« first day (2506 days earlier)      last day (2279 days later) » 

02:06
I think that redirect to github release may better
I actually want a Continuous Integration solution that automatically updates the version number, separates and packages the document package and the minimum package, and automatically updates the version number.
I used to write one in the Shell, but I think it can support more complicated settings with high-level languages like NodeJS.
The release page follows some kinds of certain specifications, and your crawler can crawl the latest information and download links.
 
2 hours later…
04:06
@GalAster interesting... that is something I could conceivably support, but I don't want to necessarily enforce? Also we might be in the same boat, as if we're calling PacletInstall on the release object I'm not sure the download counter gets triggered.
How are you imagining the CI would work? What I mean is it's not exactly trivial to get the version info and such things from the .paclet file.
 
1 hour later…
05:33
Ugh. There's no good way to get a redirect that PacletInstall can work with.
There only options available are HTML-based methods, but PacletInstall only responds to true server redirects.
06:08
Here's my current thinking: 1) there's no way to force PacletInstall to redirect on GitHub (it's certainly possible on Wolfram Cloud, but literally no one will ever go to a Wolfram Cloud address I figure). 2) PacletInstall called on the repository paclet asset download URLs do trigger a download logging event
What I imagine I can do is provide the main paclet server on GitHub and mirror the PacletSite.mz onto a ghost paclet server at www.wolframcloud.com/paclets/PacletServer
This ghost server would just provide redirects to GitHub content, either the proper paclet site, if it's not a release URL, or the release asset URL, if that exists.
PacletInstall requests would go to the ghost server. The site and all main content would remain on GitHub (this would also free us from having to duplicate paclets that are already release assets, too).
I am also thinking I'll provide a mechanism to register a webhook from the paclet server to your repo that will automatically open a rebuild request or do a notification in some way like that so that people can just simply add a new release and the build can be automatic-ish.
Now, there are at least 50 pitfalls here, notably a) this uses the Wolfram Cloud which is basically still a beta product b) I don't have a good way to remove myself from this build process at this stage, which is what's necessary to get proper CI
But it's a start and wouldn't be terribly hard to do. @GalAster @Szabolcs what do you think?
06:47
Omg the cloud is so bad. If I want any short ID I can register the account to [email protected] (even if I don't own that account) then I can "update email" to some email I do own.
And now I have that ID. I just did this because I remembered someone already owned [email protected] so I didn't have user:paclets as a short name.
So I just registered with that email and changed mine.
07:19
:)
07:59
@b3m2a1 What kind of redirect? Do u mean set the "Location" to a url in PacletSite.mz, or the web site do the redirect
Are you fed up with MenuSetup.tr / KeyEvenTranslations.tr customization? Then NotebookActions feature from DevTools` v0.10.0+ (https://github.com/kubaPod/DevTools) is for you! :-p

And seriously, I think it can be quite useful even if in an initial stage. Here's an example: https://mathematica.stackexchange.com/a/184568/5478
U can pack ghost Paclet Information in PacletSite.mz in fact, do not have to use automatically generated one.
`"*.paclet"` is a zip, generated by `com.wolfram.paclet.PacletPacker`, but
`"PacletSite.mz"` is a hard-coded compressed package. I don't know how to do it. I often generate corrupted compressed packages.
08:38
@GalAster I want a proper redirect, so that PacletInstall will follow it
Set "Location" also follow.
Interesting. I didn't know that. That's a pretty easy change, then.
Well I'll still need to maintain some list of redirects to apply to the server list, but that's not too hard.
I think you can also use a PacletSite.m instead of a PacletSite.mz
Which is easily writable
Ah no looks like it enforces .mz
Shame
There's a ...<>"/PacletSite.mz" in the source
I will work on CI, vuepress also needs NodeJS, these thing can be done together
In terms of the packer, I'm pretty sure it's just doing a .zip compression
fine, thank you
08:50
What's your vision for the CI?
 
1 hour later…
09:53
:47301373, I've just had an email from Wolfram to ask if I want to make my BVP eigenvalue package available on the FunctionRepository (resources.wolframcloud.com/FunctionRepository)
Although it looks like it isn't really helpful, given that my package has multiple functions that need to interact
Hmm. It looks like my mac version happily overwrites csv files that are already open in Excel (although you don't notice until you close the file and reopen it), but Windows throws an error.
10:09
@KraZug The function repository is...borderline useful I think. Actually one potential person who could really contribute is @HenrikSchumacher as he has a lot of powerful, standalone functions that only really exist here.
But most useful things aren't one function...
10:24
@b3m2a1, yes, I see that. Wolfram sent out an email plugging my package as an advertising email, but attributed it to my former PhD supervisor instead of me. They have now sent out an email correction to the same list, and then they have asked me: "Would you consider adding your functionality to our Function Repository to make it available to a broader audience?"
"We'd like to put you in touch with a co-worker of ours who'd be happy to assist and guide you through the process of adding your functions. ", so I'll see what they come back with.
Ask them when the paclet repo is coming out :)
3
10:36
I want to export a string to a subfile in a ZIP file, but have it be interpreted as a package when imported, does anyone know how to do this?
Just for example we can try ExportString[<|"asdasd.txt"->"a[1]"|>, "ZIP"] but I want that .txt file to really be a .m file, but the contents to still be a[1].
Oh nice, nevermind, this is how you do it:
bleh = ExportString[<|{"asdasd.m", "Text"} -> "a[1]"|>, "ZIP"];

