Changes in Word2007 object model (Explanation of performance changes (read "degraded")

[Greg Maxey]

Fellow MVP Helmut Weber and I where comparing efficiencies of our macros for
deleting duplicate paragraphs from a document.

In this discussion (and tests) I have learned that: 1) my macro is
considerably faster that his using Word2003. 2) Helmut's is considerably
faster than mine using Word2007. 3) Both are considerably faster using
Word2003.

Assuming Word2007 is a better product than its predecessor, its seems that
at the very least both would be faster in Word2007! I am interested if
anyone knows of the changes to the object model that would account for
degrade in performance using both methods and the flip-flop in the result
with my method in Word2003 compared to Word2007.

Jezebel or Steve Hudson, I don't remember which, once explained to me how
using Do ... Loop Until x Is Nothing was faster than a For Each ... Next x.
The results in our tests bare that out in Word2003, but it seems the Do ...
Until x Is Nothing method has taken the slow road in Word2007.

Here is a sample of results observed using 201 and 801 paragraphs. The
first 200/800 paragraphs are unique. The last paragraph (201/801) is a
duplicate of the 200/800th paragraph.

Word 2003 200 800
My Method 4.375 sec 68.68 sec
Helmut's Method 7.48 sec 120.81 sec
Word 2007 200 800
My Method 14.42 sec 9.98 sec
Helmut's Method 229.56 sec 158.44 sec

If you are interested running your own comparisons, here is code for
generating the test paragraphs, my code, and Helmut's code.

Sub BuildTestParagraphs()
Dim oRng As Word.Range
Dim i As Long
Set oRng = ActiveDocument.Range
oRng.Delete
For i = 1 To 200
If i = 1 Then
oRng.InsertAfter "The quick brown fox jumped over the lazy dog." & vbCr
Else
oRng.InsertAfter "The quick and extremely agile brown fox" _
& " jumped over " & i & " the lazy dogs." & vbCr
End If
Next i
oRng.InsertAfter "The quick and extremely agile brown fox" _
& " jumped over " & i - 1 & " the lazy dogs."
End Sub

Sub GregsKillDuplicates()
Dim eTime As Single
Dim oParRef As Paragraph
Dim oParChk As Paragraph
eTime = Timer
Set oParRef = ActiveDocument.Range.Paragraphs(1)
Set oParChk = ActiveDocument.Range.Paragraphs(2)
Do
Do
'An empty last paragraph may throw an error on the last loop.
On Error GoTo Err_Exit
If oParRef.Range = oParChk.Range Then
oParChk.Range.Delete
Else
Set oParChk = oParChk.Next
End If
Loop Until oParChk Is Nothing
Set oParRef = oParRef.Next
On Error Resume Next
Set oParChk = oParRef.Next
On Error GoTo 0
Loop Until oParRef Is Nothing
Err_Exit:
MsgBox Timer - eTime
End Sub

Sub HelmutsKillParagraphs()
'AKA Makro6x
Dim t As Single
t = Timer
Dim prg1 As Paragraph
Dim prg2 As Paragraph
For Each prg1 In ActiveDocument.Range.Paragraphs
For Each prg2 In ActiveDocument.Range.Paragraphs
If prg1.Range.Text = prg2.Range.Text Then
If prg1.Range.Start <> prg2.Range.Start Then
prg2.Range.Delete
End If
End If
Next
Next
MsgBox Timer - t
End Sub

 

[Karl E. Peterson]

Assuming Word2007 is a better product than its predecessor, its seems that
at the very least both would be faster in Word2007!

Hmmmm, an odd conclusion at that, and founded on an assumption unsupported by facts
in evidence.

Is there anything, other than marketing, that would lead you to either? If
anything, history refutes the conclusion outright.

 

[Greg Maxey]

Karl,

That "assuming" remark was a thin veal for an increasing dislike of Word2007
;-)

 

[Karl E. Peterson]

That "assuming" remark was a thin veal for an increasing dislike of Word2007
;-)

Ahhhhh... Reads much better with the sarcasm duly noted. <g>

Sorry to hear of your findings, but not at all unsurprised. Take VB.NET (please!
<g>) for example. Sucks hind tit in every way, when held up against what was.
MSFT's answer to crappy performance? Nothing that a quick trip to "Bigger Hammer
Hardware" won't solve!

 

[Tony Jollans]

Hi Greg,

Ì find the results you post odd - especially Helmut's method being faster
for 800 than 200 paragraphs in 2007 - and not consistent with the statements
about which is faster in each case. Some quick tests gave me different, but
still dramatic, results:

Word 2003 200 paras 800 paras
Greg's way 2.45 secs 42.09 secs
Helmut's way 4.42 secs 76.03 secs

Word 2007 200 paras 800 paras
Greg's way 214.17 secs forever and a day
Helmut's way 6.53 secs 95.05 secs

There are many factors that affect performance. In this case I wonder (a
complete guess, I'll grant) if hardware is significant - might there be some
optimisation for dual core processors that isn't quite right? Do you have a
dual core chip? Does Word 2007 or Windows Vista do any such optimisation?

I am not surprised that Word 2007 is slower than Word 2003. I have no idea
what might be causing it in this particular case but with documents becoming
ever more complex and the reasonable assumption that users of newer software
have, in general, newer and faster machines I don't see that as a big issue,
per se.

The dramatic differences do, however, suggest that something is amiss
regardless of any particular issues related to this particular process and I
will try and dig some more when I've caught up from being at Summit (sorry I
didn't get to meet you there, btw).

