Sneak Peek

daniel_the_smith · **#41**

John Fairbairn wrote:

I would have preferred that Daniel made this clear himself, but can I just mention that GoGoD has indeed given permission in principle. Daniel and T Mark will iron out the details at the US Congress, but the general thrust is that we will expect a restriction on the number of moves per game made available, and also on the amount of header data supplied.

We welcome creative uses of our database. We have alreaady given similar permission to quite a range of people, companies and academic institutions and (I think - T Mark can confirm) we have never charged. But in return we do expect some sensible restrictions, and all the features requested here must go through that filter.

This was the first question asked, if you look back at the beginning of the thread.

I didn't state I had permission to begin with because I sorta figured it went without saying... In retrospect, perhaps that was silly of me. I don't think anything I've described here is in conflict with the email T Mark sent me back when I originally asked if this was ok. Anyway, if people want to review the games, they can buy GoGoD for themselves-- whatever my site ends up like it will definitely not be a pro-game-viewing site. T Mark and I will hash out any remaining details at congress. I very much appreciate the amount of work you guys have gone through to put GoGoD together-- hopefully people realize I won't be including features that reduce the value of buying their own copy.

lorill wrote:

Use title for tooltips instead of alt.

Thanks! I will probably start using that later.

ethanb wrote:

I found it browsing through taisha variations - one pattern was played with minor variations in only 3 games, but in places where all the "next moves" were the same move, or if I went to a track with only one game, the ghost stones weren't rendered.

Right now I have it set to stop showing the averaging when there's fewer than 5 games to average together. I'm not sure, but it sounds like that is the cause of what you're seeing?

CarlJung wrote:

Put link in sig.

Haha, ok.

@everyone:
My wife got home and I was able to try the new stuff in safari on her mac. Apparently the ugliness is limited to the iPad, so I updated the main site. Should be faster now.

(I don't know why my iPad is showing things wrong-- it's grabbing pieces of other parts of the sprite image, so you see like an extra row of pixels at the edges of the letters and stones. I suspect it's related to the scaling it does-- I think they're clipping the source image after scaling instead of before. If so, bad, Apple, bad! Hopefully I can find a way around it...)

kokomi · **#42**

Hey, i like it. But it's a bit slow? Or maybe my internet is slow :-?

kokomi · **#43**

wineandgolover wrote:

Daniel,

This is very impressive. Well done.

The first upgrade I would recommend is filters. Examples include "date range" "country" and "player". For example, this would be a great way to determine how the Koreans are currently answering the one space high approach pincer, rather than seeing a percentage that includes potentially obsolete joseki.

Also, I am a bit slow and am not sure I 100% understand how the surrounding positions works. Let's say that in a joseki played 100 times, the opposite starpoint was black 25 times, white 25 times, and empty 50 times. How would that display? If it was 50 black and 50 white, would it look the same? 50 white and 50 empty? 75 white, 25 empty?

I am sure this would be hard, but it would be cool if you could hover your pointer over a ghost stone and it would tell you, "black 25%, white 25%, empty 50%.

Will your db update as GoGod updates?

Thanks for sharing. Let me know if you need a beta tester for future features.

I'll keep playing with it and let you know if I have other ideas.

See you at the congress.

Again, great job!

-Brady

I'd love a filter on komi. But then black or white matters, which is too much work.

daniel_the_smith · **#44**

kokomi wrote:

Hey, i like it. But it's a bit slow? Or maybe my internet is slow :-?

I seem to recall you're in China-- if so, you're a loooong way from the server, which could explain it. Unfortunately there wouldn't be much I could do about that at the moment.

lorill · **#45**

ethanb wrote:

daniel_the_smith wrote:

Hmm, do you have an example position?

Assuming that's addressed to my question about the ghost stones, it happens on anything where there's only one path (no matter how many times it was played.)

I found it browsing through taisha variations - one pattern was played with minor variations in only 3 games, but in places where all the "next moves" were the same move, or if I went to a track with only one game, the ghost stones weren't rendered.

Example:
http://dailyjoseki.com/browse/bwpwxibuzwwjbyfwvsbuc

Edit: oops, just read the above explanation. Indeed, there's less than 5 games.

ethanb · **#46**

daniel_the_smith wrote:

ethanb wrote:

I found it browsing through taisha variations - one pattern was played with minor variations in only 3 games, but in places where all the "next moves" were the same move, or if I went to a track with only one game, the ghost stones weren't rendered.

Right now I have it set to stop showing the averaging when there's fewer than 5 games to average together. I'm not sure, but it sounds like that is the cause of what you're seeing?

Yeah, I guess so (did that behavior change some recently? When I made that initial post, it seemed like the trigger was different) I guess I don't understand why this is a feature and not a bug.

Li Kao · **#47**

daniel_the_smith wrote:

kokomi wrote:

Hey, i like it. But it's a bit slow? Or maybe my internet is slow :-?

I seem to recall you're in China-- if so, you're a loooong way from the server, which could explain it. Unfortunately there wouldn't be much I could do about that at the moment.