ImportString[bleh, {"ZIP", "asdasd.m"}]

a[1]
10:51
Yeah, I may well ask that. They have a "Publication Watch Team" - at least that is who I've been communicating with. But they don't seem to blog about it in any way, which would be the best way to actually highlight packages.
11:11
Hmm, is there a way to make a safeSave, similar to the safeExport function given here: mathematica.stackexchange.com/questions/75745/…
11:25
@KraZug replacing Export with Save does not work because...
?
it evaluates the argument
@KraZug HoldAll / HoldRest etc?
yes, almost certainly a Hold problem
I mean, are you aware of those attributes?
Set them for your export function to pass arguments to Save without evaluation.
I'm vaguely aware of them yes, but not in practice
list = RandomReal[1, 50];
safeSave[file_String, args___] :=
If[! FileExistsQ[file], Save[file, args], $Failed]
safeSave["test.csv", list]
ok, so SetAttributes[safeSave, HoldRest]
thanks
12:28
Why do they even provide that shitty Stack Trace feature...it's almost never properly useful and it has a bad tendency to crash the FE.
12:49
@GalAster bad news: PacletFindRemote calls into PacletManager`Package`setLocation which manually reassigns the location specified in the PacletSite.mz file to wherever they downloaded from.
13:22
Okay I have a workaround, but rather than purely using PacletInstall it will use a custom function that does a slight override of setLocation. I just tested that it works, so we're good on that.
13:51
More of a TeX question than M; however, I've got a script with a line Export["out.tex",RandomChoice[list of long text strings],"TexFragment"] that needs to be line-wrapped in order to render properly in LaTeX. Does anyone have any suggestions either in M or LaTeX on how I can get the string to wrap if needed?
I would say the Function repository is actually quite useful especially in a stack exchange context. There are so many nice functions that only live in answers here and it would be nice to be able to store them somewhere (and update them if it's useful)
ResourceFunction["ReadableForm"], I just had a hard time to creat a function like this, LOL.
14:21
@Szabolcs we can now use the PublicPacletServer interface to install directly from the download URL in your release, provided we're willing to use a function I wrote to overcome some of the hardcoded difficulties in the PacletManager (I've already emailed Todd Gayley about them)
<< paclets.github.io/PacletServer/Install.wl
PublicPacletInstall["IGraphM"]
That will install "IGraphM" from your link, but it uses the PacletServer architecture and everything PacletInstall does for that.
@KraZug Would you mind sending me the emails of Wolfram Review Team? I have a lot of neural network models want to submit, but I filled out that table several times and no one contacted me.
My only changes were to take two of the implementation functions and fix a few bugs
For performance (and also a little bit safety) I'm split the repo into two different repos, too. There's a "Repository" repo which is basically a bare paclet server. We can do all the normal PacletInstall stuff off it. Then there's the PacletServer repo which provides a customized PacletSite.mz file and an install function that respects the "Location" parameter supplied in it.
My email is [email protected], help needed
14:35
@CarlLange We should have a Github repository for this, I think there was some talk of having that once but it never happened
One big problem is that WRI might at any time add a function with the same name as those in the function repository...
14:57
@GalAster, my contacts have been through [email protected]. They sent out a blanket email to advertise my package (very poorly in my opinion, could have explained my package much better if they'd just asked me beforehand). I replied to an email this morning where they suggested putting it on the repository, haven't had a direct contact since.
@GalAster, generally the moderators put email addresses in something like Compress, so it can't be read directly
@KraZug What is [email protected] ?
@b3m2a1 Thanks! I will test this.
15:11
@C.E. Yeah let's do it!
@b3m2a1 Version 10.0 fun:
Strangely, the Get (instead of Import) fails silently. It just does nothing. No package is actually loaded.
Also, this thing broke my auto-install script too: URLFetch from GitHub releases breaks with the same error.
Is this because GitHub changed certificates?
https from raw.githubusercontent.com does load fine. I was actually trying to make an easier install script before I saw your message.
If I start using this, I would like it to be very very robust though.
Can you get rid of StringEndsQ?
I suggest something like:
If[$VersionNumber > 10.0,
 stringEndsQ = StringEndsQ,
 stringEndsQ[str_, patt_] := StringMatchQ[str, ___ ~~ patt]
]
URLSaveAsynchronous has a nasty habit of failing in some versions on some systems. As I remember, the issues were around 11.0 (+- some), primarily on Windows.
This is why I use URLFetch in my script.
setNonRemoteLocation[paclets:_Paclet,location_String] --> should always use full context, PacletManager`Paclet, or Needs the PacletManager where needed
Sorry, I use URLSave to be precise
I should not that I just skimmed the code, so I only noticed local issues. I should take the time to understand how it works, and which bit runs when. Maybe some of the parts I mentioned above do not actually run.
15:39
@Szabolcs, they sent a blanket marketing email (Dear Researcher...) about a week ago from that email address, publicising my package.
15:52
@b3m2a1 Maybe supporting such old releases isn't reasonable after all. It was just pointed out to me that Wolfram stopped supporting it a while ago.
16:07
@b3m2a1 How about I put this method as an alternative "experimental" install method and ask people to report issues?
 
