Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.clipper.visual-objects > #4536 > unrolled thread

Speed Orderkeycount

Started byD.J.W. van Kooten <public@ic2remove.this.com>
First post2012-12-31 15:29 +0100
Last post2013-02-02 09:27 +0100
Articles 16 on this page of 36 — 12 participants

Back to article view | Back to comp.lang.clipper.visual-objects


Contents

  Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2012-12-31 15:29 +0100
    Re: Speed Orderkeycount dlzc <dlzc1@cox.net> - 2012-12-31 10:39 -0800
      Re: Speed Orderkeycount Karl Faller <k.faller_withoutthat_@onlinehome.de> - 2012-12-31 19:44 +0100
      Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-02 15:58 +0100
    Re: Speed Orderkeycount "Gerhard Bunzel" <g.bunzel@domonet.de> - 2013-01-02 13:52 +0100
      Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-03 11:12 +0100
        Re: Speed Orderkeycount "Gerhard Bunzel" <g.bunzel@domonet.de> - 2013-01-03 15:30 +0100
          Re: Speed Orderkeycount Carlos Rocha <carlos.rocha@doossier.com.removethis> - 2013-01-03 15:13 +0000
            Re: Speed Orderkeycount "Jamal" <vodotnet_nospam@yahoo.com> - 2013-01-04 23:58 -0500
              Re: Speed Orderkeycount "Willie Moore" <williem@wmconsulting.com> - 2013-01-05 15:11 -0600
                Re: Speed Orderkeycount "Jamal" <vodotnet_nospam@yahoo.com> - 2013-01-08 13:26 -0500
          Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-03 16:52 +0100
            Re: Speed Orderkeycount Carlos Rocha <carlos.rocha@doossier.com.removethis> - 2013-01-03 18:00 +0000
              Re: Speed Orderkeycount Karl Faller <k.faller_withoutthat_@onlinehome.de> - 2013-01-06 22:39 +0100
                Re: Speed Orderkeycount Carlos Rocha <carlos.rocha@doossier.com.removethis> - 2013-01-07 01:30 +0000
                  Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-08 09:03 +0100
                    Re: Speed Orderkeycount "Er!k \\/!$$Er" <nospam@nospam.url> - 2013-01-08 12:57 +0100
                    Re: Speed Orderkeycount Carlos Rocha <carlos.rocha@doossier.com.removethis> - 2013-01-08 16:26 +0000
    Re: Speed Orderkeycount Dave Francis <suilvenassociates@googlemail.com> - 2013-01-02 12:08 -0800
    Re: Speed Orderkeycount "Urs Eggmann" <Urs.Eggmann@shinternet.ch> - 2013-01-03 00:02 +0100
      Re: Speed Orderkeycount dlzc <dlzc1@cox.net> - 2013-01-02 20:29 -0800
        Re: Speed Orderkeycount "Urs Eggmann" <Urs.Eggmann@shinternet.ch> - 2013-01-03 13:53 +0100
          Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-03 14:47 +0100
      Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-03 12:04 +0100
    Re: Speed Orderkeycount "Paul D B" <polleke@NOMORESPAMhnt.be> - 2013-01-16 14:12 +0100
      Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-17 12:04 +0100
        Re: Speed Orderkeycount "Paul D B" <polleke@NOMORESPAMhnt.be> - 2013-01-22 13:16 +0100
          Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-01-22 16:02 +0100
            Re: Speed Orderkeycount "Paul D B" <polleke@NOMORESPAMhnt.be> - 2013-01-24 11:52 +0100
    Re: Speed Orderkeycount "Jamal" <vodotnet_nospam@yahoo.com> - 2013-01-25 15:19 -0500
      Re: Speed Orderkeycount D.J.W. van Kooten <public@ic2remove.this.com> - 2013-02-01 14:54 +0100
        Re: Speed Orderkeycount "Jamal" <vodotnet_nospam@yahoo.com> - 2013-02-01 09:31 -0500
    Re: Speed Orderkeycount "Jamal" <vodotnet_nospam@yahoo.com> - 2013-01-27 23:51 -0500
      Re: Speed Orderkeycount Camille Kater <katercam@pt.lu> - 2013-02-01 18:55 +0100
        Re: Speed Orderkeycount "Jamal" <vodotnet_nospam@yahoo.com> - 2013-02-01 18:06 -0500
          Re: Speed Orderkeycount Camille Kater <katercam@pt.lu> - 2013-02-02 09:27 +0100

