Kombilo 0.7.1

ugoertz · #1

Hello,

there is a new Kombilo version which includes some small fixes and, mainly, comes with a Windows installer. Read more: http://u-go.net/kombilo/.

Reports about bugs/problems (and success stories) are welcome!

Best regards

Ulrich

daal · #2

Just a quick first-impression comment: The new interface is a significant improvement over the old version, and the searches seem to run a good deal faster. I'm looking forward to trying it out more extensively. Thanks!

mw42 · #3

Hi, thanks for the update on a great program.

mw42 · #4

When editing a .sgf file the ctrl click functionality does not work on the the top row (19th rank). I installed Kombilo using the Windows executable.

ugoertz · #5

mw42 wrote:

When editing a .sgf file the ctrl click functionality does not work on the the top row (19th rank). I installed Kombilo using the Windows executable.

Ouch. Thanks for the report. Will be fixed in version 0.7.2, probably some time next week.

Best,

Ulrich

TMark · #6

I am a fairly heavy user and I did twenty or so searches and wanted to clear them from the system, so I chose complete reset. Unfortunately, it seems to only clear about 3 or 4 of the previous searches, so I had to do it several times.

Best wishes.

stalkor · #7

for the updates, but just a suggestion, dont make a new thread for every small version update:)

daal · #8

stalkor wrote:

:tmbup: for the updates, but just a suggestion, dont make a new thread for every small version update:)

AFAIK, this is the first update for the Windows version since I started playing go 600 years ago.

ugoertz · #9

TMark wrote:

I am a fairly heavy user and I did twenty or so searches and wanted to clear them from the system, so I chose complete reset. Unfortunately, it seems to only clear about 3 or 4 of the previous searches, so I had to do it several times.

Best wishes.

Thanks for the report. Will be fixed, I hope, in 0.7.2.

@stalkor: Thanks for the "thumbs up". I won't open threads for bugfix releases, but I thought the Windows installer will be of interest for many people. Maybe I should have mentioned it in the title.

Best regards,

Ulrich

ez4u · **#10**

Hi Ulrich,

Thanks for the update. It is greatly appreciated! A few comments naturally. :-)

Database -> Export search results
1. The diagrams have lost the extra spaces between the dots so they are hard to read (this is OK in -> Export current position).
2. The first search results use capital letters (A, B, C) rather than small letters (a, b, c), which is the reverse of version 0.5. As a result, the search results can not be cut and pasted into L19 or SL since capitals have special meanings.
For example:

3. The total result line under Statistics is missing.
4. Version 0.7 and 0.5 give me slightly different results. This may be due to the handling of a single game in GoGoD - 1998-04-21a - between Rui Naiwei and Jiang Zhujiu. It was the world's first game of environmental go. It began with 13 passes. When I search on the empty board (to obtain the starting set of even games played on an empty board), both 0.5 and 0.7 report a total of 65,527 matches (GoGoD Christmas 2011 edition). However in the table listing next moves, 0.7 appears to count the environmental game 14 times and reports 17 games with W 4-4 as the first play. Meanwhile 0.5 does not report the game at all and shows only 3 games (yes, GoGoD contains 12 even games that show White playing the first stone!). As a result, the detailed listing for 0.7 totals to 65,540 games while that of 0.5 totals to 65,526 games. So both lists are incorrect. :blackeye:

Date Profile
This is a great addition. I have spent the last few months working on this manually in excel. However, it should be part of the Export search results file instead. Working with GoGoD (at least) the granularity needs to be higher than it is possible to show in the lower right corner area. There are other issues as well. For example...

In order to create the graph shown here, I rearranged the directories in GoGoD to give the periods that I wanted to use. I do not know how an automated selection can give an effective result for all queries. In addition, here I used all even games as the denominator. In another analysis we might want all games, all games starting with B1 on 3-4, or all games with a taisha joseki as the denominator. It would be wonderful to select a query as the "base case" for further analysis and produce statistics based on that. What do you think? :rambo:

ugoertz · **#11**

Hi Dave,

thanks for the detailed feedback.

ez4u wrote:

Database -> Export search results
1. The diagrams have lost the extra spaces between the dots so they are hard to read (this is OK in -> Export current position).

