Committed:

* Extended Ya to unpad arrays

* YO, ZO now take one input by default and two as alternative default

Release 18.6.0:

* Corrected bug in RLE for empty input

* Function o with cell string input now pads to the right by default

* Yv (strtrim) with char array input now counts char 0 too as insignificant whitespace

* Octave compatibility: YX (regexeprep) now allows fourth input equal to 1 in Octave

* & has been defined as FT# for XS (sortrows)

* Extended Ya to unpad arrays

* YO, ZO now take one input by default and two as alternative default

* Changed order of perdefined contents of clipboard L. 1L and 2L are interchanged with 9L, 10L

@Suever The change from 1L to 9L affects this answer of yours. Just in case you want to update it. It seems not to be working anyway, surely due to other changes in the language

Oh, nvm. It doesn't affect it because you use 1L but not its predefined content

@Suever matl.suever.net seems to be giving errors. Also if I select a previous release

Oh, working now

It must have been a transient thing

It fails again

It says > error: unable to find current directory several times

@DrGreenEggsandIronMan MATL uses 32-bit integers Actually it uses double (floating point) by default. This means integers are exact up to 2^53

@LuisMendo Just fixed it

it was a directory permission issue. I changed which user is running the octave sessions and it couldn't create folders where new source code gets stored so old source code worked but new source code did not

I just updated the link on that question to a matl.suever.net link pointing to the active version when it was created

@Suever Do you think that the new online compiler can be used in general yet? I'll try the timing issue later, to see if it's gone

@LuisMendo So I was messing around with those timeouts the other day and tried upgrading the version of the web server to see if that fixed some of the issues, it still seemed kind of sporadic. I was going to do a little more debugging on that.

that's really the only thing left before we can make it "official"

there are a few things that I still want to do for displaying images but none of them are 100% necessary to launch it.

It would be nice to come up with a good challenge which utilizes some of the functionality to "release" it

@Suever I was trying now but I get 405 error. I guess you are fiddling with it

@LuisMendo Just tweaked something

@Suever Most definitely one with images. I was trying to solve flawr's algebraic curve challenge. I manage to compute the two-variable polynomial easily (gives a 101x101 matrix), but the hard part is to find the zero-cross of that

Yea that is a really good / interesting challenge

Working now

I tried several conditions to detect the zero line, but I haven't found one that works for all cases

The one I linked fails for the cross

Actually plotting the 2-var polynomial is more interesting :-)

Oooo that's pretty cool

Yes :-) Pity that the final step (finding the zero line) resists me

The problem is, it needs to be "adaptive"

I tried colouring the "n" points closest to zero

Also finding local minima, accross any of the two axes, of the abs of the function

Wait where in the challenge does it specify this?

maybe I misunderstood the question

You need to plot the implicit curve f(x,y) = 0

Oh the third test case?

I'm computing f(x,y)

Getting from f(x,y) to f(x,y)=0 as an implicit x-y curve is hard

oh ok got it

so you need the zero level set of your function

Computing f(x,y) used 4D arrays. I was so proud :-)

@Suever Exactly. Doing that numerically is tricky

And we don't have contour in MATL

No :-(

Not that I use it in Matlab, either LOL

Oh it's super handy for some of the stuff I do

like segmentation etc.

I was trying to see what algorithm contour uses internally. But the actual algorithm must be somewhere else

The source code is too short

yea probably mex

could you mess with the colormap?

to make ~0 equal to white?

@Suever The problem is, I don't need to find exactly zeros, but values close to zero. How close? That's the hard part. They are not exact zeros because the domain is sampled (an x-y grid)

right

Probably there's no value that is exactly zero

and I guess "how close to zero" can vary across the domain

Yes. And across cases

The example I linked works perfectly for the parabola, but not for the cross

Setting to 1 all points below a quantile is not a good idea either. The quantile would depend on the case too

(depending on line length for example)

some skeletonize functions would come in handy with this

"Skeletonize" sounds very close to what I want, yes (from my non-image processing experience)

so you basically want the local minima of the absolute value of a function

Ugh. tricky

but contour would be perfect. Maybe we need to add to MATL :)

@Suever We'll add that to the other image processing functions. You (and maybe Ray) need to do that list :-) Maybe 5 or 6 functions, the most common ones. Anyway, no hurry, I still have some other things pending

I know very little about image processing. Some of the bw... ones seem to be often used

Yea I definitely want to add some of those. That would make MATL a very strong contender in image processing questions

since there aren't many (any?) golfing languages that do advanced image processing well

Also on the website, I added a few debugging outputs to the javascript console. If it's not responsive, you should be able to pop the console open and see if it even connected to the server

basically it should show "connected" when the socket connects, followed by a UUID for the session, and then tell you when a job is complete and when it received output

Apparently this is a widely reported issue but my server configuration SHOULD be correct based on all posts I can find about it, but I just want to narrow down at what point it's failing

@Suever Yes, it would be a good advantage. Not many languages with those features

@Suever Ok, if I see the issue I'll look at that

@Suever I've just had that issue. The "error console" tab shows nothing. I can't see my IP (79....172) in https://matl.suever.net/admin/

The task doesn't appear in https://matl.suever.net/admin/tasks either

(latest one is the algebraic curve code)

by error console you mean chrome's error console right?

when you go to the more tools->developer tools in your browser

Hm I meant the tab called "error console" on the page https://matl.suever.net/, visualized using Chrome

Let's see that

oooohhhhh no

sorry yea I meant the browser's error console

Ok, so what should I look for there?