Page 2 of 2 — ← Prev page 1 [2]


#4545

Fromdlzc <dlzc1@cox.net>
Date2013-01-02 20:29 -0800
Message-ID<6c6173d5-4edd-48ee-b301-b44adff27578@googlegroups.com>
In reply to#4544
Dear Urs Eggmann:

On Wednesday, January 2, 2013 4:02:03 PM UTC-7, Urs Eggmann wrote:
> 
> I introduced my time measurement into the Init as
> follows. All methods were very fast with 55'000
> records.  How many records do you have in your dbf
> and how many are deleded?
...
> ? SELF:oDCbBrowser1:Server:RECCOUNT

Reccount is supposed to poll only the record counter, stored in the file header.  It should be 0 mSec every time.

...
> ? SELF:oDCbBrowser1:Server:OrderKeyCount()

Apparently this causes a span of the index file.  Since the orderbag key count is stored in *its* header, this also should be 0 mSec (if it does not span the file).

...
> ? nEnde - nBeginn , ' milliSec'    // 9 mSec

Dick indicated that it got longer after the second and subsequent invocations.  Do you see this?

David A. Smith

[toc] | [prev] | [next] | [standalone]


#4548

From"Urs Eggmann" <Urs.Eggmann@shinternet.ch>
Date2013-01-03 13:53 +0100
Message-ID<akletbFbvl7U1@mid.individual.net>
In reply to#4545
Hi David,

>> ? nEnde - nBeginn , ' milliSec'    // 9 mSec
>
> Dick indicated that it got longer after the second and subsequent 
> invocations.  Do you see this?

Three subsequent calls gave 9 resp. 8 and 8 mSecs ??
(SP4b 2838)

I guess the  delay Dick observed has another reason than OrderCount alone.

Urs.

[toc] | [prev] | [next] | [standalone]


#4549

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2013-01-03 14:47 +0100
Message-ID<hn2be8hd48dlnnmla3nb49p6c02ja6dcbb@4ax.com>
In reply to#4548
On Thu, 3 Jan 2013 13:53:53 +0100, "Urs Eggmann"
<Urs.Eggmann@shinternet.ch> wrote:

Hello David, Urse,
>
>> Dick indicated that it got longer after the second and subsequent 
>> invocations.  Do you see this?
>

No, other way around: it gets faster after the second call. If I make
a new exe or reset the server/PC and re-start the exe it's again slow
the first time, and faster subsequent times.


>I guess the  delay Dick observed has another reason than OrderCount alone.


Must be. OrderkeyCount hasn't changed from SP3->SP4 but there's no
source of the underlying call: VODBOrderInfo( DBOI_KEYCOUNT, cTarget,
uOrder, @uRetVal )

I've posted the issue as well to ADS. The delay may be because of
going from ADS 10->11. But we have the impression (here) that
'suddenly' was much slower, after some Windows updates. Makes little
sense, but so it looked.

Dick

[toc] | [prev] | [next] | [standalone]


#4547

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2013-01-03 12:04 +0100
Message-ID<2voae8d37d329hp1vamivj9s00tghlgak9@4ax.com>
In reply to#4544
On Thu, 3 Jan 2013 00:02:03 +0100, "Urs Eggmann"
<Urs.Eggmann@shinternet.ch> wrote:

>Hi Dick,
>
Hello Urse, Dave,

>How many records do you have in your dbf and how many are deleded?

Thanks for your comments.

I've got 85500 records, 0 deleted. About your tests, I tried them,
here's my comment:
>
>? SELF:oDCbBrowser1:Server:RECCOUNT

This is 0  ms indeed but it gives the same result as lastrec(). It
doesn't take the amount in the current orderscope into consideration.


