Extend Base.gradient instead of defining a new one on 0.4

[yuyichao]

Not sure if this is the best long term solution but at least it makes all tests pass and IMHO produce the least breakage for users.

[mlubin]

gradient is on 0.3 also, no need to special case for 0.4.

[yuyichao]

I C. So it's the recently introduced handling of conflict export that breaks this, rather than the conflicting names themselves......

[mlubin]

Appears so

[yuyichao]

Updated

[mlubin]

I don't think this was the right solution. There's more than one method to compute the gradient of a function. I don't think the Calculus version should be any special and take over the gradient(::Function) method.

[johnmyleswhite]

I agree.

@mlubin: What do you think of trying to pull this package apart into separate packages for finite differentiation and symbolic differentiation? Now that ForwardDiff is out, I feel like I could take the work I did long ago on Calculus2 and try to adopt your API. We could then start to deprecate this package.

[mlubin]

Splitting into FiniteDiff and SymbolicDiff seems like a good idea. Both
could be hosted under JuliaDiff. Are the integration features in Calculus
useful?
@jrevels
On Sep 9, 2015 9:43 AM, "John Myles White" notifications@github.com wrote:

I agree.

@mlubin https://github.com/mlubin: What do you think of trying to pull
this package apart into separate packages for finite differentiation and
symbolic differentiation? Now that ForwardDiff is out, I feel like I could
take the work I did long ago on Calculus2 and try to adopt your API. We
could then start to deprecate this package.

—
Reply to this email directly or view it on GitHub
#66 (comment)
.

[johnmyleswhite]

No, the integration functionality is far inferior to what's in Base.

[jrevels]

What do you think of trying to pull this package apart into separate packages for finite differentiation and symbolic differentiation?

Splitting into FiniteDiff and SymbolicDiff seems like a good idea.

I like this idea.

Now that ForwardDiff is out, I feel like I could take the work I did long ago on Calculus2 and try to adopt your API.

That would be great. I stumbled upon Calculus2 a while ago and was hoping the effort would be revived soon! Let me know how I can help, and feel free to file issues in ForwardDiff.jl if anything needs to change API-wise there to make things work out.