Meziantou.FrameworkOverview
editEvery language package builds the same kind of tree, with the types of Meziantou.Framework.Language. What you learn with one language applies to the others: only the node types and the SyntaxKind values differ.
A SyntaxTree holds the text it was parsed from and a root node. Nodes hold other nodes and tokens, the words and the punctuation of the text; tokens hold trivia, the whitespace and comments around them. Every character of the text belongs to exactly one token or trivia, so the tree writes the text back exactly.
| Page | Covers |
|---|---|
| Nodes and tokens | The parts of a tree, kinds, lists, invalid text in the tree, and finding a node |
| Trivia | Whitespace and comments, and the difference between Span and FullSpan |
| Source text and positions | SourceText, TextSpan, lines, and Location |
| Diagnostics | The problems a parser reports, their severity, and their position |
| Editing trees | Replacing, inserting and removing nodes, building them, and editing by text |
| Annotations | Marking a node to find it again after an edit |
| Visitors and rewriters | Walking a tree, and building a new one from it |
Immutable and shared
A tree never changes. An edit returns a new tree that shares every node it did not touch with the old one, so:
- you can keep several versions of a document for little more memory than one;
- you can read a tree from several threads;
- a node you hold stays valid after an edit, but belongs to the old tree: find its counterpart in the new tree with an annotation.
Parsing never throws
Whatever the text says, the parser returns a tree that reproduces it, and reports what is wrong as diagnostics. It keeps the shape the author most likely meant, so one mistake gives one diagnostic rather than a cascade, and the rest of the document is read as usual.