>? SELF:oDCbBrowser1:Server:Count()

This takes betweeen 904 and 52932 (!) ms (much faster when run on the
server, despite ADS) . It looks like it starts counting record by
record and that the result is somehow buffered when you run it a
second time with the same exe.

>? SELF:oDCbBrowser1:Server:OrderKeyCount()


This is between the 296 and 17098 ms. I agree with David who states
that this should be 0 ms too, as the result should be retrievable from
the CDX. And I am sure it worked like that for a while when I replaced
Count by Orderkeycount March this year. Maybe it always took a few 100
ms, but never the 17 seconds (or more) I get now. There must be
something changed, and the only suspects are Windows updates,
SP3->4/4a or ADS  10->11. 

Difficult where to begin. We leave showing the # of mails out for now,
not sure if trying to find the reason is worth it.

Dick

[toc] | [prev] | [next] | [standalone]


#4594

From"Paul D B" <polleke@NOMORESPAMhnt.be>
Date2013-01-16 14:12 +0100
Message-ID<50f6a72a$0$3117$ba620e4c@news.skynet.be>
In reply to#4536

[Multipart message — attachments visible in raw view] — view raw

Dick,

sorry to step in late on this thread. Have you already solved your problem?

For what it is worth, I copy this little doc I made for myself back in 2005...
It is in Dutch, sorry don't have time to translate it.


1. Row count



Als je het aantal records wil weten of tonen, rekening houdende met filter en index condities, dan is dit de meest voor de hand liggende methode:



DBServer:OrderKeyCount() methode, die onafhankelijk is van de bBrowser instellingen. 



Er bestaat ook bBrowser:GetRowcount(). Echter, deze is slechts correct indien:



- bBRowser:RowcountMode := #KeyCount

- bBrowser:RowMode := #Relative  (hoewel ik niet 100% zeker ben dat dit een must is)

- indien Optimisatie TRUE staat.  Staat deze FALSE, dan geeft GetRowCount() een verkeerde waarde. Want dit hangt nl. af van de DBServer var OrderKeyCount, en die geeft dan ook een verkeerde waarde!

Dit heeft dan weer andere implicaties, want bBRowser gebruikt dat nl. om bv. al dan niet een verticale scrollbar te tonen. Dus als als je optimisatie af hebt staan, is de scrollbar is niet meer betrouwbaar (verschijnt bv. als er niks te scrollen valt)



Besluit: gebruik nergens bBRowser:GetRowCount()







2. Performantie



Alle Orderkey* functies worden buitengewoon belastend indien Setdeleted(TRUE) en/of Setfilter() in gebruik



Dit omdat dan de hele database over het netwerk naar de client wordt gestuurd bij elke call van een Orderkey functie!

Het probleem wordt pas voelbaar bij files met een grotere size (meerdere megabytes). Het aantal records speelt minder een rol. 



Nu gebruikt bBRowser deze functies intensief (o.m. bij het scrollen) indien men Rowmode op #Relative heeft staan en/of rowcountmode op #Keycount. 



Rowmode #Relative is enkel nodig indien men een correct werkende multiple selection wil hebben. In dat geval moet men echt wel inschatten hoe groot de DBF zal worden. Indien ze meerdere MB wordt, dan moet men echt afstappen van Setfilter() en enkel indexes met FOR-conditions of orderscopes gebruiken. En uiteraard steeds Setdeleted(FAL­SE) maar dan in alle indexorders een FOR !Deleted() steken.

NB: multiple selection werkt nog wel in rowmode:#absolute, doch een block-selection met shift-click kan soms niet goed werken, afh. van de gebruikte order. 



De bBRowser:RowcountMode zet je best steeds op #RecCount van zodra de file een substantiele size heeft. 

Multiple selection wordt hierdoor NIET beïnvloed. 

Enige nadeel is dat de scrollbar ook verschijnt als er te weinig records zijn om te scrollen, maar ze werkt wel razendsnel.





Het optimisatieprobleem



Default staat Set Optimize TRUE.

Echter als een setfilter actief is, en er wordt extern (dwz in een andere instance van de server) een record bijgevoegd, dan verschijnt deze niet in de server zolang je niet een Clearfilter() en nieuwe SetFilter() hebt gedaan. 

