The idea of the optimizations used in app-demo-optimized
(see the repo)
is to parse the document text when the
document is first opened. The resulting parse
tree is stored as
model.lastAst. Each block in the AST carries
a label (version, id): (Int, Int), where
the id is unique to each block.
Each time the text changes, a new AST is computed
with an incremented version number. The
function Diff.mergeWith equals is applied to
the old and new ASTs
to compute an updated AST. The updated AST
is identical to the new AST except for the id's.
The id of a node in the updated AST is
the same as in the old AST if and only if the type, level,
and content of the node has not changed.
This information is used to signal MathJax not
to re-render mathematical text that is unchanged.
To see where these optimizations are applied,
look for the places in app-demo-optimized/Main.elm
where functions in the modules
ParseWithId and Markdown.ElmWithId are called.