make tensor type representation internal #176
Conversation
Codecov Report
@@ Coverage Diff @@
## dev #176 +/- ##
==========================================
+ Coverage 67.05% 67.09% +0.04%
==========================================
Files 13 13
Lines 4835 4835
Branches 1202 1202
==========================================
+ Hits 3242 3244 +2
+ Misses 930 928 -2
Partials 663 663
|
|
@gbaydin I think we can just pull this - these union cases should be hidden going forward, since the API is used to access/create them |
|
@dsyme ok let's do it. Note: I think that the Tensor, TensorF, TensorR cases should be visible to the user and seeing the discriminated union cases in tensor's string representation was taking care of this in a simple way. This is so that they are aware of what their code is doing in debugging and prototyping. I don't know if making this internal hides this from the user. I think we can still give this information to the user by simply adding "F" and "R" suffixes somewhere in the Tensor's string representation (which I hope will be shown by default by fsi and dotnet interactive). |
The cases in the discriminated union should be hidden from the rest of the world so we can change it in the future, and control how the reverse/forward cases are generated
opUnary and opBinary need to be hidden because they are inlined. This relates to #89 where we can provide public entry-points for user-defined extensions