@b3m2a1 Exporting a Circle to svg leads to crazy values for the radius and the position

@halirutan huh yeah and that seems not to bedevil Disk... ah I see I did my coordinate adjustment twice instead of once as it should have been

@halirutan just pushed the fix

@b3m2a1 Thanks!

I like how much more responsive Image3D is in 12.0

@Szabolcs @b3m2a1 IIRC, both of you also have a strong opinion about WRI using an open bug tracker. As you might remember, Around pointed out a few days ago that Stephen might want to discuss this. I have prepared a first version of a blog post where I discuss this as open as possible.

Would you, or anyone else, want to read over it or even make substantial edits and act as a co-author?

@halirutan I can bring my fifth-grade English reading/writing skills to bear on this if you want a grammar checker

Also Around might prefer being called Arnoud :)

@b3m2a1 I'm more afraid that I have forgotten some obvious arguments. Szabolcs always has a very clear way of writing and pointing out the important things.

Furthermore, if the post includes opinions from several people, it gives a better overview what the community thinks.

@b3m2a1 Sometimes I also write Arnould for whatever reason :)

Ah my thought process is basically this but I can look it over and see if I have anything to contribute

@b3m2a1 Alright :)

@halirutan Absolutely. I didn't have time to respond here when Arnoud mentioned it and you pinged me.

How to convert an Image to a Raster?

It seems Show @ Image does this, but is there a direct way?

@Kuba You mean Raster and not Graphics@Raster[]?

@Szabolcs I want to have something like Graphics[{ Dynamic @ image, Dynamic @ line}] but unless I convert image to raster Graphics complains it can't handle images.

@Kuba I think Raster@ImageData[image, DataReversed->True]

Will it keep relevant metadata?

color scheme etc?

@Szabolcs I am doing First @Show@RandomImage[1, 5] which is good enough

What metadata is relevant for display?

but ugly

ColorFunction?

Does Image store a ColorFunction? I thought only Raster does. I thought Image hard codes the result.

Image doesn't even have such an option. It only has ColorSpace.

Yes, and it is converted to color function, or am I wrong?

Yes, you are right. You would lost information this way. Probably Show is the way to go. Is it too slow?

Ideally, Raster should be able to contain an unmodified Image

@Szabolcs I think not but it feels strange so I thought I am missing something.

How about Inset @Kuba?

No conversion done, no overhead.

@Szabolcs not sure if it would be easy to match coordinates between the image and what I about to overlay.

What's wrong with Inset? I guess one thing is that you don't get auto-conversion from pixel coordinates to graphics coordinates

Yes

It would not be trivial, but it is definitely possible. It's not like Overlay where it's not really possible

@Szabolcs yep, but First @ Show wins so far, I will go with it

Graphics[{
  {Dynamic[First @Show@RandomImage[1, 5], UpdateInterval -> .1]},
  {Red, Thick, Arrow[{{.5, .5}, {4.5, 4.5}}]}
  }, ImageSize -> 400, PlotRangePadding -> None, Frame -> True]

@Kuba See above. That's a reason not to be confident about Inset

Still unfixed in 12

I use Inset when I only care about position not sizes.

@Kuba It does appear to align very well. Examine the pixels in the screenshot.

Maybe you should give Insert another chance :-)

@Szabolcs right. So you would go with Inset instead of Show? It is probably faster

My hunch is that it is faster. I haven't tried. It won't process the image, it just inserts the boxes. That's as fast as you can get

@Szabolcs I will give it a try, thanks.

Sorry, I don't remember the exact issues I had with Inset, but you may want to read the post I linked ... just in case it applies ...

ok

will read it once it fails :P

I have an idea but I am not 100% sure this is a good idea and I wanted to get some feedback first. I was thinking about asking the SE community what FEM features should be added next. This question is, due to my work with WRI, a bit different than other questions here, I understand that. I have a draft text that I can post if people here are not completely opposed to the idea.

@Kuba Would it be off-topic on meta?

@halirutan with all my love for that idea, it would, it is not about the site, is it? On meta otoh fewer would complain about that :)

@Kuba The thing is that I would rather have it here, where the right people see it and we have a clear approach of voting and commenting.

@halirutan I know, so do I. That is independent of it being off topic :) But again, we have many off topic an highly upvoted threads so at the end meta can be a way to go.

@user21 The above being said, I would very much like to see this on meta although it's not on-topic.