Will be fixed in 0.7.2.

Quote:

2. The first search results use capital letters (A, B, C) rather than small letters (a, b, c), which is the reverse of version 0.5. As a result, the search results can not be cut and pasted into L19 or SL since capitals have special meanings.

I will put the option to use lower case labels back in the Options/Advanced menu in 0.7.2. Currently, if you want to use it, you could edit the kombilo.cfg file (set uppercaseLabels to 0), see http://dl.u-go.net/kombilo/doc/manual.html#the-kombilo-cfg-configuration-file. You should edit this file when Kombilo is not running.

Quote:

3. The total result line under Statistics is missing.

Fixed in 0.7.2.

Quote:

4. Version 0.7 and 0.5 give me slightly different results. This may be due to the handling of a single game in GoGoD - 1998-04-21a - between Rui Naiwei and Jiang Zhujiu. It was the world's first game of environmental go. It began with 13 passes. When I search on the empty board (to obtain the starting set of even games played on an empty board), both 0.5 and 0.7 report a total of 65,527 matches (GoGoD Christmas 2011 edition). However in the table listing next moves, 0.7 appears to count the environmental game 14 times and reports 17 games with W 4-4 as the first play. Meanwhile 0.5 does not report the game at all and shows only 3 games (yes, GoGoD contains 12 even games that show White playing the first stone!). As a result, the detailed listing for 0.7 totals to 65,540 games while that of 0.5 totals to 65,526 games. So both lists are incorrect. :blackeye:

This is really a question in which way continuations are recognized. In some sense, this environmental game has several empty board positions where the next "real" move (not pass) is W 4-4. From another point of view, the pass is a continuation which Kombilo (0.5) does not consider.

Since environmental go did not really become popular, I decided that for now other things are more important. (There is a note in the documentation.)

You should be able to reproduce the Kombilo 0.5 behavior by disabling the "hashing" algorithms in the search options.

Quote:

Date Profile
This is a great addition. I have spent the last few months working on this manually in excel. However, it should be part of the Export search results file instead. Working with GoGoD (at least) the granularity needs to be higher than it is possible to show in the lower right corner area. [...]

In another analysis we might want all games, all games starting with B1 on 3-4, or all games with a taisha joseki as the denominator. It would be wonderful to select a query as the "base case" for further analysis and produce statistics based on that. What do you think? :rambo:

Your graph is impressive. I also have been thinking about a good way to expand the Date profile function in Kombilo. Right now, you can already "do it yourself" by using a Python script. You could check out the sgftree.py script. Here is an example output file of this script: http://dl.u-go.net/facing-nirensei.sgf. But of course this will require some learning, unless you are already familiar with Python.

I expect to add an interface to this script to Kombilo itself in version 0.8. I will also add the date profile to the "Export search results" function.

Best

Ulrich

RBerenguel · **#12**

Is there any way to speed up (even a little!) kombilo's start up time? In my Mac it takes 1 minute 50 seconds to start up the graphical mode, and then 50 seconds to load the database (this part is no problem, but almost 2 minutes to start it's odd, maybe something wrong in my side)

ugoertz · **#13**

RBerenguel wrote:

Is there any way to speed up (even a little!) kombilo's start up time? In my Mac it takes 1 minute 50 seconds to start up the graphical mode, and then 50 seconds to load the database (this part is no problem, but almost 2 minutes to start it's odd, maybe something wrong in my side)

That's very weird. The Kombilo window should come up immediately. Loading the databases will take some time, depending on their size and the speed of your hard disk. So the second part sounds about normal, but the first one I do not understand. I will try to find out some more about this (but without a Mac at hand it might be hard).

Best,

Ulrich

RBerenguel · **#14**

If you know how I can get some profiling working let me know and I'll look into it. Of course in a Mac the X system has to launch, but today I had it already opened, so this does not look like the issue.

