add 🐬 functionality #189
add 🐬 functionality #189
Conversation
|
Ok this is not correct yet. I will have to investigate further. |
There was a problem hiding this comment.
Definitely like the ability to have this!
Just as a sidenote, I'm fine if this is not possible or we don't have the time for this yet:
Do you know if there is any way to make this consistently work with fermions? For example, what happens if we define flip to be:
iin codomain and!isdual-> flipiin codomain andisdual-> flip + twistiin domain and!isdual-> flip + twistiin domain andisdual-> flip
I think In some sense, what I am thinking of is thatcontractwould basically become equivalent toflipping all contracted indices to be<bra|ket>instead of the other way around, and then contracting. I'm just not entirely sure if this is consistent, as in that it commutes with arbitrary permutations + contractions
Also +1 on the PR title
src/tensors/indexmanipulations.jl
Outdated
| end | ||
| return t | ||
| end | ||
| flip(t::TensorMap, I) = TensorMap(t.data, flip(space(t), I)) |
There was a problem hiding this comment.
I think here it's probably best to also include a copy keyword, similar to permute, which can be used to control the sharing of data. At least, it's good to have this in the docstring somewhere
|
Ok, this turned out slightly more complicated as anticipated, because of Frobenius Schur factors. However, I think the implementation is now fixed, and includes the suggestion of @lkdvos to also include twists. I think this guarantees that if both spaces in a pair of |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #189 +/- ##
==========================================
+ Coverage 82.95% 83.16% +0.20%
==========================================
Files 42 42
Lines 5223 5257 +34
==========================================
+ Hits 4333 4372 +39
+ Misses 890 885 -5 ☔ View full report in Codecov by Sentry. |
Flipping tensor indices, for @Adwait-Naravane and @VictorVanthilt .
Be careful with fermions, as shown in the tests, simply flipping both of a pair of contracted indices does not yield the same contraction when fermions are involved.