The filter problem where new records are not visible is caused by the DBFCDX driver that optimizes the filters. You can switch off the optimizer, and then your problems will go away (but the filter will be slower :-()RddInfo(_SET_OPTIMIZE,FALSE)Note: This is a GLOBAL setting, even though there is also a method RddInfo for the DbServer class !What the driver does is this: once it has read the whole database it stores the record numbers of the records that meet the criterium of the filter. That way it doesn't have to go out and read all the records again. Robert van der Hulst



ECHTER!!



Optimisatie afzetten schept niet enkel een zeer zware performance degradatie, maar andere problemen ook; bv. Orderkeycount() is niet meer correct!!


Paul


"D.J.W. van Kooten"  schreef in bericht news:3983e852pa5rpb5avb2scr5og9t5ttu854@4ax.com... 

March this year we found that COUNT() litteraly counted all records.
We use it to show the number of e-mails in a bBrowser and obviously
this caused unacceptable delay in displaying them. We also found this
was not the case with OrderKeyCount.

However, without a change in our code, we found that the delays are
back, and this time in OrderKeyCount. This use to directly give an end
result and now it can take up to some minutes!

I am not sure when it started, SP4, SP4a, Windows update...

Any ideas about it would be appreciated.

Dick

[toc] | [prev] | [next] | [standalone]


#4598

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2013-01-17 12:04 +0100
Message-ID<82mff89pf7lbvpjc7e986p852egaphv1je@4ax.com>
In reply to#4594
On Wed, 16 Jan 2013 14:12:11 +0100, "Paul D B"
<polleke@NOMORESPAMhnt.be> wrote:

Hello Paul,
>
>sorry to step in late on this thread. Have you already solved your problem?

This is VERY interesting. Especially your remark about that
Orderkeycount causes the whole database to be moved over the network
and we have setdeleted on. That is exactly what happened -we have a
peer-2-peer config here, with my PC being the 'server'-, although I
also found that the first time OrderKeyCount also caused a huge delay
on my PC.

It isn't explained yet why it used to be fast, but your research is a
good starting point to do some testing, thanks for it!

Dick

[toc] | [prev] | [next] | [standalone]


#4627

From"Paul D B" <polleke@NOMORESPAMhnt.be>
Date2013-01-22 13:16 +0100
Message-ID<50fe830f$0$3108$ba620e4c@news.skynet.be>
In reply to#4598
Hi Dick,

Long ago I had also some unexplainable delay opening DBF's over the network.
To make the long story short: it turned out to be a defective switch.
So always check the hardware (NIC-cable-switch etc) before rewriting your 
VO-application :-)

Anyway, in your configuration you can be sure of one thing: each "filter" 
you set on the DBF (wether it is setdeleted, setfilter) will mean that the 
local client will have to do the evaluation and will need the whole DBF. If 
furthermore you are using orderkeycount (which is extensively used by 
bBrowser, as you will have read in my paper) then the DBF will be pumped 
over the network hundreds of times. Just watch the stats from your network 
card.

Using ADS will solve your problem. Otherwise, avoid filters and use 
conditional orders instead.

Paul


"D.J.W. van Kooten"  schreef in bericht 
news:82mff89pf7lbvpjc7e986p852egaphv1je@4ax.com...

On Wed, 16 Jan 2013 14:12:11 +0100, "Paul D B"
<polleke@NOMORESPAMhnt.be> wrote:

Hello Paul,
>
>sorry to step in late on this thread. Have you already solved your problem?

This is VERY interesting. Especially your remark about that
Orderkeycount causes the whole database to be moved over the network
and we have setdeleted on. That is exactly what happened -we have a
peer-2-peer config here, with my PC being the 'server'-, although I
also found that the first time OrderKeyCount also caused a huge delay
on my PC.

It isn't explained yet why it used to be fast, but your research is a
good starting point to do some testing, thanks for it!

