---

yaml
---
r: 126695
b: refs/heads/snap-stage3
c: 1a1a9d5
h: refs/heads/master
i:
  126693: b5323b5
  126691: 0f926f3
  126687: 3504803
v: v3

Author: emberian

[refs]

@@ -1,7 +1,7 @@
 ---
 refs/heads/master: 7be8f0af0393dcdb077c2f6b1653836fd3fba235
 refs/heads/snap-stage1: e33de59e47c5076a89eadeb38f4934f58a3618a6
-refs/heads/snap-stage3: 19e1f5cdb6a47070fd5f12993e947ea6db0eb5dd
+refs/heads/snap-stage3: 1a1a9d54456355cc0ebdd397fd04871abe27f78c
 refs/heads/try: 502e4c045236682e9728539dc0d2b3d0b237f55c
 refs/tags/release-0.1: 1f5c5126e96c79d22cb7862f75304136e204f105
 refs/heads/ndm: f3868061cd7988080c30d6d5bf352a5a5fe2460b

branches/snap-stage3/src/grammar/raw-string-literal-ambiguity.md

@@ -0,0 +1,29 @@
+Rust's lexical grammar is not context-free. Raw string literals are the source
+of the problem. Informally, a raw string literal is an `r`, followed by `N`
+hashes (where N can be zero), a quote, any characters, then a quote followed
+by `N` hashes. This grammar describes this as best possible:
+
+    R -> 'r' S
+    S -> '"' B '"'
+    S -> '#' S '#'
+    B -> . B
+    B -> ε
+
+Where `.` represents any character, and `ε` the empty string. Consider the
+string `r#""#"#`. This string is not a valid raw string literal, but can be
+accepted as one by the above grammar, using the derivation:
+
+    R : #""#"#
+    S : ""#"
+    S : "#
+    B : #
+    B : ε
+
+(Where `T : U` means the rule `T` is applied, and `U` is the remainder of the
+string.) The difficulty arises from the fact that it is fundamentally
+context-sensitive. In particular, the context needed is the number of hashes.
+I know of no way to resolve this, but also have not come up with a proof that
+it is not context sensitive. Such a proof would probably use the pumping lemma
+for context-free languages, but I (cmr) could not come up with a proof after
+spending a few hours on it, and decided my time best spent elsewhere. Pull
+request welcome!