1 hour later…
17:33
@KraZug How does one get Wolfram to publicize one's package?
17:59
@Szabolcs, no idea. I only found out about it when they sent me the marketing email. I'd try emailing them on that address
 
2 hours later…
19:35
@b3m2a1 Could you please correct the path to the MaTeX paclet in PacletSite.m?
20:04
@Szabolcs Ah yeah, that's fixed now.
If you're interested I register these data points in a file ReviewQueue/ExtraPacletInfo.wl which can be used to register data for the PacletSite.mz file that doesn't show up in the raw PacletInfo.m.
That way you can add these non-standard "Extensions" like {"PacletServer", ...} that I use for building the pages without worrying about adding it to your actua paclet
It's also where I do custom overloads of "Location".
20:41
It seems SE have already made up their mind that no matter what the community says, they are going to force new themes on us, without listening to most of the criticism. Bad attitude ...
-26
Q: English Language & Usage's updated site theme is ready for testing!

CatijaAs part of implementing the new unified themes across the network, we're gradually rolling out updated site themes for each site. As of today, we have enabled your updated site theme for testing. If you can't see it right now, that's by design! We're hoping to get feedback from you before rolli...

-26 votes
53
Q: Make TeX.SX look nice again!

samcarterA collection of user scripts to make the main site look a bit better again, after the new design has gone live on August 3, 2018. The scripts can be used e.g. with Tampermonkey (Chrome) Greasemonkey (Firefox) A fix for recent versions of Firefox and GreaseMonkey (4.x) by @PaulGaborit. A use...

-136
Q: Mathematics new site theme is live

Joe FriendWe rolled out the new site theme for Mathematics. It is live now. What new theme? If you're like, "What the heck are you talking about?", then you should read the Meta Stack Exchange post entitled Rollout of new network site themes (and maybe the posts it links to for the full background). Y...

-136
On our site, there wasn't even a comment about the single most upvoted complaint, which should in principle be easy to fix.
Actually, some of the themes look quite fine to me (not counting the left nav bar and all that). Our does not look good.
@Szabolcs This is what really bothers me. I'm fine with the march of time and I understand their wish to redesign the site, but I feel like our site-specific feedback, which they specifically asked for has been so far totally ignored.
ah, I take it back, Jon just yesterday responded to a few of the bugs: mathematica.meta.stackexchange.com/questions/2374/…
although the syntax highlighting and the logo, seemingly the biggest issues to this community, have no response on them.
also, yuck, 'centred by design' mathematica.meta.stackexchange.com/a/2380/57593
21:09
I miss the old stars. The new ones are just a plain 5-star. The old one was the combination of two 5-stars which evoked the stellated dodecahedron which is the classic Mathematica symbol.
I believe these are among the non-negotiable standardized elements.
And so are the tags.
But are these really such a big maintenance burden as they claim?
I mean, just look at how simple the TeX.SE hack was to restore {tags}: tex.meta.stackexchange.com/a/7791/6302
Have they still not fixed the font coloring?
Not a single comment on it, even after I asked explicitly.
I worry that they want to standardize that too.
Actually, TeX.SE has one that's different from SO. I'm relieved. It looks like this is not among the standardized ones.
Also @Szabolcs the star image is an easy CSS fix too: .star-off{ background-image: url(old/image/path) } and presumable one for .star-on too
Anyone have the old layout who can tell me where those old icons were hosted?
Use a private browsing window.
I got it. If you use this:
.star-off {
  background-image: url(cdn.sstatic.net/Sites/mathematica/img/…)
}
Both will come back as both sprites are just at different spots on that image
(both the on and off appearances)
21:19
When the new theme is made the default, I will use these tweaks. But until then I want to give it a chance, so I can actually comment on the biggest issues and hopefully influence the design.
Sounds good. The upvote and downvote arrows have changed size.
So when/if this change becomes permanent I'll also need to tweak that
I'm annoyed by the changes, but I can't help but feel that the vitriol on some of these posts is...unproductive.
Like some people just comment "YOU GUYS RUINED MY LIFE $%##$@#$%$", which while emotionally gratifying doesn't really help.

« first day (2506 days earlier)      last day (2279 days later) »