Dick 

[toc] | [prev] | [next] | [standalone]


#4630

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2013-01-22 16:02 +0100
Message-ID<g9atf8dt2fdjlmrlij5hbbgrtg15js5pcb@4ax.com>
In reply to#4627
On Tue, 22 Jan 2013 13:16:14 +0100, "Paul D B"
<polleke@NOMORESPAMhnt.be> wrote:

Hello Paul,

>Using ADS will solve your problem. Otherwise, avoid filters and use 
>conditional orders instead.
>
That's what I expected but it doesn't. Because of the complex
combination of criteria to show, a filter was the only useable
solution. But orderkeycount is many times slower on the workstations
than on the servfer, while I would expect ADS to take care of that.

I've posted this to ADS as a bug; keep you posted.

Dick

[toc] | [prev] | [next] | [standalone]


#4652

From"Paul D B" <polleke@NOMORESPAMhnt.be>
Date2013-01-24 11:52 +0100
Message-ID<51011252$0$3117$ba620e4c@news.skynet.be>
In reply to#4630
Dick check out this group's archives, for a thread I started about 10 years 
ago: "Terrible network degradation when using Setfilter and OrdKeyNo and 
ADS"

It won't help you (there is no solution to the problem) but it might comfort 
you that others experienced the same problem (and that not much has 
changed).

What you could try is play with the read-ahead cache:
// als ADS, zet read-ahead cache op 30 recs

IF PDBAppObj():GLOcDriver == "AXDBFCDX"

hTableHandle := SELF:oDBTkmain:INFO( DBI_GET_ACE_TABLE_HANDLE )

IF hTableHandle <> 0

AdsCacheRecords( DWORD(hTableHandle) , 30)

ENDIF

ENDIF

And then really try to abandon the setfilter() because that is the real 
solution.

It is not complicated to imitate the setfilter with a temporary index with a 
for-condition (which is the same as your index experssion). I can send you a 
code sample if you like.
Anyway, in one application where I had this problem, this is what I wrote in 
my "history of changes":
- TK Overzicht: compleet afgestapt van het Setfilter gedoe omdat dit 
afgrijselijk slechte performantie gaf in combinatie met

ADS en bBRowser in #relative mode. Ik werk nu volledig met een tijdelijke 
indexfile. Omdat dit via ADS loopt, gaat dit bliksemsnel.

Deze temp CDX files staan in de \Data\Temp\ folder op de server.

Alle performantieproblemen zijn hiermee van de baan!!! OOk de 
netwerkvertraging vanwege WINXP valt nog nauwelijks op.

Een nadeel: aangezien de temp CDX file voor elke user uniek is; zal een 
record die een andere user append/wijzigt, niet meer bij de

andere users verschijnen, tenzij zijn temp CDX wordt heropgebouwd...!


Finally: even simply doing Setdeleted(true) is already a filter and will 
already degrade performance!
Solution: systematically put !Deleted() in every index condition. Then you 
can put Setdeleted(false) and the deleted records will not show up.


Paul

"D.J.W. van Kooten"  schreef in bericht 
news:g9atf8dt2fdjlmrlij5hbbgrtg15js5pcb@4ax.com...

On Tue, 22 Jan 2013 13:16:14 +0100, "Paul D B"
<polleke@NOMORESPAMhnt.be> wrote:

Hello Paul,

>Using ADS will solve your problem. Otherwise, avoid filters and use
>conditional orders instead.
>
That's what I expected but it doesn't. Because of the complex
combination of criteria to show, a filter was the only useable
solution. But orderkeycount is many times slower on the workstations
than on the servfer, while I would expect ADS to take care of that.

I've posted this to ADS as a bug; keep you posted.

Dick 

[toc] | [prev] | [next] | [standalone]


#4664

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2013-01-25 15:19 -0500
Message-ID<ke4ul6$k4g$1@dont-email.me>
In reply to#4536
Dick,

Do you have SetDeleted(TRUE) ?

If so, change it to SetDeleted(FALSE) and let me know how it goes.

From reading various news groups, this can act as a "FILTER" and can slow 
down your DBF access.

Jamal

"D.J.W. van Kooten"  wrote in message 
news:3983e852pa5rpb5avb2scr5og9t5ttu854@4ax.com...

March this year we found that COUNT() litteraly counted all records.
We use it to show the number of e-mails in a bBrowser and obviously
this caused unacceptable delay in displaying them. We also found this
was not the case with OrderKeyCount.

