part of the implementation
of the class'. In other words a number of methods of the class delegate to a
StringBuilder.

Similar to how the System.IO.Strin gWriter class wraps a StringBuilder class.

Otherwise I agree for most things its not a good idea. I suppose I should
have emphasized that more.

Thanks for bringing up this caution!

Jay


"Jon Skeet" <skeet@pobox.co m> wrote in message
news:MPG.19da87 6fde6779d998972 4@news.microsof t.com...[color=blue]
> Jay B. Harlow [MVP - Outlook] <Jay_Harlow@ema il.msn.com> wrote:[color=green]
> > Depending on intended use of the class, I will actually allocate a[/color][/color]
single[color=blue][color=green]
> > StringBuilder in the constructor then continue to reuse the single
> > StringBuilder. Setting StringBuilder.L ength = 0 each time I started.[/color][/color]
This[color=blue][color=green]
> > causes the contents of the buffer to be 'cleared' without allocating a[/color][/color]
new[color=blue][color=green]
> > StringBuilder buffer.[/color]
>
> I don't believe this is a good idea, personally. Creating a single
> empty StringBuilder is cheap - and the code is more natural. When you
> want to build a string, you create a StringBuilder and use it. When you
> want to build the next string, you create a new StringBuilder. No
> threading worries, no extra member which has no logical reason for
> existing, etc.
>
> Also, things are likely to stay in generation 0 - why get stuff pushed
> up to higher generations (where write barriers might become an issue)
> unnecessarily?
>
> Unless I had some reason to think that the StringBuilder code was the
> bottleneck *and* I had profiled both ways of working, I'd use the more
> natural code.
>
> --
> Jon Skeet - <skeet@pobox.co m>
> http://www.pobox.com/~skeet
> If replying to the group, please do not mail me too[/color]

[Jon Skeet]

Re: StringBuilder

Jay B. Harlow [MVP - Outlook] <Jay_Harlow@ema il.msn.com> wrote:[color=blue]
> That's why the statement is prefixed with 'depending on intended use of the
> class'. Read as 'The StringBuilder is an integral part of the implementation
> of the class'. In other words a number of methods of the class delegate to a
> StringBuilder.[/color]

Ah, to a single StringBuilder. I'd thought you'd meant it would be
effectively started afresh each time, and had nothing to do with the
logical state of the object.
[color=blue]
> Similar to how the System.IO.Strin gWriter class wraps a StringBuilder class.[/color]

That makes a lot of sense, yes.
[color=blue]
> Otherwise I agree for most things its not a good idea. I suppose I should
> have emphasized that more.[/color]

Nah - don't blame yourself for my lack of comprehension skills :)

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too

[Alvin Bruney]

Re: StringBuilder

Now hold on. I said they weren't scientific. I prefaced the results by that.
I don't have a lot of time on my hands to go about making sure this will
stand up to the hounds of hell. I leave it up to MS for that.

What difference does it make what i do with them. The question I was after
was which way is faster to build strings. FYI, i take these queries and
write to the database. I don't see what is gained by that tidbit.

This is a real world program by the way. Not an arbitrary couple lines of
code make up. If I can't see results there, and I should, then either I am
doing something wrong. In this case point it out so I can fix the code or MS
is on a marketing hype with this stringbuilder/.NET crap.



