Custom printing for Taylor1

[lbenet]

Fixes #211

... by adding a new global variable.

[dpsanders]

So you can't do a Taylor1 of a Taylor1 with different variable names? It seems to me like we could just store the variable name inside the Taylor1 object.

[coveralls]

Coverage increased (+0.03%) to 94.336% when pulling 3039fb0 on lb/iss211 into da84c93 on master.

[lbenet]

Thanks for the remarks!

Indeed, you can't have different symbols which still makes those objects pretty confusing. There is some work on nested Taylor1s (see #126) though it has some problems; I have some (old) implementation which I haven't pushed, which would allow to distinguish those variables nicely. Perhaps I should also push that for evaluation.

I also thought about adding a fieldname in the struct as you propose. The question is what would we do when we have an operation (say addition) involving two Taylor1 objects with different display symbols? Independently of what the answer to that question will be, we would need to check the field of the displayed symbols, which would impact negatively the performance. (Perhaps, the solution is to promote the result to a Taylor1{Taylor1{T}} since we are actually talking about two distinct independent Taylor1 variables, but this is also debatable.)

So I simply decided not to get into those issues.

[lbenet]

@dpsanders Do you agree to go ahead and merge this?

[lbenet]

I'll go ahead and merge this.

[dpsanders]

Sorry for the delay, thanks for merging!