However, without a change in our code, we found that the delays are
back, and this time in OrderKeyCount. This use to directly give an end
result and now it can take up to some minutes!

I am not sure when it started, SP4, SP4a, Windows update...

Any ideas about it would be appreciated.

Dick 

[toc] | [prev] | [next] | [standalone]


#4702

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2013-02-01 14:54 +0100
Message-ID<l5ing811jp0frrmrsikte3am88ek7qhjgm@4ax.com>
In reply to#4664
On Fri, 25 Jan 2013 15:19:36 -0500, "Jamal"
<vodotnet_nospam@yahoo.com> wrote:

Hello Jamal,

>Do you have SetDeleted(TRUE) ?

Yes, but there were no deleted records in the database so it shouldn't
make a difference. 

I'll try your DbSetRestoreWorkarea later, and let you know if
something influences it.

Thanks!

Dick

[toc] | [prev] | [next] | [standalone]


#4703

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2013-02-01 09:31 -0500
Message-ID<kegjin$5ab$1@dont-email.me>
In reply to#4702
Hi Dick,

Yes, it does! Because it acts a "FILTER" and pulls the whole database across 
the network.

Jamal

"D.J.W. van Kooten"  wrote in message 
news:l5ing811jp0frrmrsikte3am88ek7qhjgm@4ax.com...

On Fri, 25 Jan 2013 15:19:36 -0500, "Jamal"
<vodotnet_nospam@yahoo.com> wrote:

Hello Jamal,

>Do you have SetDeleted(TRUE) ?

Yes, but there were no deleted records in the database so it shouldn't
make a difference.

I'll try your DbSetRestoreWorkarea later, and let you know if
something influences it.

Thanks!

Dick 

[toc] | [prev] | [next] | [standalone]


#4667

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2013-01-27 23:51 -0500
Message-ID<ke5047$n95$1@dont-email.me>
In reply to#4536
Dick

In your Start method, add DbSetRestoreWorkarea(FALSE)

Also, on all network computers, configure NIC cards "Interrupt Moderation" 
option to Disabled (if it has it).

Please let us know if this makes a difference.

HTH,
Jamal


"D.J.W. van Kooten"  wrote in message 
news:3983e852pa5rpb5avb2scr5og9t5ttu854@4ax.com...

March this year we found that COUNT() litteraly counted all records.
We use it to show the number of e-mails in a bBrowser and obviously
this caused unacceptable delay in displaying them. We also found this
was not the case with OrderKeyCount.

However, without a change in our code, we found that the delays are
back, and this time in OrderKeyCount. This use to directly give an end
result and now it can take up to some minutes!

I am not sure when it started, SP4, SP4a, Windows update...

Any ideas about it would be appreciated.

Dick 

[toc] | [prev] | [next] | [standalone]


#4708

FromCamille Kater <katercam@pt.lu>
Date2013-02-01 18:55 +0100
Message-ID<510C018E.4000901@pt.lu>
In reply to#4667
Jamal,

What is the purpose of DbSetRestoreWorkarea(TRUE/FALSE) ?

Camille

On 28/01/2013 05:51, Jamal wrote:
> Dick
>
> In your Start method, add DbSetRestoreWorkarea(FALSE)
>
> Also, on all network computers, configure NIC cards "Interrupt
> Moderation" option to Disabled (if it has it).
>
> Please let us know if this makes a difference.
>
> HTH,
> Jamal
>
>
> "D.J.W. van Kooten"  wrote in message
> news:3983e852pa5rpb5avb2scr5og9t5ttu854@4ax.com...
>
> March this year we found that COUNT() litteraly counted all records.
> We use it to show the number of e-mails in a bBrowser and obviously
> this caused unacceptable delay in displaying them. We also found this
> was not the case with OrderKeyCount.
>
> However, without a change in our code, we found that the delays are
> back, and this time in OrderKeyCount. This use to directly give an end
> result and now it can take up to some minutes!
>
> I am not sure when it started, SP4, SP4a, Windows update...
>
> Any ideas about it would be appreciated.
>
> Dick

[toc] | [prev] | [next] | [standalone]


#4709

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2013-02-01 18:06 -0500
Message-ID<kehhpa$g6n$1@dont-email.me>
In reply to#4708
Camille,

Setting it to FALSE disables the constant restoring of of DB work areas 
after every call in DBServer class. Just look in the RDD SDK. Which in turn 
speeds up your DB access.

