first attempts at writing the serializers for the new types PhylogeneticModel and GroupBasedPhylogeneticModel #4010
Conversation
…ticModel and GroupBasedPhylogeneticModel
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #4010 +/- ##
==========================================
+ Coverage 84.58% 84.73% +0.14%
==========================================
Files 596 630 +34
Lines 81969 84510 +2541
==========================================
+ Hits 69336 71606 +2270
- Misses 12633 12904 +271
|
| save_object(s, graph(pm), :tree) # already implemented | ||
| save_object(s, number_states(pm), :states) # already implemented | ||
| save_object(s, probability_ring(pm), :probability_ring) # already implemented | ||
| save_object(s, root_distribution(pm), :root_distribution) # needs to be implemented, or we change the entry type to QQFieldElem |
There was a problem hiding this comment.
you can using save_typed_object here if the type isn't always know, or if your using a vector any because not all entries are the same type you'll need to use a tuple
src/Serialization/Combinatorics.jl
Outdated
| @@ -19,6 +20,19 @@ function load_object(s::DeserializerState, g::Type{Graph{T}}) where T <: Union{D | |||
| return g(smallobj) | |||
| end | |||
|
|
|||
| function save_object(s::SerializerState, e::Edge) | |||
| save_data_dict(s) do | |||
There was a problem hiding this comment.
this can be done using save_data_array instead, it'll make file sizes smaller
…d object Co-authored-by: antonydellavecchia <[email protected]>
…zation of dicts to catch the case where the key is not serialized with params
|
|
||
|
|
||
| function load_object(s::DeserializerState, g::Type{PhylogeneticModel}) | ||
| graph = load_object(s, Graph{Directed}, :tree) |
There was a problem hiding this comment.
here you would need to use
| graph = load_object(s, Graph{Directed}, :tree) | |
| graph = load_typed_object(s, :tree) |
…ests more elaborate. Changed testing for graphs so that false negatives are avoided.
|
In commit 8723fc9, tests did not pass so I investigated it and found out that apparently object identity does not the same with polymake-derived objects? Therefore, the graph objects of the original model and the loaded one were not identical, although they were the same tree. I wrote a workaround using |
I am currently trying to serialize the new data types introduced in the algebraic statistics (sub)project on algebraic phylogenetics by @marinagarrote et al (cf. #3812). Following the documentation and writing the code as naive as possible, I got problems when testing the functions by trying to save an instance of GroupPhylogeneticModel, with the following error message:
Unsupported type 'GroupBasedPhylogeneticModel' for encoding.