update the docs

libnixf/README.md

@@ -1,3 +1,5 @@
+[TOC]
+
 ## `libnixf`, The nix frontend
 
 This is a nix language frontend (lexer + parser, semantic analysis) focusing on editing experience.
@@ -11,3 +13,64 @@ Currently the package has these following components.
 * Sema (Semantic Analysis, e.g. duplicated keys)
 * Bytecode (experimental)
 
+## The `Basic` module
+
+This module mainly exposing nixf interfaces, and it's definitions are shared among other modules.
+
+### `Diagnostic` and Fixes
+
+A `Diagnostic` have a set of `Note`s attached to it.
+And each diagnostic has a set of fixes. Each fix has a name briefly describe the plan.
+For example
+
+* "insert ;"
+* "remove ...".
+
+And each fix has a series of `Edit`s. The removing range, and the new text to replace it.
+
+These "fixes" could be applied automatically, and also may pop a "💡" and let you choose that fix.
+
+### AST Nodes
+
+Each node of AST is owned by it's parent.
+The node is *immutable* right after it is constructed.
+
+libnixf AST has some semantic information attached to it.
+These semantic information, is constructed by *Semantic Actions*, in the `Sema` module.
+
+## The `Parse` module
+
+### The parsing algorithm
+
+This is a recursive descent parser not completely consistent with the official one in case of error recovery.
+
+
+The goal is:
+
+* Provide consistent experience in case of the code can be correctly *evaluated*
+* Provide much better experience in case of the code definitely cannot be correctly *evaluated*.
+
+For example: the text below will be parsed differently in libnixf.
+
+```nix
+{
+    a = 1  # <-- look at here
+    b = 2;
+}
+```
+
+
+In official nix parser it will be parsed as `a = (Call 1, b)`, for `a` attribute.
+
+However in libnixf parser, since `1` will definitely NOT be evaluated to lambda.
+Thus definitely you will not "Call" it.
+
+So the result would be:
+
+
+```nix
+{
+    a = 1  # <-- missing semicolon ";"
+    b = 2; # Correctly parse this as "b = 2"
+}
+```