According to Mr. Data "If you want to use server objects in a filter you 
need to enable the workarea restoring (using DbSetRestoreWorkArea(TRUE)) for 
the complete lifetime of the filter".

In my case I NEVER use filters. I have been using 
DbSetRestoreWorkArea(FALSE) for several years with no problem.


Jamal

"Camille Kater"  wrote in message news:510C018E.4000901@pt.lu...

Jamal,

What is the purpose of DbSetRestoreWorkarea(TRUE/FALSE) ?

Camille

On 28/01/2013 05:51, Jamal wrote:
> Dick
>
> In your Start method, add DbSetRestoreWorkarea(FALSE)
>
> Also, on all network computers, configure NIC cards "Interrupt
> Moderation" option to Disabled (if it has it).
>
> Please let us know if this makes a difference.
>
> HTH,
> Jamal
>
>
> "D.J.W. van Kooten"  wrote in message
> news:3983e852pa5rpb5avb2scr5og9t5ttu854@4ax.com...
>
> March this year we found that COUNT() litteraly counted all records.
> We use it to show the number of e-mails in a bBrowser and obviously
> this caused unacceptable delay in displaying them. We also found this
> was not the case with OrderKeyCount.
>
> However, without a change in our code, we found that the delays are
> back, and this time in OrderKeyCount. This use to directly give an end
> result and now it can take up to some minutes!
>
> I am not sure when it started, SP4, SP4a, Windows update...
>
> Any ideas about it would be appreciated.
>
> Dick 

[toc] | [prev] | [next] | [standalone]


#4710

FromCamille Kater <katercam@pt.lu>
Date2013-02-02 09:27 +0100
Message-ID<510CCDD9.6020205@pt.lu>
In reply to#4709
Thanks Jamal,

Camille

On 02/02/2013 00:06, Jamal wrote:
> Camille,
>
> Setting it to FALSE disables the constant restoring of of DB work areas
> after every call in DBServer class. Just look in the RDD SDK. Which in
> turn speeds up your DB access.
>
> According to Mr. Data "If you want to use server objects in a filter you
> need to enable the workarea restoring (using DbSetRestoreWorkArea(TRUE))
> for the complete lifetime of the filter".
>
> In my case I NEVER use filters. I have been using
> DbSetRestoreWorkArea(FALSE) for several years with no problem.
>
>
> Jamal
>
> "Camille Kater"  wrote in message news:510C018E.4000901@pt.lu...
>
> Jamal,
>
> What is the purpose of DbSetRestoreWorkarea(TRUE/FALSE) ?
>
> Camille
>
> On 28/01/2013 05:51, Jamal wrote:
>> Dick
>>
>> In your Start method, add DbSetRestoreWorkarea(FALSE)
>>
>> Also, on all network computers, configure NIC cards "Interrupt
>> Moderation" option to Disabled (if it has it).
>>
>> Please let us know if this makes a difference.
>>
>> HTH,
>> Jamal
>>
>>
>> "D.J.W. van Kooten"  wrote in message
>> news:3983e852pa5rpb5avb2scr5og9t5ttu854@4ax.com...
>>
>> March this year we found that COUNT() litteraly counted all records.
>> We use it to show the number of e-mails in a bBrowser and obviously
>> this caused unacceptable delay in displaying them. We also found this
>> was not the case with OrderKeyCount.
>>
>> However, without a change in our code, we found that the delays are
>> back, and this time in OrderKeyCount. This use to directly give an end
>> result and now it can take up to some minutes!
>>
>> I am not sure when it started, SP4, SP4a, Windows update...
>>
>> Any ideas about it would be appreciated.
>>
>> Dick
>

[toc] | [prev] | [standalone]


Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | comp.lang.clipper.visual-objects


csiph-web