DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2022-10-27 00:00:00Z and 2022-10-28 00:00:00Z

k9 mail doesn't work for you?

[ddnet-maps:master] 1 new commit

69da889 A milkyway, M Depressed II - ddnet-maps

the R^2 on my huffman benchmark for compression is incredibly low but on decompression is close to 1

06:20

idk how to get it higher so its more accurate

06:20

fuckyousnail

06:21

Image attachment

Image attachment

Learath2

Do you understand groups?

The thing is, i understand what they do, like avg, sum, count, min and max, i know what they are doing but finding the right code for them is hard for me. It's probably just learning by doing.. but it's frustrating :/

https://www.cybertec-postgresql.com/en/multiranges-in-postgresql-14/

Multiranges in PostgreSQL 14 - | CYBERTEC Data Science & PostgreSQL

Multiranges are a new feature in PostgreSQL 14. Here's a brief introduction on how to use them - learn the basics & important new uses.

07:58

new

Meekrioz

The thing is, i understand what they do, like avg, sum, count, min and max, i know what they are doing but finding the right code for them is hard for me. It's probably just learning by doing.. but it's frustrating :/

Right code? Can you give an example thay confuses you? Maybe I can figure out what you are getting stuck on. Aggregates are pretty simple except for very few caveats

Learath2

Right code? Can you give an example thay confuses you? Maybe I can figure out what you are getting stuck on. Aggregates are pretty simple except for very few caveats

I will, when i get home

Image attachment

10:23

Updated teemasterparser

10:24

Now its a cli tool

10:24

Ill improve it with more stats

Ryozuki

https://www.cybertec-postgresql.com/en/multiranges-in-postgresql-14/

is this the map coke made that everyones talks about

Any rust pros, what does it even mean?

Image attachment

Looks like one too many square brackets at the start, though idk the context at all

11:33

tiles[i][j] instead of tiles[[i][j]]

Learath2

tiles[i][j] instead of tiles[[i][j]]

nah, its ndarray

Image attachment

11:42

result is Vec<Vec<bool>>, tiles is Array2<Tile>, im transmutating bool into u8

11:42

but even if i use simple u8 number, it will give same error

Anime.pdf

nah, its ndarray

oh, wait, i actually need to use tiles[[i,j]] instead of tiles[[i][j]] (edited)

11:49

oopsie doopsie

Anime.pdf

oopsie doopsie

Image attachment

I recommend the itertools crate too

11:51

Magick

Yeah, I didn't know what the type of tiles was, so just made a random guess

Wrapping a reader that has its datq in memory around a buffered reader makes sense?

13:01

Probs not

@heinrich5991 https://doc.rust-lang.org/std/cmp/struct.Reverse.html

Reverse in std::cmp - Rust

A helper struct for reverse ordering.

14:08

did you knwo about this?

14:08

its awesome

14:08

 let mut total_game_types: Vec<(String, usize)> = total_game_types.into_iter().collect();
 total_game_types.sort_by_key(|x| Reverse(x.1));

14:09

if u want to sort but in the other order, simply wrap it around that

Ryozuki

@heinrich5991 https://doc.rust-lang.org/std/cmp/struct.Reverse.html

ye, seen it

https://raw.githubusercontent.com/edg-l/teemasterparser/master/example.svg

14:32

discord cant display svg shit

need "players playing multeasymap"

14:34

14:34

but nice graph

14:34

how did you generate it?

Parsing the master srv data

14:36

Oops wrong link

14:36

https://github.com/edg-l/teemasterparser

GitHub - edg-l/teemasterparser: Parses the data from one day in htt...

Parses the data from one day in https://ddnet.tw/stats/master/ and generates svg images for each day. - GitHub - edg-l/teemasterparser: Parses the data from one day in https://ddnet.tw/stats/master...

[build]
rustflags = ["-C", "target-cpu=native"]

Yeah thats old

14:37

Not required anymore xd

that probably belongs into your user config instead

14:38

~/.cargo/config.toml

Parsing things in a streaming way is nice, but some buffering helps great to reduce time, i found a optimization where each json inside the file went from 5ms to 1ms by allocating a buffer from the known size stored in the tar metadata

14:38

1ms on a not so good cpu

ah nice

@heinrich5991 nah, its cuz i used simd_json before

14:38

Xd

does it distinguish between 0.6 and 0.7 servers btw?

14:40

probably not

Not yet

14:40

Also i found some servers dont even have a max_clients field

14:40

Gotta be rly careful parsing

Ryozuki

Also i found some servers dont even have a max_clients field

those don't get shown in the client, I think

14:41

are you sure they exist?

Well my json parser sure said so

ah. you could probably treat them as invalid

Im still amazed how well zstd compressed those json files

14:43

From 9mb to 8gb iirc

[ddnet/ddnet] Pull request opened: #5987 Fix physics change by weak hook fix (fixes #5769)

is just the fixed version #5770 but fokko disabled that i can write into his repo fixes #5769

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested ...