To cite stackexchange policy:

> Meta is for Mathematica Stack Exchange users to communicate with each other about Mathematica Stack Exchange (asking questions about how the websites work, or about policies and community decisions)

I hope we can stretch the interpretation enough to acknowledge that this would be a community discussion, although not specifically about this site. The situation that we as a community get asked about possible features is unique and exciting enough for me.

@Kuba, never heard of of that blog before.

@halirutan, I am not sure I understand, do you propose that I ask my actual question on meta or that the discussion about such a question should be on meta?

@user21 judging by the content it seems to be driven by Wolfram Cloud team so you can ask JP.

@user21 ask the final question on Meta.

@b3m2a1 any idea why it does not work?

DynamicModule[{right},

 right = Function[, Print@RandomReal[]];

 EventHandler[
  Graphics@Disk[],
  {
   "MouseClicked" :> right[]}
  ]
 ]

@halirutan Here you go: Resulting SVG (now 18kb), Notebook

@user21 this is somehow linked with the thoughts being shared on an eventual public list of bugs, isn't it?... public formalized discussion on functionality...

@P.Fonseca, no, it's purely my own initiative to make a better product.

@Kuba it looks like the FE mechanism for figuring out where Function should evaluate isn't used to Null as the first argument. Any of these work:

DynamicModule[{right, wrap}, right = Function[Print@RandomReal[]];
 EventHandler[Graphics@Disk[], {"MouseClicked" :> right[]}]
 ]

DynamicModule[{right, wrap}, right[] := Print@RandomReal[];
 EventHandler[Graphics@Disk[], {"MouseClicked" :> right[]}]
 ]

DynamicModule[{right, wrap}, right = Function[Null, Print@RandomReal[]];
 EventHandler[Graphics@Disk[], {"MouseClicked" :> wrap@right[]}]
 ]

It looks as if it tries to evaluate it in the FE but fails due to not understanding how Function works?

Ah yeah definitely an issue in how the FE implemented Function as this works too:

DynamicModule[{right, wrap}, right = Function[Null, Print@RandomReal[]];
 EventHandler[Graphics@Disk[], {"MouseClicked" :> right[1]}]
 ]

This works too:

DynamicModule[{right, wrap}, right = Function[{}, Print@RandomReal[]];
 EventHandler[Graphics@Disk[], {"MouseClicked" :> right[]}]
 ]

I've not seen this before:

I cannot reliably reproduce it (yet).

@MichaelE2 ah yeah that's the deepest fallback when Dynamic fails

You can make it happen by providing broken DynamicBox constructs often.

The cell expression is:

Cell[BoxData[
 DynamicBox[ToBoxes[
   Out[-2], StandardForm],
  ImageSizeCache->{305., {16., 24.}}]], "Output",
 GeneratedCell->False,
 CellAutoOverwrite->False]

Out[-2] should be the result of the input 2, i.e., 2. It sometimes works with similar input/output.

Once instead of the ToBoxes[Out[-2],..] in the cell expression, the box form of the value Out[-2], which was the symbol x, appeared.

Inconsistent behavior seems like a bug, at least to an ordinary user.

@b3m2a1 Mathematica wrote the DynamicBox expression, so I can't take credit.

BTW, I'm going to try to use your package/paclet dev stuff to create a diff. eq. utility package for a course I'm teaching in the fall. Thanks for putting that together for us. It's looks cool.

@MichaelE2 glad it will be put to use! Let me know if/where/when you get stuck or things are unclear or aren't as smooth as they should be.

I'm always down to write tutorials and polish things up if they will be useful

@b3m2a1 Thanks! I'm hoping stylesheets will work seamlessly. Years ago, I didn't really understand package structure, except for the .m file. I'd distribute a stylesheet separately to the students; but they wouldn't follow directions, wouldn't put it in the right place, would just ignore the errors and press on without letting me know there was a problem. I just stopped using them.

@user21 I propose to ask your final question on meta.

It's such a pity that the new RelationalDatabase stuff doesn't support postgis geometry types.