Laman · **#15**

regarding the database loading at startup, was the implementation in 0.5 the same as it is now in 0.7? i don't use kombilo often, but i think that the start was faster (instant), of course for the price of slow lookups.

if there is a workaround / other setting i don't know about, please tell me and excuse my ignorance

otherwise the most comfortable solution for a user would be a fast start with the database loading in the background and finished during the first query (resp. before it, but with a waiting query already enqueued), if not earlier. but i realize this might not be the easiest thing from the programmer's point of view, so take it just as an suggestion

ugoertz · **#16**

RBerenguel wrote:

If you know how I can get some profiling working let me know and I'll look into it. Of course in a Mac the X system has to launch, but today I had it already opened, so this does not look like the issue.

I do not have a great idea how to do this. Here is one possibility that might help: Instead of kombilo.py, let us work with the viewer script v.py, so that loading the databases does not come in.

In v.py, at the very beginning of the file, add

Code:

## OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE⌴
## SOFTWARE.

import time                     # add these
t = time.time()                 # two lines

import cPickle

Then, add some print statements at the very end of the file, for example:

Code:

if __name__ == '__main__':

    print 'Step 1', time.time() - t

    root = Tk()
    root.withdraw()
    print 'Step 2', time.time() - t

    if sys.path[0].endswith('library.zip'): SYSPATH = os.path.split(sys.path[0])[0]
    else: SYSPATH = sys.path[0]

    try:
        if os.path.exists(os.path.join(SYSPATH, 'kombilo.app')):
            root.option_readfile(os.path.join(SYSPATH, 'kombilo.app'))
    except TclError:
        showwarning('Error', 'Error reading v.app')

    app = Viewer(root)
    print 'Step 3', time.time() - t

    root.protocol('WM_DELETE_WINDOW', app.quit)
    root.title('v.py')

    app.boardFrame.focus_force()
    root.tkraise()

    if len(sys.argv) > 1:              # load sgf file given as first argument
        app.openFile(os.path.split(sys.argv[1])[0], os.path.split(sys.argv[1])[1])
        app.changeCurrentFile(None, 1)
        app.delFile()

    if len(sys.argv) > 2:              # jump to move given as second argument
        try:
            for i in range(int(sys.argv[2])):
                app.next()
        except: pass
    print 'Step 4', time.time() - t

    root.mainloop()

When you invoke v.py, what does it print to the console? Maybe we can narrow down the problem a little in this way.

@Laman: thanks for your input. I agree that loading the databases should occur in the background, but it does not seem trivial to implement this because of the way the Python and C++ parts of the program are connected.

Best regards

Ulrich

ugoertz · **#17**

Hello,

I just released Kombilo 0.7.2, see http://u-go.net/kombilo/. There have been several bug fixes, and some small additions (mouse wheel support to scroll through games, "always maximize window" option (on Windows), update of references to commentaries, ...).

Thanks everybody for the feedback!

Best regards

Ulrich

RBerenguel · **#18**

Well, this pinpoints a little the slowness:

Code:

Step 1 1.80542612076
Step 2 2.39424014091
Step 3 89.2470321655
Step 4 89.2475252151

ugoertz · **#19**

Hi,

RBerenguel wrote:

Well, this pinpoints a little the slowness:

Code:

Step 1 1.80542612076
Step 2 2.39424014091
Step 3 89.2470321655
Step 4 89.2475252151

thanks for carrying this out. Unfortunately, this does not give a simple answer - the step which takes long is the one where most of the initialization process takes place. Currently the only way I see to proceed from here, would be to add further print statements to the __init__ method of the Viewer class (starting around line 2690). Even then, the end result could be that each of the initialization steps takes "too long".

It's really puzzling, how this can be so slow ...

Best regards,

Ulrich

RBerenguel · **#20**

Hi Ulrich,

I can't do it right now (I'm preparing a few network graphs in Gephi for work and will eat most of the day) but I'll set to do it when I find a few minutes to tinker with the file. I'm curious to find the slow part! I'm thinking about installing an X server and all required stuff in a remote machine (I use a 6sync VPS) to use Kombilo in my iPad and probably faster than in my Mac :/

Kombilo 0.7.1

Who is online