On Dec 9, 2006, at 2:01 AM, Kem Tekinay wrote:
After some experimentation, I have found the bottleneck to my code
and can
now write an RB program that runs *faster* than perl.
The problem is, it doesn't do just what I want.
Consider this perl code:
perl -le 'my $isOdd=1 ; while (<>) { chomp ; if ($isOdd) {print
('\''case
'\'' , $_) ; $isOdd=0} else {print("r = \"" , $_ , "\"\n") ;
$isOdd=1} }'
test.txt > perltest.txt
Here is the equivalent RB code:
#pragma BackgroundTasks false
#pragma BoundsChecking false
dim fIn, fOut as FolderItem
dim t as Double = microseconds
fIn = DesktopFolder.Child( "test.txt" )
fOut = DesktopFolder.Child( "output.txt" )
dim tIn as TextInputStream = fIn.OpenAsTextFile
dim tOut as TextOutputStream = fOut.CreateTextFile
dim caseStr as string = "case "
dim varStr as string = "r = """
dim closingStr as string = """" + chr( 13 )
dim line as string
dim isOdd as boolean = true
do until tIn.EOF
line = tIn.ReadLine
'if IsNumeric( line ) then
if isOdd then
tOut.Write( caseStr )
tOut.WriteLine( line )
isOdd = false
else
'line = line.ReplaceAll( """", """""" )
tOut.Write( varStr )
tOut.Write( line )
tOut.WriteLine( closingStr )
isOdd = true
end if
loop
tIn.Close
tOut.Close
t = microseconds - t
t = t / 1000000
MsgBox( format( t, "#,0.000" ) + " seconds" )
If you look at the commented lines, you will see two of the four
differences
from my original code.
First, I use multiple "Write" statements instead of concatenating a
string
and writing it once. This alone provided a 7-second speedup.
Second, I assigned the static text to variables that get written
out to the
output file.
Third, rather than testing a line to see if it's entirely a number,
I use a
switch to process every other line.
How about:
If (Asc(line) >= 48) And (Asc(line) <= 57) Then
?
This should be very fast, as Asc() is just type converting the
leftmost character of the passed-in string to an integer (in C, it
would just typecast a 'char' to 'int'), whereas IsNumeric() has
performance O(N) as - in the worst case - it has to search all N
characters of the passed in string just to see that one (the last
one) isn't numeric.
This can be improved on even further by caching the value of Asc
(line) into an Integer, then testing the integer for the range 48..57.
Fourth, I stopped replacing quotes with double-quotes.
What the last two have in common is that they replace RB functions
(IsNumeric, ReplaceAll), and that is where the bottleneck is.
Without these
calls, the RB program is slightly faster than the perl version.
With either
of these, the RB program slows down, and with both, slows down
dramatically.
The bottom line: Function calls slow down RB significantly, but,
other than
the simplest applications, there is no way to avoid them. By
definition,
until this is "fixed", RB programs will run slower than they can, and
probably slower than other, better optimized, languages.
______________________________________________________________________
____
Kem Tekinay (212)
201-1465
MacTechnologies Consulting Fax (914)
242-7294
http://www.mactechnologies.com Pager (917)
491-5546
To join the MacTechnologies Consulting mailing list, send an e-
mail to:
[EMAIL PROTECTED]
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives of this list here:
<http://support.realsoftware.com/listarchives/lists.html>
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives of this list here:
<http://support.realsoftware.com/listarchives/lists.html>
