Refactored sheerka class: splitted to use sub handlers. Refactored unit tests to use classes.

2020-01-22 17:49:28 +01:00
parent 821614a6c4
commit c489a38ebc
120 changed files with 7947 additions and 8190 deletions
@@ -0,0 +1,43 @@
+# How to serialize ?
+
+## General rule
+- 1 byte : type of object code
+- int : version of the encoder
+- data : can be the json representation of the object
+
+### Current supported types
+- E : events
+- J : Json object (with history management)
+- P : pickle (no history)
+- S : state (history, but not managed by the serializer )
+- C : concept (with history management)
+- D : concept definitions (no history management)
+- R : executionContext ('R' stands for Result or ReturnValue, no history management)
+
+## How concepts are serialized ?
+- get the id of the concept
+- get the hash of the concept −> it will be its unique key  
+structure of the serialisation:
+```json
+{
+  "id" : "id",
+  "parent": <hash code of the previous version of the concept> or "",
+  "name": <name of the concept>,
+  "where": "",
+  "pre": "",
+  "post": "",
+  "body": "",
+  "desc": "",
+  ...
+}
+``` 
+
+## Idea to manage ObjectSerializer
+Problem:  
+During serialization, there is no issue. The match() method is the unique way to get the correct serialier.  
+During the deserialisation, all Object serializer have type = '0' and version = 1.  
+So how to choose the correct one ?  
+    A possible solution will be to add the type of the object to deserialize to the saved stream  
+    --> SHA256 for every object. Too much data saved.  
+The id is to let to inc the version automatically in the Serialiser (during the registration) and to keep the mapping within sdp.state
+