apgsearch v3.1

[muzik]

Noticed something else: the output after compiling is this

Code: Select all

$ sh recompile.sh
Skipping updates; use --update to update apgmera automatically.
Rule unspecified; assuming b3s23.
Symmetry unspecified; assuming C1.
Configuring rule b3s23; symmetry C1
recompile.sh: line 65: python: command not found

[Apple Bottom]

I was meaning that about the ./apgmera command, I already knew about the name. (Isn't Mera also a real name?)

I'll see if I can compile it again. Is there a way to change the name of the zip folder? Typing that cffe-lotsofnumbers is honestly kind of exhausting.

EDIT: In case this is a problem, I didn't run "unzip apgmera.zip", but rather the apgmera-master-thatstupidlineoftext.zip. Inputting the latter allowed the process to continue up to the point I described earlier.

EDIT2: Looks like I shouldn't have gone past that point. Typing the latter by itself doesn't work either, but with the file adress it worked.

Re: the name: "." refers to the current directory (the same way that ".." refers to the parent directory), and "/" is the path separator (Windows uses "\" instead, for historic reasons). The reason you have to type "./apgmera" and not just "apgmera" is that the current directory's not usually included in the search path (and on multi-user systems like Unix that's for a very good reason).

Re: renaming the file: use the "mv" command. Also, try hitting Tab after typing part of a file's or directory's name to complete it. (Very handy feature, that.)

Noticed something else: the output after compiling is this

Code: Select all

$ sh recompile.sh
Skipping updates; use --update to update apgmera automatically.
Rule unspecified; assuming b3s23.
Symmetry unspecified; assuming C1.
Configuring rule b3s23; symmetry C1
recompile.sh: line 65: python: command not found

There we go. You need to install the "python" package - apparently that's not a default in Cygwin either.

[muzik]

Looks to be working pretty much perfectly now!

In all honesty, though, we really kind of need a YouTube tutorial for this

[muzik]

Definitely seems to be working. Uploading some 10M soup hauls for b3/s12 C1 (and kind of encouraging everyone else to do so because I want to confirm the existence of spaceships from asymmetric soups)

[Apple Bottom]

Definitely seems to be working. Uploading some 10M soup hauls for b3/s12 C1 (and kind of encouraging everyone else to do so because I want to confirm the existence of spaceships from asymmetric soups)

Glad to hear it's working!

And you mean this ship? I'm going to go out on a limb and predict that won't make a C1 appearance anytime soon.

[muzik]

And you mean this ship? I'm going to go out on a limb and predict that won't make a C1 appearance anytime soon.

That one's probably the one I'm most thinking of. I also doubt it'll appear within the next few months, but we'll see.


Why doesn't apgsearch come as a standalone exe application? With a gui? Would that be too much work?

[calcyman]

Why doesn't apgsearch come as a standalone exe application? With a gui? Would that be too much work?

It wouldn't be able to change rules. Basically, in order to simulate cellular automata really quickly, the rule needs to be hardcoded into the program. So whenever the rule needs to be changed, the program needs to be able to recompile itself. (An alternative would be self-modifying code, but I don't think that's possible in a modern operating system.)

[muzik]

When I try to apgsearch b3478/s5678, it appears to search fine, but upon completing a haul it gives something along the lines of this, if I remember:

Code: Select all

Bad status 500
Connection unsuccessful. Continuing search...

Now the recently updated section for this rule does get bumped to the top of the list, and for a while has an uncommitted hauls link (which is an empty table), so there is definitely some connection being made.

Other rules appear to work fine.

What the heck is going on?

[calcyman]

When I try to apgsearch b3478/s5678, it appears to search fine, but upon completing a haul it gives something along the lines of this, if I remember:

Code: Select all

Bad status 500
Connection unsuccessful. Continuing search...

Now the recently updated section for this rule does get bumped to the top of the list, and for a while has an uncommitted hauls link (which is an empty table), so there is definitely some connection being made.

Other rules appear to work fine.

What the heck is going on?

You're trying to upload hauls which exceed one megabyte in size, thereby causing an 'entity too large' exception.

[muzik]

You're trying to upload hauls which exceed one megabyte in size, thereby causing an 'entity too large' exception.

Suspected something like that. On the contrary, I'm well able to send massive hauls of b35678/s because of its high death rate. Honestly the console is really weirdly satisfying to look at, with the rapidly changing numbers and stuff

[Apple Bottom]

It wouldn't be able to change rules. Basically, in order to simulate cellular automata really quickly, the rule needs to be hardcoded into the program. So whenever the rule needs to be changed, the program needs to be able to recompile itself. (An alternative would be self-modifying code, but I don't think that's possible in a modern operating system.)

The GUI could be separate from the backend. Simply supply one binary per rule/symmetry combination, and invoke the appropriate one -- after all there's only 2^18 possible rules, and 17 symmetries, so that's only about 4.5 million different binaries...

...on second thought, scrap that idea.

Unrelated -- I've noticed that Catagolue's not (yet?) committed some of my recent Day & Night hauls. It's skipped a few, then started committing them again, then skipped some more, etc. Other rules are also affected.

I don't know if it's related to Day & Night, but I've stopped all activity on that rule for now just to be on the safe side and avoid overloading the site even further, and will focus on less productive rules; would be nice to get your OK to start again when this is taken care of.

(I've been thinking Day & Night's thoroughly searched by now anyway, but I'd like to take it to 100 billion objects before stopping for good.)

[drc]

Will apgmera ever support non-totalistic rules?
Edit: Ah, I see. But what's stopping it from using ruleloader?