--
Enjoy,
Tony

"Greg Maxey" <[email protected]> schreef in bericht
Fellow MVP Helmut Weber and I where comparing efficiencies of our macros for
deleting duplicate paragraphs from a document.

In this discussion (and tests) I have learned that: 1) my macro is
considerably faster that his using Word2003. 2) Helmut's is considerably
faster than mine using Word2007. 3) Both are considerably faster using
Word2003.

Assuming Word2007 is a better product than its predecessor, its seems that
at the very least both would be faster in Word2007! I am interested if
anyone knows of the changes to the object model that would account for
degrade in performance using both methods and the flip-flop in the result
with my method in Word2003 compared to Word2007.

Jezebel or Steve Hudson, I don't remember which, once explained to me how
using Do ... Loop Until x Is Nothing was faster than a For Each ... Next x.
The results in our tests bare that out in Word2003, but it seems the Do ...
Until x Is Nothing method has taken the slow road in Word2007.

Here is a sample of results observed using 201 and 801 paragraphs. The
first 200/800 paragraphs are unique. The last paragraph (201/801) is a
duplicate of the 200/800th paragraph.

Word 2003 200 800
My Method 4.375 sec 68.68 sec
Helmut's Method 7.48 sec 120.81 sec
Word 2007 200 800
My Method 14.42 sec 9.98 sec
Helmut's Method 229.56 sec 158.44 sec

If you are interested running your own comparisons, here is code for
generating the test paragraphs, my code, and Helmut's code.

Sub BuildTestParagraphs()
Dim oRng As Word.Range
Dim i As Long
Set oRng = ActiveDocument.Range
oRng.Delete
For i = 1 To 200
If i = 1 Then
oRng.InsertAfter "The quick brown fox jumped over the lazy dog." & vbCr
Else
oRng.InsertAfter "The quick and extremely agile brown fox" _
& " jumped over " & i & " the lazy dogs." & vbCr
End If
Next i
oRng.InsertAfter "The quick and extremely agile brown fox" _
& " jumped over " & i - 1 & " the lazy dogs."
End Sub

Sub GregsKillDuplicates()
Dim eTime As Single
Dim oParRef As Paragraph
Dim oParChk As Paragraph
eTime = Timer
Set oParRef = ActiveDocument.Range.Paragraphs(1)
Set oParChk = ActiveDocument.Range.Paragraphs(2)
Do
Do
'An empty last paragraph may throw an error on the last loop.
On Error GoTo Err_Exit
If oParRef.Range = oParChk.Range Then
oParChk.Range.Delete
Else
Set oParChk = oParChk.Next
End If
Loop Until oParChk Is Nothing
Set oParRef = oParRef.Next
On Error Resume Next
Set oParChk = oParRef.Next
On Error GoTo 0
Loop Until oParRef Is Nothing
Err_Exit:
MsgBox Timer - eTime
End Sub

Sub HelmutsKillParagraphs()
'AKA Makro6x
Dim t As Single
t = Timer
Dim prg1 As Paragraph
Dim prg2 As Paragraph
For Each prg1 In ActiveDocument.Range.Paragraphs
For Each prg2 In ActiveDocument.Range.Paragraphs
If prg1.Range.Text = prg2.Range.Text Then
If prg1.Range.Start <> prg2.Range.Start Then
prg2.Range.Delete
End If
End If
Next
Next
MsgBox Timer - t
End Sub

 

[Greg Maxey]

Tony,

My data was skewed and your right it didn't make much sense.

Actually my results where more like this:

Word 2003 200 800
My Method 4.375 sec 68.68 sec
Helmut's Method 7.48 sec 120.81 sec
Word 2007 200 800
My Method 14.42 sec 229.56 sec
Helmut's Method 9.8 sec 158.44 sec

The jist of the post was my method is faster in 2003, Helmut's faster
in 2007 and both faster in 2003. Other than being interested in what
changes caused the differences, the main question is why does it take
a new and improved product longer to perform the same task.

I don't know if I have a dual processor or not. I have a Dell
Dimension 8400 which when purchased 3 years ago I think it was near
cutting edge. Today it is probably overdue for the dustbin.

Thanks.

 

[Tony Jollans]

Thank you, Greg, it makes more sense now.

Clearly something has changed and I can't, for the moment at least, explain
it. The new and improved product has new features which, inevitably, include
a processor load so, as I think I said, slightly worse performance across
the board does not surprise, or concern, me. The different relative
performance interests me and the extreme nature of my results does concern
me.

If your computer is three years old it does not have a dual processor - I
don't know if it's significant.

--
Enjoy,
Tony

"Greg Maxey" <[email protected]> schreef in bericht
Tony,

My data was skewed and your right it didn't make much sense.

Actually my results where more like this:

Word 2003 200 800
My Method 4.375 sec 68.68 sec
Helmut's Method 7.48 sec 120.81 sec
Word 2007 200 800
My Method 14.42 sec 229.56 sec
Helmut's Method 9.8 sec 158.44 sec

The jist of the post was my method is faster in 2003, Helmut's faster
in 2007 and both faster in 2003. Other than being interested in what
changes caused the differences, the main question is why does it take
a new and improved product longer to perform the same task.

I don't know if I have a dual processor or not. I have a Dell
Dimension 8400 which when purchased 3 years ago I think it was near
cutting edge. Today it is probably overdue for the dustbin.

Thanks.