Hacking apgsearch

[pifricted]

I met UnicodeDecodeError:

gbk.PNG (38.37 KiB) Viewed 8103 times

[Resu]

I met UnicodeDecodeError:
gbk.PNG

I think it's because the script had ' instead of " somewhere.

[pifricted]

I met UnicodeDecodeError:
gbk.PNG

I think it's because the script had ' instead of " somewhere.

No, it isn't.

After I replaced all ' to ", it's still get error.

Edit: soloved. Save the codes in gbk formart.

[PK22]

Some of the censuses are broken! There are tons of junk objects (presumably from searching bs012345678 and uploading it to the wrong census). I cannot find the haul(s) these junk objects originated from, and hence I cannot determine if it was due to a bug or the work of Uploady McGetYourOwnSymmetry.
It’s affecting the pseudo variants of C1, D4_x4, and D4_x1. I am currently searching these, but none of my hauls appear to be the issue - I checked my hauls for D4_x1, since I am the only person to have contributed (the anonymous haul was me doing testing), and no erroneous objects were there.

Please help find the issue as soon as possible!

[rabbit]

Please help find the issue as soon as possible!

The two hauls were uploaded by you, and are on this page. Here are their pages.

The false still lifes all come from the same "page" of soups in either haul. I believe this has happened with hacked apgsearch far in the past, but I do not know what it is caused by.

[PK22]

Since the censuses are now heavily contaminated, should I rename them in future versions of the script, or is there a way to remove objects from the censuses?

[Resu]

I can probably implement stdin symmetries, and I would like to add more custom 'symmetries', provided that the 'soups' can be generated by the program. If anyone has any ideas for additional features or symmetries, please post them here or PM me with information about them.

Inflated symmetries would be useful.

[PK22]

Alright, I am able to start work on development and bug fixing again.

Inflated symmetries would be useful.

I will implement these, they shouldn't be too difficult (I just need to take each cell in a cell list as (x, y) and add (2x, 2y), (2x+1, 2y), (2x, 2y+1), (2x+1, 2y+1) to a new cell list for each inflation).
I am considering implementing stdin symmetries, but the problem is that I need to find how to ensure that the RLE can be taken from Catagolue - how exactly does apgsearch v5 do this for stdin symmetries?

To prevent the nightmarish still-life issue from happening again, I will add some basic program-side verification for still lifes, oscillators, and spaceships - I can simply check that the hashes are the same for each object after one full period, and toss out the apgcodes which fail the test. My main worry is that the bug could contaminate official censuses - peer review might(?) prevent faulty objects from being uploaded, but what about for censuses which aren't peer reviewed, like b3s2/C1 or a census for another rule with <10^12 objects?

[Resu]

I think this will be useful: a way to automatically switch census so that you can see which census is good for which patterns.
(Possibly uploading a haul per each census, than repeat?)

[PK22]

(Moving discussion to a thread where it is more relevant)

I don't remember implementing storage of symmetries, what do you mean?

No, I mean storing which symmetry the user chose to search.

I never added such a thing. I did plan to have text files that stored the occurrence data of objects for b3s23 symmetries, but I never implemented it. (The idea was that the program could calculate the scores for each object on the fly based on actual data, without needing a human to manually calculate them based on their rarity relative to the pentadecathlon.) You might be thinking of the long-term caching that I had to remove, where the program stored the decompositions of b3s23 objects, with one text file for pseudo symmetries and one text file for normal symmetries.

I have implemented object verification and inflated symmetries (although they are rather slow), and once I get stdin symmetries working, I will release all the changes as version v1.2 Py3. Should I add long-term decomposition caching, storage of searching history (not to be confused with browsing history), or storage of occurrence data for v1.2?

[Resu]

I never added such a thing. I did plan to have text files that stored the occurrence data of objects for b3s23 symmetries, but I never implemented it. (The idea was that the program could calculate the scores for each object on the fly based on actual data, without needing a human to manually calculate them based on their rarity relative to the pentadecathlon.) You might be thinking of the long-term caching that I had to remove, where the program stored the decompositions of b3s23 objects, with one text file for pseudo symmetries and one text file for normal symmetries.

It must at least have a variable that stores what symmetry was entered by the user at the start of the script.

[PK22]

Oh, you mean symmstring. In v1.1.4, it is taken from the user at line 4197, and it is stored by RuleGenerator in a slashed form (B****/S****) as self.slashed (or indirectly by Soup as self.rg.slashed).

EDIT: v1.2 is out now! I believe that it might be too difficult to implement stdin symmetries, so I have skipped them for now.

[Resu]

What symmetries does apgsearch Py3 not support that is mentioned here: Apgsearch#Higher_symmetries (other than the GPU symmetries)?

[PK22]

It supports all of the non-GPU ones mentioned except for the discontinued 25pct and 75pct (but it supports 25p and 75p in place of them).

