1 # Design and open questions about libsyntax
4 The high-level description of the architecture is in RFC.md. You might
5 also want to dig through https://github.com/matklad/fall/ which
6 contains some pretty interesting stuff build using similar ideas
7 (warning: it is completely undocumented, poorly written and in general
8 not the thing which I recommend to study (yes, this is
13 The centerpiece of this whole endeavor is the syntax tree, in the
14 `tree` module. Open questions:
16 - how to best represent errors, to take advantage of the fact that
17 they are rare, but to enable fully-persistent style structure
18 sharing between tree nodes?
20 - should we make red/green split from Roslyn more pronounced?
22 - one can layout nodes in a single array in such a way that children
23 of the node form a continuous slice. Seems nifty, but do we need it?
25 - should we use SoA or AoS for NodeData?
27 - should we split leaf nodes and internal nodes into separate arrays?
28 Can we use it to save some bits here and there? (leaves don't need
29 first_child field, for example).
34 The syntax tree is produced using a three-staged process.
36 First, a raw text is split into tokens with a lexer (the `lexer` module).
37 Lexer has a peculiar signature: it is an `Fn(&str) -> Token`, where token
38 is a pair of `SyntaxKind` (you should have read the `tree` module and RFC
39 by this time! :)) and a len. That is, lexer chomps only the first
40 token of the input. This forces the lexer to be stateless, and makes
41 it possible to implement incremental relexing easily.
43 Then, the bulk of work, the parser turns a stream of tokens into
44 stream of events (the `parser` module; of particular interest are
45 the `parser/event` and `parser/parser` modules, which contain parsing
46 API, and the `parser/grammar` module, which contains actual parsing code
47 for various Rust syntactic constructs). Not that parser **does not**
48 construct a tree right away. This is done for several reasons:
50 * to decouple the actual tree data structure from the parser: you can
51 build any data structure you want from the stream of events
53 * to make parsing fast: you can produce a list of events without
56 * to make it easy to tweak tree structure. Consider this code:
63 Here, the attribute and the `pub` keyword must be the children of
64 the `fn` node. However, when parsing them, we don't yet know if
65 there would be a function ahead: it very well might be a `struct`
66 there. If we use events, we generally don't care about this *in
67 parser* and just spit them in order.
69 * (Is this true?) to make incremental reparsing easier: you can reuse
70 the same rope data structure for all of the original string, the
71 tokens and the events.
74 The parser also does not know about whitespace tokens: it's the job of
75 the next layer to assign whitespace and comments to nodes. However,
76 parser can remap contextual tokens, like `>>` or `union`, so it has
79 And at last, the TreeBuilder converts a flat stream of events into a
80 tree structure. It also *should* be responsible for attaching comments
81 and rebalancing the tree, but it does not do this yet :)
85 Parser and lexer accept a lot of *invalid* code intentionally. The
86 idea is to post-process the tree and to proper error reporting,
87 literal conversion and quick-fix suggestions. There is no
88 design/implementation for this yet.
93 Nothing yet, see `AstNode` in `fall`.