[muzik]

Unrelated -- I've noticed that Catagolue's not (yet?) committed some of my recent Day & Night hauls. It's skipped a few, then started committing them again, then skipped some more, etc. Other rules are also affected.

I don't know if it's related to Day & Night, but I've stopped all activity on that rule for now just to be on the safe side and avoid overloading the site even further, and will focus on less productive rules; would be nice to get your OK to start again when this is taken care of.

This (or something similar) is happening to a lot of rules. Such as b3s12, b3s01367, and (boom) b3s23.




EDIT: now all rules are doing it?

[calcyman]

Everything is broken!!!

Basically, the Day&Night census is simply too large for Catagolue to store in working memory without crashing. I'll need to make some rather drastic modifications to the codebase before Catagolue can resume operation. And since someone else is cohabiting my bedroom at the moment, it might take several days before I have time to do so.

Once things are working again, there should be a whopping spike in the number of objects in the main census (i.e. several trillion objects committed to b3s23/C1 in a single day!). I imagine this should be rather fun to observe.

[drc]

Can we still search?

[muzik]

Basically, the Day&Night census is simply too large for Catagolue to store in working memory without crashing. I'll need to make some rather drastic modifications to the codebase before Catagolue can resume operation. And since someone else is cohabiting my bedroom at the moment, it might take several days before I have time to do so.

I thought it was my fault because I left my computer apging for a good few days while I was away.

It was uploading big hauls for b3478s5678 like every 30 seconds...


I've stopped searching in the meantime. Is it safe if I continue to upload?

[calcyman]

I've stopped searching in the meantime. Is it safe if I continue to upload?

Probably not. The crash was due to a combined effort from b3478s5678 and b3678s34678.

[Apple Bottom]

Basically, the Day&Night census is simply too large for Catagolue to store in working memory without crashing. I'll need to make some rather drastic modifications to the codebase before Catagolue can resume operation. And since someone else is cohabiting my bedroom at the moment, it might take several days before I have time to do so.

Oh, that's not good. I hope you'll be able to get this fixed soon.

For the time being, as some intermediary bandaid, would it be possible to simply blacklist Day & Night (C1 at the very least) to keep it from being updated -- i.e. just skip uncommitted Day & Night hauls and don't touch its census data? I don't know how Catagolue works behind the scenes, but I imagine that this way, other rules could keep getting updated as normal while you're working on a more permanent fix.

FWIW I've stopped all my searchers for the time being.

[muzik]

Probably not. The crash was due to a combined effort from b3478s5678 and b3678s34678.

Ok I've stopped searching that rule, and continuing on with b35678s. Is that safe?

[calcyman]

For the time being, as some intermediary bandaid, would it be possible to simply blacklist Day & Night (C1 at the very least) to keep it from being updated -- i.e. just skip uncommitted Day & Night hauls and don't touch its census data?

I've done that for both b3678s34678 and b3478s5678; Catagolue seems to be working properly again now.

Of course, this is only a temporary solution, and I should really make the changes to allow supermassive censuses to be handled without loading all tabulations into memory simultaneously. Day&Night is actually a good test case for this; once I get it working again without running out of heap space, it will mean that b3s23/C1 could safely expand to quadrillions of objects.

[Apple Bottom]

I've done that for both b3678s34678 and b3478s5678; Catagolue seems to be working properly again now.

Of course, this is only a temporary solution, and I should really make the changes to allow supermassive censuses to be handled without loading all tabulations into memory simultaneously. Day&Night is actually a good test case for this; once I get it working again without running out of heap space, it will mean that b3s23/C1 could safely expand to quadrillions of objects.

Excellent! I've just started my B3/S23 searchers again. (Will still hold off on Day & Night until you give the green light.)

Quadrillions of objects, here we come!

[muzik]

I still can't help feeling that this is my fault though

[dvgrn]

I still can't help feeling that this is my fault though

Absolutely! Very good work finding and pointing out a serious vulnerability.

It's much better to get these little problems out of the way before some Catagolue-related news item catches the attention of Hacker News or some similar news site again, and there are suddenly a hundred or a thousand times as many apgmera users. Wouldn't want them all to lose interest and disappear as quickly as they showed up, because some little technical issue like this brought Catagolue to a screeching halt.

[Apple Bottom]

Absolutely! Very good work finding and pointing out a serious vulnerability.

It's much better to get these little problems out of the way before some Catagolue-related news item catches the attention of Hacker News or some similar news site again, and there are suddenly a hundred or a thousand times as many apgmera users. Wouldn't want them all to lose interest and disappear as quickly as they showed up, because some little technical issue like this brought Catagolue to a screeching halt.

I'll also take some of that praise. Who knew that obsessively searching Day & Night would help bring Catagolue down, and who knew that this would be a good thing? It's true, though: that which does not kill us makes us stronger.

While we're on the subject of things breaking, getting a /textcensus for b3678s34678/C1 has been broken since at least May 1st, likely for similar reasons.

[muzik]

I still can't help feeling that this is my fault though

Absolutely! Very good work finding and pointing out a serious vulnerability.

It's much better to get these little problems out of the way before some Catagolue-related news item catches the attention of Hacker News or some similar news site again, and there are suddenly a hundred or a thousand times as many apgmera users. Wouldn't want them all to lose interest and disappear as quickly as they showed up, because some little technical issue like this brought Catagolue to a screeching halt.

Guess it's time to commit some really obscure crimes that nobody thought about doing then.

In all seriousness, I've never been more proud of being an idiot.


Is it simply because of the amount of variation in objects, and the rapid rate that new objects are being added, that the site brokeded? Because there's quite a few rules that fit that criteria. Might need to stabilise a few of them.


Also, why do rules like b/s012345678 take absolutely AGES to search?