"Jon Skeet" <skeet@pobox.co m> wrote in message
news:MPG.19da83 d5647df5e498972 3@news.microsof t.com...[color=blue]
> Alvin Bruney <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote:[color=green]
> > On the advice of a user, I've timed stringbuilder v string.[/color]
>
> Doing what with them? You haven't actually said what operations you're
> performing on them in the first place.
>
> Note that for doing simple benchmarks like this you might like to use
> my benchmarking framework:
>
> http://www.pobox.com/~skeet/csharp/benchmark.html
>
> That's much easier than using breakpoints, and it means you can run
> without the debugger attached (which is going to be much more like
> real-world performance).
>[color=green]
> > My conclusion is: I think these microsoft guys have been feeding us a[/color][/color]
bunch[color=blue][color=green]
> > of bullshit with this stringbuilder efficiency crap.[/color]
>
> StringBuilder *is* more efficient for building strings - but as you
> haven't said what you did in your tests, it's rather hard to tell what
> you've actually measured.
>
> If you really doubt that using StringBuilder is a good idea for
> building up strings, I suggest you try these two programs:
>
> using System;
> using System.Text;
>
> public class TestStringBuild er
> {
> public static void Main(string[] args)
> {
> DateTime start = DateTime.Now;
> StringBuilder sb = new StringBuilder() ;
> for (int i=0; i < 100000; i++)
> {
> sb.Append ('x');
> }
> DateTime end = DateTime.Now;
> Console.WriteLi ne (sb.ToString(). Length);
> Console.WriteLi ne (end-start);
> }
> }
>
>
>
> using System;
> using System.Text;
>
> public class TestString
> {
> public static void Main(string[] args)
> {
> DateTime start = DateTime.Now;
> string s = "";
> for (int i=0; i < 100000; i++)
> {
> s = s+'x';
> }
> DateTime end = DateTime.Now;
> Console.WriteLi ne (s.Length);
> Console.WriteLi ne (end-start);
> }
> }
>
> --
> Jon Skeet - <skeet@pobox.co m>
> http://www.pobox.com/~skeet
> If replying to the group, please do not mail me too[/color]

[Alvin Bruney]

Re: StringBuilder

True. I haven't tested for that. This was beyond the scope of the exercise.

"Hasani" <HJB417@hotmail .c0m> wrote in message
news:_F_bb.1582 2$nU6.3116914@t wister.nyc.rr.c om...[color=blue]
> You didn't track the memory usage. which is one of the points of using the
> stringbuilder over string
> "Alvin Bruney" <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote in
> message news:ebY1L7dgDH A.4024@TK2MSFTN GP11.phx.gbl...[color=green]
> > On the advice of a user, I've timed stringbuilder v string. Here are the
> > results.
> >
> > Here are the numbers:
> > Total # queries 3747
> > Time in Milliseconds
> >
> > StringBuilder: String
> > 460.6624 320.4608
> > 350.504 220.3168
> > 240.3456 230.3312
> > ----------- -----------
> > Ave. 350.504 Ave. 257.0362
> >
> > Sample query:
> > "Update ds_active_rate SET (rate, surcharge, min_sec, increment_sec,
> > begin_date, new_rate, new_rate_date, location_desc, category_id,
> > country_code, customer_number , user_name)=('0' ,'0','0','0','2 150-01-01
> > 12:00:00','','' ,'Afghanistan', 2,93,'1STAMERIC AN','') where rate_deck =
> > 'COPYRATE' and location_id = 1001 and world_id = 1 and rate_level =[/color][/color]
1"[color=blue][color=green]
> >
> > Specs on my machine
> > 1Gig SDRAM
> > P-IV 2Gig Intel
> > 20Gig Hard Drive
> > Dell C-640
> >
> > What I did: I let the routine run till it hit a breakpoint after a[/color]
> timespan[color=green]
> > object. Noted the time, and pointed execution back to the first time[/color]
> object[color=green]
> > and let it run to the break point. This gave me my three times.[/color][/color]
Obviously,[color=blue][color=green]
> > some sort of optimization was going on in the background for successive[/color]
> runs[color=green]
> > which accounts for increasingly lower times. Then I converted the code[/color][/color]
to[color=blue][color=green]
> > string and repeated the procedure. I don't claim that these numbers are
> > scientific. But I did my best to control the process.
> >
> > My conclusion is: I think these microsoft guys have been feeding us a[/color]
> bunch[color=green]
> > of bullshit with this stringbuilder efficiency crap.
> >
> >
> >[/color]
>
>[/color]

[Alvin Bruney]

Re: StringBuilder

I used the editor to replace .append( with +
compiled and fixed the compiler issues. nothing more. nothing less. I have
projects due, i can't spend 10 hours preparing a bench mark.
[color=blue]
> I strongly suggest you refactor this code in a way that lets you do a[/color]

I think you are grasping at straws here. This is the same code, one running
stringbuilder, the other running string (formatted as described above). I'm
not seeing the results that was hyped about stringbuilder straight out of
the box. I wouldn't want to COOK the results (refactor or what ever the
latest terminology is) to get these results either. If it is as good as they
say, I shouldn't be jumping thru hoops to get the results. This isn't C++.