HSTORE is one thing (although that should be easily represented just with Associations), but the geometry would be so useful

@MichaelE2 if you use a paclet it should all work well

If you want an easy place to point your students to to install the paclet you can submit it here and the installation should be no issue

Extra pity that it seems like I can't define EntityFunctions with arbitrary SQL on EntityStores backed by RelationalDatabase :(

@CarlLange the biggest pity is that the core implementation code isn't well documented or exposed

Otherwise you might be able to write that in rather than wait for 12.1

@b3m2a1 Yes, I agree. Not for the first time I've wished for the thing to be open source so I can just fix my problem instead of complain about it

@CarlLange well this part you might have the source for

Like you can dig this up probably

The issue is that the low-level stuff never gets documented or exposed or any real love from WRI

@b3m2a1 Yeah, sure, but I'm working against the system :(

Not-strictly-related complaint - it feels a lot like they've moved away from writing Tutorials and towards just having loads of guide pages that are essentially just different-ordered lists of functions

@chuy 2 mins

Seems like it should be fine, I'm just not sure about the way to run the query against an entity class

I think the way the SPARQL stuff works for Entities was slightly tacked-on

(I do love seeing SPARQL in use though)

I think there is an interesting set of opportunities if one could use SPARQL to access Entities (especially in conjunction with other knowlegde graphs).

@chuy There certainly is. You can already do this to some extent. They give an example somewhere of linking the Knowledgebase with Wikidata using a "linking set"

It's not clear to me that you can use the SPARQL stuff with entity classes or lists of entities in this way, unfortunately. Your query looks fine though.

Is there a straightforward(ish) way to push a dataset (or group of associations) to an RDF Dataset (maybe this question doesn't even make sense)

Ah, the linking set stuff is in the docs for SPARQLLoad

@chuy Not that I'm aware of, but let me think about it for a bit. We're talking about eg ExampleData[{"Dataset", "Planets"}], right?

yeah something like that, could even start with something simpler (tabular data)

@chuy On a simple level this is pretty easy, you create an RDFStore with triples for each part of your dataset. Then you can export it as ttl or whatever

ex = "http://example.com/" <> ToString@# &;
RDFStore@Flatten@
  Normal[dataset[All, <|#, "id" -> CreateUUID[]|> &][All,
    With[{id = #id},
      KeyValueMap[RDFTriple[ex@id, ex@#1, ex@#2] &, #]] &]]

with dataset = ExampleData[{"Dataset", "Titanic"}][[;; 5]]

@CarlLange thanks an example here is very useful for me

Ah, made small mistake, ex@#2 should just be #2

This is just a toy example, there's a lot more you could do, but it should get you started :)

indeed

and then you can do stuff like

SPARQLSelect[
  {
   RDFTriple[SPARQLVariable["id"], ex@"class", "1st"],
   RDFTriple[SPARQLVariable["id"], ex@"sex", "female"],
   RDFTriple[SPARQLVariable["id"], ex@"survived", False]
   }
  ][store]

Showing the five women in first class who didn't survive :)

I am envisioning combining URLDispatcherwith this

so the URLs are real

A lot (a lot) of the time in these cases, the URLs don't actually resolve, or resolve to a landing page

But that's definitely a good idea

I have stuff like this :wolframcloud.com/objects/jhernandez/kb/Country/Spain/Population

Nice :)

in principle, one could have any entity listed there, you can swap Spain with Portugal

or any Entity + Property

Yeah!

it's my naive semantic web

Now you're playing with power ;)

@b3m2a1 @Kuba @C.E. @CarlLange @Szabolcs Can we live with that? I made the spacing between the & a bit larger since I agreed with b3m that this looks a bit clearer

@b3m2a1 Thanks again. Distributing the paclet was something I would eventually have to worry about.

@halirutan Looks good to me!

@CarlLange Thanks again for the code (which doesn't run in 12 btw).

@halirutan Which code? Might not have been me you're thinking of :)

@CarlLange The code that Lucas gave me :)))

;)

I'm confused as always.

Huh, quite surprised that there's no built-in venn diagram function

@halirutan There's very little left of the right-alignment now, I wonder if one might not just as well justify the text so that it takes up the whole line. I'd prefer the space to be equally divided between the ampersand and Wolfram and Wolfram and Language.

That will still leave more space after the ampersand, while not disturbing the rhytm.

@C.E. For me this is no solution. The space between two words is important for a good look&fell imo. It looks ripped apart if we distribute the space

@halirutan it doesn't look acceptable to me either, but thanks for trying it out.

@halirutan I can live with that. But I don't trust mysense of aesthetics. Should we ask SE team for feedback?

@Kuba @C.E. What I do is that I finalize both the normal and the meta logo. Then I create a new meta post where people can again comment. When we find a compromise that works for everyone, we simply turn this post into a feature request.

And then I can probably talk to our designer if he has further comments.

@halirutan Sounds good