Adding support to DiracDelta for generic points in the domain

[kishore-nori]

Hi @santiagobadia and @amartinhuertas, I added methods to DiracDelta to support generic points in the domain, which are not grid entities with tags, would be glad to know your suggestions for improvement and changes. This PR is aimed at resolving the issue [#357]. Given that CellFields can now be evaluated at a generic point in the domain, the method make use of this to calculate the integral of the composition of DiracDelta with a CellField, and also works for generic functions compatible to evaluate at a Gridap Point.

So as to conform with DiracDelta struct, a DiscreteModel is constructed with given set of Points to generate Triangulation and Measure. If we are not required to conform with the existing framework, we could just store the Points as a struct and evaluate the CellField point-wise directly, but the above doesn't create much overhead. And probably the later is not generic enough to be extendable to line and surface forcing terms.

The usage is same as earlier but the constructor is as follows:

model = CartesianDiscreteModel((0,1,0,1),(3,3))
p = Point(0.1,0.2)
pvec = [p, 1.5*p]
δ = DiracDelta(model,p)
δ_comb = DiracDelta(model,pvec)

[WIP] TO DO: support for line and surface loads! (like presented here #408 (comment)). I think the line load case is extendable in a straight forward manner using CartesianDiscreteModel (but may be needs to be done separately for each line - defined by a pair of points..)

[amartinhuertas]

HI @kishore-nori ! Thanks for your work. Can you please add an entry to the NEWS.md file pointing to this work ?

[amartinhuertas]

Hi @kishore-nori ! I have finished a first round of review. Please address or rebut my comments/remarks/etc..

[kishore-nori]

Thank you very much for the comments Alberto, I ll make changes accordingly following our discussion in comments here.

[kishore-nori]

Hi @amartinhuertas and @santiagobadia, I have rewritten the implementation (realised I could do it in a good way, after our discussion to approach it as quadrature). This now works for evaluations and also solving the PDE with DiracDelta at a single or even a set of Points. I added tests to compare with both the exact solution and the earlier tag based implementation of DiracDelta. I also incorporated the earlier review comments from our discussions. Would be happy to know further comments for improvement.

[amartinhuertas]

Hi @kishore-nori ! nice work! I finished with my review. Please answer/rebut my comments.

[codecov-commenter]

Codecov Report

Merging #837 (c5d38b1) into master (173e625) will increase coverage by 0.04%.
The diff coverage is 97.82%.

@@            Coverage Diff             @@
##           master     #837      +/-   ##
==========================================
+ Coverage   88.29%   88.33%   +0.04%     
==========================================
  Files         164      164              
  Lines       19079    19120      +41     
==========================================
+ Hits        16845    16889      +44     
+ Misses       2234     2231       -3     

Impacted Files	Coverage Δ
src/CellData/DiracDeltas.jl	93.42% <97.82%> (+7.70%)	⬆️
src/Geometry/Triangulations.jl	89.29% <0.00%> (+0.73%)	⬆️
src/CellData/CellQuadratures.jl	79.78% <0.00%> (+1.06%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more