My analogy. If car A is faster than car B then car A should be faster on an
old dirt road or the autobahn. They shouldn't have to be refactored to race
downhill to show these results.


"Jon Skeet" <skeet@pobox.co m> wrote in message
news:MPG.19da92 f765fb511b98972 7@news.microsof t.com...[color=blue]
> Alvin Bruney <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote:[color=green]
> > I apologize for not providing the code. The batch command explanation[/color][/color]
passed[color=blue][color=green]
> > over my head. I'd like you to explain this a little more. I believe the[/color][/color]
code[color=blue][color=green]
> > follows your quidelines. This is a raw cut and paste by the way.[/color]
>
> Yes - if you could change the formatting so it came out looking
> slightly more readable in future, that would be appreciated :)
>
> You've shown the StringBuilder code, but not the String code.
>
> I strongly suggest you refactor this code in a way that lets you do a
> straight, repeatable benchmark (i.e. it's a complete program which
> doesn't depend on any external factors, it has all the data constructed
> before the loop is entered, etc) and post both cases of that. Certainly
> based on that code, I would expect it to be significantly faster at
> constructing the queries than straight string concatenations.
>
> <snip>
>
> --
> Jon Skeet - <skeet@pobox.co m>
> http://www.pobox.com/~skeet
> If replying to the group, please do not mail me too[/color]

[Jon Skeet]

Re: StringBuilder

Alvin Bruney <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote:[color=blue]
> I used the editor to replace .append( with +
> compiled and fixed the compiler issues. nothing more. nothing less. I have
> projects due, i can't spend 10 hours preparing a bench mark.
>[color=green]
> > I strongly suggest you refactor this code in a way that lets you do a[/color]
>
> I think you are grasping at straws here.[/color]

I'm not. I'm suggesting that posting only *one* of the pieces of code,
and not doing scientific tests in the first place, isn't a good way of
coming up with the kind of conclusions you're doing.
[color=blue]
> This is the same code, one running
> stringbuilder, the other running string (formatted as described above).[/color]

But not as actually shown. I'd like to *see* the code. For one thing
then I could come up with some sample tests myself.
[color=blue]
> I'm
> not seeing the results that was hyped about stringbuilder straight out of
> the box. I wouldn't want to COOK the results (refactor or what ever the
> latest terminology is) to get these results either.[/color]

I wasn't suggesting "cooking" anything. I was suggesting that before
you start saying that people are lying, you should have some good
grounds.
[color=blue]
> If it is as good as they
> say, I shouldn't be jumping thru hoops to get the results. This isn't C++.
>
> My analogy. If car A is faster than car B then car A should be faster on an
> old dirt road or the autobahn. They shouldn't have to be refactored to race
> downhill to show these results.[/color]

Your analogy is horribly flawed. Car A can be faster than car B, but
without timing them properly you won't know that.

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too

[Jon Skeet]

Re: StringBuilder

Alvin Bruney <vapordan_spam_ me_not@hotmail_ no_spamhotmail. com> wrote:[color=blue]
> Now hold on. I said they weren't scientific. I prefaced the results by that.[/color]

Yes, and then accused MS of lying, in quite strong terms. Doing that
without good evidence is a really bad idea, IMO.
[color=blue]
> I don't have a lot of time on my hands to go about making sure this will
> stand up to the hounds of hell. I leave it up to MS for that.
>
> What difference does it make what i do with them. The question I was after
> was which way is faster to build strings. FYI, i take these queries and
> write to the database. I don't see what is gained by that tidbit.[/color]

Did your tests include the time taken executing the queries? If so, it
could easily be the database which was being slower, or something
similar. In order to test string vs StringBuilder performance, you
should *only* include that.
[color=blue]
> This is a real world program by the way. Not an arbitrary couple lines of
> code make up. If I can't see results there, and I should, then either I am
> doing something wrong.[/color]

If the bulk of the time is spent doing the database query, then you
*wouldn't* see much difference.
[color=blue]
> In this case point it out so I can fix the code or MS
> is on a marketing hype with this stringbuilder/.NET crap.[/color]

Or maybe you could accept that your tests were quite possibly entirely
bogus.

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too