mview 5.1.0 delay during compound view

Hi,


I have files with some millions of structures and mview


manages to read them within 3-4 minutes with some delay with caching the structures,


but then it takes almost hours for loading several million times the current font (in my case its C:\windows\fonts\tahoma.ttf.





Would it be possible to exclude that read for the sake of speed or


cache this font (the problem is, its 375 kByte in my case). I remember


that was not the case in older versions (but I am not 100% sure).





From the filemon analysis it only loads 3.4 kByte but this would still be 4 Mio x 3.4 kByte


and thats 13.6 Gbyte read for nothing. And it doesn't do this in an intelligent way, it really takes hours to finish.





Tobias

Dear Tobias,





We will need some time to discuss this issue. We will notify You in this topic.





Regards,


Gabor

Gabor,


that happens only in the spreadsheet view, not in the matrix view. The SDF file under investigation has some fields included.


If I turn on <show fields> in matrix options view its quite fast.





Tobias

Dear Tobias,





Could it be possible for you to give us (if its size permits then as an attachment in this forum) the above mentioned SDF file (if it is not confidential)?





Regards,


Gabor

Quote:

Could it be possible for you to give us (if its size permits then as an attachment in this forum) the above mentioned SDF file (if it is not confidential)?

... or you can send us in (private) email. We handle confideantally the provided data.


Do you know my or Gabor's email address?

Hi,


the SDs are several GByte (too much for our web server) but I included


a small 400 Mbyte file which shows the same symptoms. Its generated from SMOG2 C10H16O.sdf [31 Mbyte ZIP].





So the deal is that the SD file loads with mview in ca. 25 seconds, with a file size of


422 MB that seems ok. As a comparison a full wc (wordcount) is through in 3.6 seconds, which would reflect a read speed of ~100 Mbyte/sec.


But then the long mview read starts, loading the WIN tahoma font. That can take hours.





Another issue, in matrix view (where the font problem does no occur) it


keeps reading from the HD even if the file is fully shown, I also had some files where mview got stuck with reading EOF (these where files larger


than 2 GByte, which was sometime ago the magic barrier but not with


WIN NTFS anymore).





Tobias

Thanks for the attached example. It helps a lot for debugging.

Hi,


I tested with my other system and it seems its very specific


to my workstation which has around 12 different JAVA versions


and running java 1.5 and 1.6. Thats a mess with JAVA.


Also switching between server/client or version JAVA 1.5 or


JAVA 1.6 did not solve the problem of eternal font loading.





Using the following system I had no problem:


Product Version: MarvinView 5.1.0


Build Date: 2008-08-07


Operating System: x86 Windows XP 5.1


Java: Sun Microsystems Inc. Java 1.6.0_07


Memory: 192.6M maximum, 37.2M total, 6.1M free





and it works very quick and zooming and scrolling is very enjoyable.


The whole file is loaded under 20 seconds if scrolled


down to the very last molecule on a Core Duo 2 Ghz.








Tobias

Hi,


I updated to JAVA 1.6 sub 007 and this resolved the problem,


now scrolling to the bottom of the C10H16O takes around


14 seconds to load and show all 452458 molecules.





Also PubChem with several million molecules is no problem,


however the mview program itself could have some heuristics, so if


I want to zoom into the middle of the file, it should do so instantly


and read from there directly (instead of looping through each


and every molecule).








Marvin 5.0.1


JAVA 1.6.0_07


WinXP 32bit.








HOWEVER,


that does not solve the EOF problem in matrix view.


After switching from automatic to matrix view and scrolling around the end of the file it mostly gets stuck with reading


EOF, I think there are still some issues, especially in large files the scroll bar is reluctant to fulfill his duty.





Cheers


Tobias

Dear Tobias,


We have tested the C10H160.sdf on several machines. We have not found any problem by jumping to the end of the file in molecule matrix view.


Although, Marvin requires some time to load molecules from the end of the file but we have not seen any stucking by moving the scrollbar up and down around the end of the file.


By the way, first time when I downloaded the above file, MarvinView behaved strange, several records could not be read. Finally, I have realized that the error was in the file. It is damaged by downloading or extracting. After downloading it again, Marvin read it fine.





How did you scroll down? (I have tried dragging down the "thumb" of the scrollbar till end of the bar.)

Hi Tamas,

Tamas wrote:

After downloading it again, Marvin read it fine. How did you scroll down? (I have tried dragging down the "thumb" of the scrollbar till end of the bar.)

I think it was a very specific WIN problem, as I said a new installation of the JAVA 1.6 Server


version seemed to solve the problem. Also the scrolling is fine (I also use the scrollbar) and MView looks fine.





Thanks for checking


Tobias