search

SciML / DiffEqOnline

It's Angular2 business in the front, and a Julia party in the back! It's scientific machine learning (SciML) for the web
http://app.juliadiffeq.org
Other
27 stars 6 forks source link

Issue tracking for release #8

Closed ChrisRackauckas closed 7 years ago

ChrisRackauckas commented 7 years ago

I'll put out a blog post on this and start promoting it, but there's a few things left to tackle:

I think that's enough to launch, and other things like exposing Sundials algorithms could be done later (and it'll probably make some more people interested in helping!)

amellnik commented 7 years ago

Looks good, added two and about to knock off one.

ChrisRackauckas commented 7 years ago

Nothing is complete without a favicon

Smart man.

ChrisRackauckas commented 7 years ago

There's a favicon for JuliaDiffEq:

https://github.com/JuliaDiffEq/juliadiffeq.github.io/blob/master/favicon.png

Just need to get the web server to use it.

ChrisRackauckas commented 7 years ago

I find that I am sometimes able to hit the memory quota. Maybe because maxiters is too high on the divergent runs? You can check with:

dx = a*x - b*x*y - 2x
dy =- d*x*y
ChrisRackauckas commented 7 years ago

Well I had a whole lot done and was about to call it a night, but then the last front-end that I sent to the server won't stop spinning. Could you take a look at it? I can't seem to find out what is up.

But the current state is really good (sans this spinning issue). It is responsive, I can do cool 3D Lorenz equation plots, equations which explode stop earlier and throw an appropriate error, the about page is filled out, etc.

ChrisRackauckas commented 7 years ago

Is there some kind of delay when updating the frontend? It seems that every time when I push to the server it goes down for a bit (at least, it went down for a bit and was up when I woke up, looks like it's doing it again, don't know if you did something in between)

amellnik commented 7 years ago

Yeah, there's some initial dyno boot time in addition to the julia startup costs. I haven't pushed anything to it in the last day or so. There might be something else going on too -- @abahm mentioned that it was down for him at some point when neither of us was pushing anything. I'll start looking into the logs and dashboard.

amellnik commented 7 years ago

Ah whoops I misread this the first time. I think Alan's issue was related to the back end.

ChrisRackauckas commented 7 years ago

The first plot does have a startup time, but any plot after that (for any user) should be snappy. So we just got to make sure we run it once after deployment... or maybe there's a way to just call the solve function once to force compilation for everything?

ChrisRackauckas commented 7 years ago

I am planning to upgrade right before release. I usually get around 500 blog views on a post day, but I assume that this will have an even wider reach (it's fun! And if it gets reblogged to HackerNews, then the hit couts skyrocket), so we may want to be prepared for at least a 1000 user day. After a week it'll calm down and we can then adjust it back down. I'll front the cost.

amellnik commented 7 years ago

Sure, if it really takes off I can also pitch in. We can bake a call to solveit(...) with default parameters into the main script before @sync serve(...).

ChrisRackauckas commented 7 years ago

We can bake a call to solveit(...) with default parameters into the main script before @sync serve(...).

That sounds like the right idea.

I just put a PR in for a tag of OrdinaryDiffEq.jl. It also includes something that should make the init dt special handling no longer necessary. If we just go with the plot recipe and this is gone, we can scrap a lot of the code I added last night but keep the benefits. I'll test it out soon.

amellnik commented 7 years ago

Re new item -- I'll work this in soon, but probably not blocking for release.

image

ChrisRackauckas commented 7 years ago

Is this good for now?

amellnik commented 7 years ago

Yep! The heroku timeout returns a 503 response but without the proper access headers so the client ignores it (and treats it as a 0 response for some reason) but I think I'm also handling that case now.

I have a few minor tweaks I want to make to improve the layout on mobile devices, but I don't think it's blocking.

ChrisRackauckas commented 7 years ago

Alright, I upped the plan to take a potential initial hit. Let's see what happens?

amellnik commented 7 years ago

Sounds good!

(Please excuse brevity and errors, sent from a mobile device.)

On Jan 17, 2017 1:02 PM, "Christopher Rackauckas" notifications@github.com wrote:

Alright, I upped the plan to take a potential initial hit. Let's see what happens?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/JuliaDiffEq/DiffEqOnline/issues/8#issuecomment-273298538, or mute the thread https://github.com/notifications/unsubscribe-auth/AGzU_sEgmu7L_K202uc2gkj5wKiXe_smks5rTSx7gaJpZM4LhUsF .