How to get native Wolfram plots

[gganapat]

I miss the ability to get high quality native Wolfram plots like before (e.g, Export["plt.png",Plot[Sin[x],{x,0,Pi}]]; Import["plt.png"]). This could sidestep many of the missing features of the WLJS Plot function. I was wondering if I could run this through a command line accessible through WLJS and then import the plot. That would be awesome addition because there are just too many nooks and crannies to fix all issues otherwise.

Thx!
Gani -

[JerryI]

Yes. There is a terminal in the command palette, where you can select Master Kernel. It is basically a frontend server, which is not affected by WLJS Packages. However, it does not have an easy access to your notebook data (unless you export the data you want to plot and then import it using master kernel to rasterize it...)

There is a better way to do it.
You can use parallel kernels (are not affected by WLJS packages)

firstly launch at least 1

LaunchKernels[1]

then you ParallelSubmitAsync (see docs)

With[{
  p = Plot[x, {x,0,1}],
  dir = FileNameJoin[{$TemporaryDirectory, "test.png"}]
},
  ParallelSubmitAsync[Export[dir, p]] // WaitAll
];

Import[%]

boom ;)

I should have been told you early about it. This idea just came to my mind, I have forgotten that they are absolutely "clean".

[gganapat]

WIll this work only in notebook? I am using the server mode and it doesn't seem to like it (plot with red outline). I something like this which happens if there is more than one kernel running?

gani -

[gganapat]

It seems to do the same with the Notebook unfortunately.

gani-

[JerryI]

Hm. Could you, please, try to move Plot inside ParallelEvaluate as well?

UPD: del, missed the latest message

[gganapat]

You are truly a genius! Thanks!Gani-Sent from my iPhoneOn Jan 25, 2025, at 12:41 PM, Kirill Vasin ***@***.***> wrote: Yes. There is a terminal in the command palette, where you can select Master Kernel. It is basically a frontend server, which is not affected by WLJS Packages. However, it does not have an easy access to your notebook data (unless you export the data you want to plot and then import it using master kernel to rasterize it...) There is a better way to do it. You can use parallel kernels (are not affected by WLJS packages) firstly launch at least 1 LaunchKernels[1] then you ParallelSubmitAsync (see docs) With[{ p = Plot[x, {x,0,1}], dir = FileNameJoin[{$TemporaryDirectory, "test.png"}] }, ParallelSubmitAsync[Export[dir, p]] // WaitAll ]; Import[%] boom ;) Screenshot.2025-01-25.at.21.35.33.png (view on web) I should have been told you early about it. This idea just came to my mind, I have forgotten that they are absolutely "clean". —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[gganapat]

Never mind. With the latest version, it works fine. Thx --
gani -

[JerryI]

I see. We added this helper for parallel computing recently

[gganapat]

I tried to make the above into a module for repeated plots which works but I get an error message like so...

Any suggestions on how to fix it?
Thx --

[gganapat]

Never mind. I had errors in the module. Fixed...

[JerryI]

You can also just apply Rasterize instead of Export. Then WaitAll will return already rasterized image to the main Kernel (no files management involved).

[gganapat]

Writing wrapper and plotting works beautifully but the file size when saving is enormous. For a single plot it is >8MB and it goes up from there. With multiple plots, the system takes forever syncing data. Not sure what's happening. The actual plot size is only ~ 146 kB. Is there any way to reduce the file size?

thx, gani-

[JerryI]

A bulky filesize is an effect of the combination of two things (writing just for other contributors as well)

• internally Wolfram Language stores Image as uncompressed byte array (already bigger than a typical PNG file with length-encoding)
• WL uses WXF format for all data (efficient binary format), however, on Javascript side we only have WXF -> JS translator (even not fully implemented), but JS -> WXF is completely absent. Therefore we have to use textual ExpressionJSON format, which increases the size of any binary object by 5-10 times.

In the future we will definitely move to a fully binary data format, but when, it is hard to say (could be somewhere in 2025).

Anyway, I have a solution for your case using good-old HTML ;)

embedImage[i_Image] := With[{uid = CreateUUID[]}, With[{
  path = FileNameJoin[{NotebookDirectory[], "images", uid<>".png"}]
},
  Export[path, i];  
  HTMLView[StringTemplate["<img src=\"/images/``.png\"/>"][uid]]
] ]

or in order not to pollute that folder with many images (if you reevaluate many times the same cell), one can use cell-id as a filename

embedImage[i_Image] := With[{uid = EvaluationCell[]//First}, With[{
  path = FileNameJoin[{NotebookDirectory[], "images", uid<>".png"}]
},
  Export[path, i];  
  HTMLView[StringTemplate["<img src=\"/images/``.png?nocache=``\"/>"][uid, RandomInteger[999999999]]]
] ]

this odd thing here nocache is to kick out the browser's cache (when you overwrite the same file).

Then if you want to share this notebook with someone, you can compress and embed all project files (in the folder of the notebook) into the notebook using Embed files method and then save it. If someone opens (or you, if you restart an app and open this notebook again) it it will self extract all files and reduce its size back to normal (1 way process).

Hope it helps as a temporal solution.

Best,
Kirill

[gganapat]

thank you! Will try it out.
gani -

[JerryI]

#344 planned