Ryozuki

Also i found some servers dont even have a max_clients field

oh yeah, I ran into that too, only started recently

"How about of having [...] teeworlds with client generated world where everyone would play" - allu2, 2015

16:12

make ai generated map with a seed

16:12

could have a daily changing world

16:14

could deliver tons of inspiration

16:17

or make vote option for generate new world

Was confused to see my own link on Hacker News: https://news.ycombinator.com/item?id=33358486

metadat

16:25

331 continuously open connections at 1% CPU, nice

deen

Was confused to see my own link on Hacker News: https://news.ycombinator.com/item?id=33358486

lol cool

Just when DDNet was getting close to overtaking time.gif in github stars

Image attachment

5

[ddnet/ddnet] Issue opened: #5988 default.png skin is bad, and questions on shipped skins

!image The default skin shipped with client has been faulty for a long time now. The picture is not a great visualization, but the borders are slightly asymmetrical and there is a saturated fringe on the inside. i pasted/rotated the border from the left side onto the right side to show that the border strength is different. ![image](https://user-images.githubusercontent.com/69405348/...

deen who doesn’t know GitHub has a follow feature casually flexing Stars hehe

deen

Just when DDNet was getting close to overtaking time.gif in github stars

thats a cool graph theme

Github stars just measure how useful your repo is to other developers sadly

Of course, non-developers rarely use Github

Yeah

18:30

Is quite amazing the ratio of contributors to stars is about 1:2 I wonder if ddnet has the most contributors of any open source game

i think when the others opacity changes, strong/weak indicators opacity should also change

Image attachment

iNtErSTinG

Image attachment

19:17

Is this a good place for the tiles?

Image attachment

19:18

I also thought of using it as a seperate file. Like this is supposed to be for DDNet/DDRace tiles only and mods should have extra tilesets?

On second thought I don't think that would be easily possible or possible at all (the second part not the first part) (edited)

[ddnet/ddnet] Pull request opened: #5989 Use `str_copy` instead of `str_format` with format `"%s"`

Using str_format(aBuf, sizeof(aBuf), "%s", pStr) is equivalent to str_copy(aBuf, pStr, sizeof(aBuf)). Using str_copy is more readable and also more efficient as there is no overhead from parsing the format string and from passing varargs.

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to i...

@Ryozuki do you have any good db website to search for helmet & weapons shops

20:02

(elite)

the one i sent u

20:04

https://inara.cz/elite/outfitting/#tab_outfittingequip

Outfitting [INARA]

The companion site for Elite:Dangerous. Market data, CMDR's logs, squadrons, logbooks, galleries, powerplay, engineers, blueprints, crafting, galaxy info, news and more...

ok ok

deen

Just when DDNet was getting close to overtaking time.gif in github stars

is that some kind of xkcd theme plot?

Ryozuki

Click to see attachment 🖼️

that misleading right y-scale

?

20:14

@ReiTW https://github.com/EDCD/EDMarketConnector

GitHub - EDCD/EDMarketConnector: Downloads commodity market and oth...

Downloads commodity market and other station data from the game Elite: Dangerous for use with all popular online and offline trading tools. - GitHub - EDCD/EDMarketConnector: Downloads commodity ma...

20:14

i suggest u use this

20:14

if u sync it with inara

20:14

it updates ur location on inara

20:14

so all searches are relative to ur location

20:14

otherwise u gotta do it manually

20:14

and u also contribute to updating the prices on the db

20:16

(it just parses logs the game outputs, which are quite detailed btw)

20:16

oh nvm

20:16

This app uses the “Companion” web API that Frontier originally supplied for their Elite Dangerous iOS app and now support for third-party apps. If that API ceases to function in the future then much of this application's functionality will be curtailed (although it could still utilise Journal files).

Ryozuki

?

some points look very far from each other while in reality there's less than 3% variation due to autoscaling

20:18

which is why its misleading

but it makes sense

it does, but it's still misleading

hmm well if u dont read the axes before nothing makes sense xd

people don't pay attention to those things unfortunately

but the point of that graph is to see samples that look outside the norm

you gotta account also for the visual impact

i mean

20:20

its pretty far

20:20

given that we deal with ms

20:20

i think the scale makes sense

20:21

anyway that benc has a rly low R^2

20:21

and idk how to make it bigger

20:21

from that i read a R^2 closer to 1 means more accurate

20:21

and its 0.002

20:21

xd

20:22

https://en.wikipedia.org/wiki/Coefficient_of_determination

Coefficient of determination

In statistics, the coefficient of determination, denoted R2 or r2 and pronounced "R squared", is the proportion of the variation in the dependent variable that is predictable from the independent variable(s). It is a statistic used in the context of statistical models whose main purpose is either the prediction of future outcomes or the testing ...

