This isn't the thread for this, but I'll bite.
That's exactly the reason why it's a good idea. The backends aren't
swappable, but the code is. And for the most part that means you can just
avoid the cons of any backend instead of having to fight against them. You
could be making all of your plots with the PGFPlots backend for some
publication, and then realize that you need a trisurf plot. You can just
switch the backend and re-save your plots without actually writing new
code, and now they can be all saved and matching in PyPlot.
Another example is when you run into issues with precision. PyPlot is a
good standard choice, so I was using it to show all of my convergence test
results. However, once I started testing things like Order 10 Runge-Kutta
methods I noticed PyPlot simply can't plot values that are below 1e-30. The
quick fix is to just change the backend, i.e. add the one line of code
`plotly()`, and now all of the convergence tests how nice plots that work
in their range.
This is not to mention that Plots adds features to each backend. Since it's
not focused on making the graphic backend, it's just focused on making
"recipes" which are convenient commands for doing things like making a
scatter plot matrix, and will work for any backend. So sure you can't use
every feature of every backend, but there are more features you can easily
use through Plots than just using the backend itself. That said, it's still
a young package so there are places where it needs to wrap more of the
backend, like with trisurfs you need to be able to specify triangles. But
Tom is working on that.
On Monday, August 29, 2016 at 6:54:44 AM UTC-7, Daniel Carrera wrote:
>
> Plots.jl is a good idea, but the backends are not really swappable. You
> can get a fairly different plot if you swap the backend.
>
>
> On Saturday, 27 August 2016 02:19:45 UTC+2, Chris Rackauckas wrote:
>>
>> You should really check out Plots.jl. It's a plotting metapackage which
>> lets you use the same plot commands to use any backend. It's nice because
>> if you're using it an one package stops getting updated, you can switch to
>> another plotting backend without changing your plot commands.
>>
>> But I can see that, although Winston hasn't been tagged in almost a year,
>> there has been some development work. Have you tried
>> Pkg.checkout("Winston")?
>>
>> On Friday, August 26, 2016 at 3:04:22 PM UTC-7, K leo wrote:
>>>
>>> so that it works with version 0.5.
>>
>>