Edit: The article does mention gutter symmetries, and although a few are supported by apgsearch v5, they aren’t supported here (D2_+1_gO1s(0,1,2), with the last number representing the skew), I could probably implement them. I could also probably implement the diagonal gutter symmetries and perhaps a few others (like gutter D8_1).

[Resu]

I can probably implement stdin symmetries, and I would like to add more custom 'symmetries', provided that the 'soups' can be generated by the program. If anyone has any ideas for additional features or symmetries, please post them here or PM me with information about them.

Would it support generations rules?

[PK22]

I can probably implement stdin symmetries, and I would like to add more custom 'symmetries', provided that the 'soups' can be generated by the program. If anyone has any ideas for additional features or symmetries, please post them here or PM me with information about them.

Would it support generations rules?

That would probably take a lot of coding which I am not skilled enough to do (e.g creating the rules used for decomposing objects, adapting the function which takes an object's apgcode, etc).

[Resu]

apgsearch Py3 v1.2.1 gives this error if you don't have the python module requests:

Code: Select all

ModuleNotFoundError: No module named 'requests'

[PK22]

Ok, this is really weird. The latest version of my script is uploading erroneous hauls, because it cannot properly census switch engines. The bizarre part is that I can reproduce this error on Ubuntu, but not on Windows 11???

I will record my observations and thoughts in this post, but if you have any clue what the issue is, please let me know.
EDIT: I did a test with Wine and Windows Golly, and I ran into the same error. This suggests that the problem is with Linux.

[PK22]

Here are precompiled binaries for all pseudo-symmetries with apgsearch v5:

windows-executables.7z

(1.33 MiB) Downloaded 75 times

linux-executables.7z

(1.8 MiB) Downloaded 73 times

They upload to Pseudo_<symmetry>_Test2 (some of the original censuses are contaminated with erroneous objects).

[Disaster16439]

I want to run 3-glider-stdin, but I searched the forums and have not seen any code. I’m wondering if anyone has the code. Thanks!

[PK22]

I want to run 3-glider-stdin, but I searched the forums and have not seen any code. I’m wondering if anyone has the code. Thanks!

You can either:
a). Get the 5Glider_stdin script here, then modify the line which sets the glider count before compilation. Then run:

Code: Select all

./recompile.sh —rule <your_rule> —symmetry 3Glider_stdin
./<compiled_program_name> | ./apgluxe <arguments>

Alternatively, you can get shipcolls here, which is better as recompilation is not necessary for different glider counts, and run:

Code: Select all

./recompile.sh —rule <your_rule> —symmetry 3Glider_stdin
./shipcolls -g <rle_of_spaceship> -t 0 -r <your_rule> -n <number_of_ships> | ./apgluxe <arguments>

[Entity Valkyrie 2]

How exactly does stdin work? I don't know how to do a custom Catagolue search.

[PK22]

How exactly does stdin work? I don't know how to do a custom Catagolue search.

If you recompile apgsearch for a symmetry with “stdin” (case insensitive) in the name, then it will read RLEs from standard input and census them.
Example:

Code: Select all

./recompile.sh —rule b3s23 —symmetry test_stdin
./apgluxe -n 1000000 -k <your_key> < soup_rles.txt
./rle_outputting_program | ./apgluxe -n 1000000 -k <your_key>

See Tutorials/Catagolue stdin symmetry for more details.

[KGNINJA]

Hi Adam,

I'm KG (KGNINJA), an individual developer in Japan.
I'm building a small, cute idle game where players can opt in to let their browser do useful computation while they play.
I'd love to make Catagolue soup searching the main mission, but I wanted to ask you before doing anything.

The plan:
- Compile lifelib/apgsearch to WebAssembly and run it in a Web Worker, only after explicit consent, throttled to about 20-30% of one core, stopping when the tab is hidden.
- The searching code would be bundled and fixed (no remotely supplied code). The game itself is free.

My questions:
1. Is it acceptable to submit hauls from an unofficial WASM build? If so, what would you require (e.g. matching versions, test vectors, a specific payosha256 key, or a distinct client name)?
2. Which rule/symmetry would be most useful for a crowd of many small, short-lived browser sessions?
3. Is there a minimum haul size? Would it be OK for my server to combine many small sessions' results into one haul, or should each browser submit directly?
4. How should verification work in this setup? Can browser clients take part in peer verification?
5. Licensing: is lifelib under the MIT licence, and how would you like it credited?

One more thing, to be fully transparent: the site also sells verified computation results to AI agents for tiny amounts of USDC (a few cents at most). I would not sell anything derived from Catagolue without your OK, and I'm happy to keep the Catagolue mission purely non-commercial.

If you'd rather I didn't do this at all, that's completely fine too. Thank you for building Catagolue!

Best regards,
KG (KGNINJA)

[NNlk05]

*snip*

I did try compiling to WASM, but unfortunately it is very hard due to the inlined assembly. You are better off using the tech in https://speedydelete.com/apgweb instead. Regarding the other questions, I do not have the answers...