20:25

i wanna write a blog post for the huffman crate i made

20:25

maybe this weekend

20:26

Copium

No ddnet is not the open source game with the most contributors. For example minetest has 600 contributors but it did surpass supertuxcart! (@Tater)

Is quite amazing the ratio of contributors to stars is about 1:2 I wonder if ddnet has the most contributors of any open source game

oh nice

Well wait that repo is also kinda claiming to be a game engine. So it might not be a fair comparison.

20:33

Ah well there is osu haha https://github.com/ppy/osu

GitHub - ppy/osu: rhythm is just a click away!

rhythm is just a click away! Contribute to ppy/osu development by creating an account on GitHub.

osu best aimtrainer!

Just a few more contributors needed to get ddnet as popular as osu…. Wait what? :D

Ryozuki

and its 0.002

uh no? you got a straight line, so R^2 is really close to 1, or you're talking about something else that is not the R2 of the linear regression you're trying to fit to data

Chairn

uh no? you got a straight line, so R^2 is really close to 1, or you're talking about something else that is not the R2 of the linear regression you're trying to fit to data

Image attachment

20:59

0.00003

21:00

idk xd

how do you compute it?

i think maybe i should make the input smaller

21:02

criterion does x iterations based on a estimate of how much iterations it can do in x time

21:02

maybe if more iterations its more accurate

no, it seems to show something completely different from what a 0 R2 would be

https://github.com/edg-l/rustyman/blob/master/benches/benchmark.rs

sorry, i don't speak rust (yet)

it calles compress a lot of times

21:03

idk criterion does all

21:03

https://github.com/bheisler/criterion.rs

GitHub - bheisler/criterion.rs: Statistics-driven benchmarking libr...

Statistics-driven benchmarking library for Rust. Contribute to bheisler/criterion.rs development by creating an account on GitHub.

can't you compute it yourself?

no

21:04

im a lazy programmer

21:04

and this tool is used by everyone

just compute least square

anyway i just wanted a overview xd

R^2 is a measure of closeness between your retrofitted model and real data

21:05

what kind of model did you retrofit?

criterion does it

21:05

it warms up for 3 seconds

you don't answer my question

iguess it gets the estimate from that

21:05

i didnt do anything thats the answer xd

21:06

So, without further ado, let's start with a general overview. Each benchmark in Criterion.rs goes through four phases:

>

Warmup - The routine is executed repeatedly to fill the CPU and OS caches and (if applicable) give the JIT time to compile the code Measurement - The routine is executed repeatedly and the execution times are recorded Analysis - The recorded samples are analyzed and distilled into meaningful statistics, which are then reported to the user Comparison - The performance of the current run is compared to the stored data from the last run to determine whether it has changed, and if so by how much

21:07

The warmup period is quite simple. The routine is executed once, then twice, four times and so on until the total accumulated execution time is greater than the configured warm up time. The number of iterations that were completed during this period is recorded, along with the elapsed time.

21:07

https://bheisler.github.io/criterion.rs/book/analysis.html

Analysis Process - Criterion.rs Documentation

User Guide and Other Prose Documentation For Criterion.rs

so either you don't give him proper input data, or its doing something wrong there

21:09

okay, i think the text is just wrong

21:09

what they call R^2 is actually the director coefficient a

21:10

can you give him a linear time function with a sleep 2x ms only ?

21:10

r^2 will then be 2

21:10

so the lib is wrong

21:10

(only in the labeling part)

Image attachment

21:11

this is the decompress

it's a straight line, R^2 should be close to 1

21:12

as i said, i think what's R^2 in your screenshot is a, the director coefficient in y=ax+b

21:12

well, in second screenshot, you have a slope output

21:12

not in the first one

yeah

21:12

i think it does that when r is closer to 1

slope is 0 in the first one (what you label as R2)

but idk

okay, in the 2nd screenshot, slope looks correct (just look x input and multiply it by slope and you find corresponding y which is on the blue line)

21:14

so i guess r^2 is correct in 2nd screenshot

21:14

and 1st screenshot r2 is actually slope, hence 0 as it's a flat line

Image attachment

21:16

now it shows correct xd

21:16

it might have been a bad bench

21:16

idk

[ddnet:master] 3 new commits

d551617 Use str_copy instead of str_format with format "%s" - Robyt3 6c8fa6b Rename variable aBuf to pBuf, use size_t for buffer size - Robyt3 203fb97 Merge #5989 - bors[bot]

21:21

[ddnet/ddnet] Pull request closed: #5989 Use `str_copy` instead of `str_format` with format `"%s"`

rust looks fun, but pnet raw socket looks scary

rust looks like you're coding a fusion reactor

Image attachment

Should we periodically clear symbols for old releases? They are taking 12 GB already: https://ddnet.org/downloads/symbols/

This video is so true

deen

Should we periodically clear symbols for old releases? They are taking 12 GB already: https://ddnet.org/downloads/symbols/

This message was not shown in matrix

Exported 197 message(s)