working changes to add defer support to algorithm

ALGORITHM.md

@@ -4,19 +4,24 @@ Execute Operation:
 
 - Inline named fragments! This is done prior to coercing variables, so is done in a separate step. (should be done later? not at all?)
 - Create a `RootPlan` with the transformed operation.
-- Create a node to hold the result.
+- **NEW FOR DEFER** create state to hold the composite new deferred fragment records and deferred grouped field sets.
+- Create a node to hold the final reconcilable object.
 - For each `SubschemaPlan` in the `RootPlan`:
   - Send the initial set of operations.
-  - As each operation result becomes available, compose them:
+  - As each operation result becomes available:
     - Create a `Stitch` with the help of the given `SubschemaPlan` and a pointer to the result node.
     - Stitch the `Stitch`.
-- Return the node.
+    - **NEW FOR DEFER** record each split deferred fragment as it completes, and complete a deferred fragment when all of the split portions have completed.  
+- **NEW FOR DEFER** Yield the initial result, parsing the final reconcilable object to exclude deferrable fields.
+- **NEW FOR DEFER** Yield subsequent results as they complete.
 
 Creating a `RootPlan`:
 
 - Coerce variables.
 - Apply skip/include directives (may depends on variables).
 - Collects root fields based on the appropriate root type.
+- **NEW FOR DEFER** update the state to hold any new composite deferred fragment records and deferred grouped field sets.
+- **NEW FOR DEFER** retain inline fragments for any deferred fragments.
 - Create a `FieldPlan` for the given root fields (with no base schema, such that the `FieldTree` for the `FieldPlan` will be empty).
 - Return the `FieldPlan`.
 
@@ -29,6 +34,7 @@ Creating a `FieldPlan`:
   - Determine where in the data structure to add the field, creating a new `SubschemaPlan` entry for `toSubschema` as necessary.
   - If the field has no sub-selections, add the field to the appropriate `SubschemaPlan` and return.
   - Split the sub-selections in two, those that belong to `toSubschema`, and those that don't.
+  - **NEW FOR DEFER** record how deferred fragments are split between subschemas, relabelling them as necessary.
   - Add the field with the sub-selections that belong to `toSubschema` to the appropriate `SubschemaPlan`.
   - Create a `FieldPlanMap` using the sub-selections that do not belong to the subschema (with `toSubschema` as a base).
   - If `toSubschema` is the same as `fromSubschema`, add the `FieldPlanMap` to the `FieldPlan`'s `FieldTree` under the response key for the field.
@@ -41,6 +47,7 @@ Creating a `FieldPlanMap`:
 - Initialize a map of possible types to `FieldPlan`s.
 - For each possible type of this field:
   - Collect the subfields from the sub-selections for the given type.
+  - **NEW FOR DEFER** retain inline fragments for any deferred fragments.
   - Create the `FieldPlan` for the given subfields using `fromSubschema`.
   - Within the map, set the entry for the given type to `FieldPlan`.
 - Return the map.
@@ -50,5 +57,8 @@ Result Composition (Stitching the `Stitch`):
 - If the data section of the result is null, null the pointer node and return.
 - Copy the `data` from all the results under the pointer node.
 - Create a map of subschemas to `Stitch`es.
-- Walk the `FieldTree` for the given `SubschemaPlan` creating new `Stitch`es for each subschema, updating the pointer as necessary, saving the stitches in the map.
+- Walk the `FieldTree` for the given `SubschemaPlan`:
+  - For each `SubschemaPlan`, creating a new `Stitch` with the appropriate pointer.
+  - Save the `Stitch` to the map.
+  - **NEW FOR DEFER** update the state to hold any new composite deferred fragment records and deferred grouped field set details.
 - Stitch the `Stitch`es.