Port guidance's JSON implementation

parser/Cargo.toml

@@ -15,6 +15,8 @@ jsonschema = { version = "0.24.0", default-features = false, optional = true }
 url = "2.5.2"
 lazy_static = { version = "1.5.0", optional = true }
 regex-syntax = "0.8.5"
+indexmap = "2.6.0"
+referencing = "0.26.1"
 
 [features]
 default = ["jsonschema_validation", "lark"]

parser/build.rs

@@ -25,7 +25,10 @@ fn main() {
             },
             |bindings| {
                 bindings.write_to_file("llguidance.h");
-                bindings.write_to_file(format!("{}/../../../llguidance.h", env::var("OUT_DIR").unwrap()));
+                bindings.write_to_file(format!(
+                    "{}/../../../llguidance.h",
+                    env::var("OUT_DIR").unwrap()
+                ));
             },
         );
 }

parser/src/earley/from_guidance.rs

@@ -97,7 +97,7 @@ fn grammar_from_json(
                 input.lark_grammar.is_none(),
                 "cannot have both json_schema and lark_grammar"
             );
-            let opts = JsonCompileOptions { compact: false };
+            let opts: JsonCompileOptions = JsonCompileOptions::default();
             opts.json_to_llg(json_schema)?
         } else {
             lark_to_llguidance(input.lark_grammar.as_ref().unwrap())?

parser/src/ffi.rs

@@ -288,7 +288,7 @@ fn new_constraint_json(init: &LlgConstraintInit, json_schema: *const c_char) ->
         .map_err(|_| anyhow::anyhow!("Invalid UTF-8 in json_schema"))?;
     let json_schema = serde_json::from_str(json_schema)
         .map_err(|e| anyhow::anyhow!("Invalid JSON in json_schema: {e}"))?;
-    let opts = JsonCompileOptions { compact: false };
+    let opts = JsonCompileOptions::default();
     let grammar = opts
         .json_to_llg(&json_schema)
         .map_err(|e| anyhow::anyhow!("Error compiling JSON schema to LLG: {e}"))?;

parser/src/grammar_builder.rs

@@ -6,7 +6,7 @@ use crate::api::{
     GrammarWithLexer, Node, NodeId, NodeProps, RegexId, RegexNode, RegexSpec, TopLevelGrammar,
 };
 
-#[derive(Clone, Copy, PartialEq, Eq, Debug)]
+#[derive(Clone, Copy, PartialEq, Eq, Debug, Hash)]
 pub struct NodeRef {
     idx: usize,
     grammar_id: u32,
@@ -90,6 +90,14 @@ impl RegexBuilder {
         self.add_node(RegexNode::Repeat(node, min, max))
     }
 
+    pub fn not(&mut self, node: RegexId) -> RegexId {
+        self.add_node(RegexNode::Not(node))
+    }
+
+    pub fn and(&mut self, nodes: Vec<RegexId>) -> RegexId {
+        self.add_node(RegexNode::And(nodes))
+    }
+
     fn finalize(&mut self) -> Vec<RegexNode> {
         let r = std::mem::take(&mut self.nodes);
         *self = Self::new();