Tim Holy

If you just want the raw image itself (no axes), you can use ImageView's `write_to_png`.

You have to use ImageView to plot the image, too. See https://github.com/timholy/ImageView.jl#programmatic-usage ``` julia imgc, _ = view(img) write_to_png(imgc, filename) ```

Currently there isn't a function like `imagesc`, but you can write it in a few lines using `ImageCmap`, something along the lines of ``` julia mn,mx = extrema(img) lookup =...

Not had time for my mail in several days (it's pure luck this was on top of the heap), and I'm leaving town momentarily for a couple days. A couple...

`n=3` probably doesn't critically depend on the QuasiNewton, "explore" (rather than "exploit") will eventually cover the search space. But you'll see some benefit even there in terms of how quickly...

One potential solution is to mimic [how Observables handles this](https://github.com/JuliaGizmos/Observables.jl/blob/0f94e69cabddd87fcf057fb95ed5dc64be9661de/src/Observables.jl#L142-L155), which at least lets you precompile the function you will be wrapping. That said, you can precompile only those things...

You may have to `@nospecialize` the function you're wrapping *and* all of its callees (or at least those without predictable or precompiled types). For example, if the callee will only...

As stated in https://github.com/SciML/DiffEqBase.jl/pull/736#issuecomment-1080011363, if the argtype is deterministic then there shouldn't be any cost to specialization, and indeed you want it for the run-time performance benefits. But you do...

I can't replicate this because the URL seems interpreted as a filename on Linux.

This is unlikely to be fixed on Julia 1.0. It may work on Julia 1.3 or higher. But in general, anyone running Julia 1.3 or higher should install ImageIO and...