It's pretty slow for me too, and I'm from Germany. After a click it takes about two seconds for the stone to appear. My ping to dailyjoseki.com is 170ms and I'm using opera 10.60.

daniel_the_smith · **#48**

Hm, each click right now requires about 20-40k of html as a response from the server, that's probably the source of Li Kao's slowness if not kokomi's. There are a few things I can try to do to get that reduced to a more reasonable number. I'm guessing it goes a lot faster if you get to a position that doesn't show the ghost stones?

daniel_the_smith · **#49**

Ugh, it's pretty broken in IE right now. I'll (try to) fix it later. For now, use a real browser.

Is there a convenient way to test with IE when you're developing in ubuntu?

fwiffo · **#50**

VMWare is your best bet. Not convenient, I'll admit.

amnal · **#51**

If you consider wine to be convenient, IE seems to at least work in it. I used playonlinux to install it without the fiddliness of setting the twiddly bits manually, and it seems to at least run.

I'm not sure if this works for the most recent versions.

kokomi · **#52**

daniel_the_smith wrote:

Hm, each click right now requires about 20-40k of html as a response from the server, that's probably the source of Li Kao's slowness if not kokomi's. There are a few things I can try to do to get that reduced to a more reasonable number. I'm guessing it goes a lot faster if you get to a position that doesn't show the ghost stones?

maybe you can make it memorise somewhere, so when someone makes a request by moving a stone, it will load from memorise, rather than do all the process again? And at the back of it, you can add the requests into your database. hmm.. anyway, i'm very noob on computer...

daniel_the_smith · **#53**

kokomi wrote:

maybe you can make it memorise somewhere, so when someone makes a request by moving a stone, it will load from memorise, rather than do all the process again? And at the back of it, you can add the requests into your database. hmm.. anyway, i'm very noob on computer...

It sounds like you're talking about caching-- I believe the slowness is caused by the total amount of data that needs to get sent, not the server processing time, so I don't think that would help in this case.

I enabled compression on the test server, which packs things down quite a bit. Can those of you in China and Germany (or anywhere else) tell me if

http://dailyjoseki.com:30080/browse

is faster than

http://dailyjoseki.com/browse

? Thanks

Li Kao · **#54**

I notice no real difference. Feels like the one at :80 is even a bit faster, but probably they're about the same. And a few dozen kB shouldn't take much time. I normally can DL about 1MB/s.

Harleqin · **#55**

I cannot really tell the difference. The number of ghost stones has a noticable effect, but not which version I use.

daniel_the_smith · **#56**

Hm, thanks guys.

Do either of you (or anyone else someplace distant) have the firebug extension to firefox (or equivalent) installed? It has a net panel which breaks down the time spent waiting. I'm curious if the majority of the time for you is spent "waiting" or "receiving". When I run, I see total times 200-600 ms, with the majority spent "waiting".

I will have to add some profiling to the server just to make sure it's not something about my code, but I don't see why my server would execute slower for you than me...

daniel_the_smith · **#57**

OK, I've figured out what's taking the time, and yes, it is the server. I've done one thing that makes a slight improvement (dailyjoseki.com:30080 if you want to check). I will probably have to change the way I'm doing things a little (I'm stressing out golang's template package

) to get the speed I'm after. It is handling most requests in < 400ms (which I consider hideously slow, but better than the 600ms it was); if two people are using it at the same time, it will obviously be slower.

Also on the 30080 test server: I've added IE's filter:alpha(x) (still untested) and removed the 5 game minimum requirement for showing ghost stones. I'm not sure if I like that or not, it's kinda distracting when it shows two games and you're trying to decide which stones go with which game. I didn't have it enabled originally because I figured five games is probably not a very representative sample. I wouldn't mind hearing people's thoughts on this.

Dusk Eagle · **#58**

Hey Daniel,

I just a have a mild bug to report. When browsing through a variation and hitting 'pass,' and then hitting 'back' (the website's back button, not the browser's), the website takes you back two moves (the move before the pass) instead of just taking back the pass itself. I think it would make more sense to simply take back the pass. Also, if you play a variation and then pass -> respond -> back, the website takes you back to before the pass.

I'm not sure if I'm making myself clear here, do you understand what I mean?

daniel_the_smith · **#59**

Dusk Eagle wrote:

Hey Daniel,

I just a have a mild bug to report. When browsing through a variation and hitting 'pass,' and then hitting 'back' (the website's back button, not the browser's), the website takes you back two moves (the move before the pass) instead of just taking back the pass itself. I think it would make more sense to simply take back the pass. Also, if you play a variation and then pass -> respond -> back, the website takes you back to before the pass.

I'm not sure if I'm making myself clear here, do you understand what I mean?

Yeah, I know what you're talking about. I will change that, most likely when I get back from go congress.

everyone:

I updated the main server (the :30080 server is no longer running). IE should work now (still untested but I'm emitting the right thing now, I think). I found a way to shave 100-200ms off each request, so it is slightly faster.

Li Kao · **#60**

The back buttons stopped working for me.

Sneak Peek

Who is online