I see a toolbar with "elements", "console", "sources", ...

Console is the one you want

that should show info about being connected

if when you get no response there is no input that's one thing

but if it connects, whatever output does/ doesn't appear can tell me where it fails

It says

Connected.
common.js:3 adcd0380362340908ffc0cd483781d06

ok and then just hangs like that when you click the run button?

Yes. The "Run" text on the button becomes "Kill" and stops there

Ah ok so it's an issue with when the client submits the job

not the initial connection, that should help

Do I try with Firefox?

Yea you could try firefox

it's just strange that it's intermittent like this

Opening

Same

Firefox console says

"Connected." common.js:3:35
"d0d92e1539ae4a4ebadf597f1c8ad1cc" common.js:3:123

while the "Run" button reads "Kill" and no output has been produced

At least it's consistent accross browsers! :-)

So this is when you're starting from a new session or when you click a link which already has code in it

New session. I open the page, paste the code from scratch

ok well at least I know where to look

I'll have to dig into it a bit

The "?" button (parse with auto comments) doesn't work either when this happens

It opens the subwindow "Explanation", and says "Parsing..." all the time. Usually it takes a second or less and the parsed code appears

I got a new message in the Firefox console

pasting screen

Sorry, that's Chrome, not Firefox

Oh found the problem....

there are like a million octave sessions running

Haha

The server must be sweating

I have to go. Later!

Alright @LuisMendo thanks for the help. I'll do a reboot and then some debugging

@LuisMendo Ok should be good to go again. I'll have to do some testing to figure out where all those sessions came from. Basically it gobbled up all the RAM

@Suever Thanks to you! It's working now. I'll let you know if something happens again

@Suever I know how to find the zero-level curve! Just remove the abs and detect a sign change along any of the two axes. The code says only the case when the signa changes needs to be considered. It works for all cases

Example

I'll post it soon

No using this compiler yet, right? It's tempting... :-)

You can use it!

this is a gerat use

You could also use 0YG instead of 3YG if you want to get rid of the white border

@LuisMendo It'll be a good stress test for the server

@Suever Much better, thanks!

@Suever I'm thinking in including something in my answer like "Since this compiler is still in an experimental stage, please report any issues to @Suever in the MATL chatwoom.". Do you find it appropriate?

BTW is it "in a stage" or "on a stage"? English "in/on" always confuses me

I think it's fine as is except "chatroom" instead of "chatwoom" :)

Hahah

Thanks

So "on" or "in" (stage)

Is the share button working for you?

to get an imgur link

"in"

@Suever Hm I didn't remember that. It seems not to work

yea just realize that's not working. Was just trying that out because I wasn't sure if you needed that to post an image in your post

It's really only important for posting now testing it out

So not a show-stopper if you have an image already for your post

I'll have to dig into it, it's either an issue with IMGUR or with the CSRF token

@Suever Yes, since the link works, pasting the image is not that important. That way people will click on the link to the actual compiler :-)

Oh ok. So it seems like it's actually a CSS issue

I hate CSS

should be fine

I'll just hide the link for now :)

Oh god now docs aren't working lol

Ok they're back

damn permissions

@Suever Done!

Beautiful!

^^ @Suever

Lol

Such a high bar

I still get the error ("Run" not doing anything) sometimes. Refreshing once or twice solves it

Does the console say "Connected." in those cases?

I think it did, with a numeber-and-letter code. Let me check

Now it's working every time :-)

Of course haha

:-)

It's definitely some nginx thing. I'll look into it more for sure. We can't have all those false starts

Now I caught it. It didn't work, and the console says

main.js:29 Connected.
main.js:36 dc8f43a5148b47c186a2ca0809f8a003

Now it says more

Let me get the image

Hope that means something to you @Suever. No idea here :-P

I have to get going again, this time for longer. Bye! Good afternoon/evening there!

Alright thanks for the info!

@LuisMendo Figured out the issue why the number of octave sessions were blowing up. For some reason the octave->python adapter would just hang when sourcing the .octaverc file when a worker was trying to restart itself so the restart process would time out and the worker manager would keep starting and killing the worker process but not killing the underlying octave process.

Should be all fixed now

Still haven't gotten to the bottom of the jobs not being submitted right away yet