Remove hardcoded newline injection from grammar generator

The grammar generator should not inject delimiters between samples —
the grammar itself defines the output format. Users who want
newline-delimited output can include "\n" in their start production.

Updated example grammars, tests, and docs accordingly.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: GeorgeHahn

examples/grammars/json.ebnf

@@ -1,3 +1,5 @@
+start = value , "\n" ;
+
 value = object | array | string | number | "true" | "false" | "null" ;
 
 object = "{" [ pair { "," pair } ] "}" ;

lading_payload/README.grammar.md

@@ -1,8 +1,9 @@
 # Grammar payload
 
 Generates structured data from EBNF, PEG, or ANTLR v4 grammar files using
-[barkus](https://github.com/DataDog/barkus). Each generated sample is a
-newline-delimited byte sequence conforming to the grammar.
+[barkus](https://github.com/DataDog/barkus). Samples are concatenated directly
+with no injected delimiters — if you need newline-delimited output, include a
+trailing `\n` in your grammar's start production.
 
 ## Using it in a lading config
 
@@ -69,6 +70,9 @@ generator:
 ## Example EBNF grammar (simplified JSON)
 
 ```ebnf
+(* Wrap the start production with a trailing newline for line-delimited output. *)
+start = value , "\n" ;
+
 value = object | array | string | number | "true" | "false" | "null" ;
 
 object = "{" [ pair { "," pair } ] "}" ;

lading_payload/src/grammar.rs

@@ -3,6 +3,10 @@
 //! This module wraps barkus — a structure-aware grammar-based data generator —
 //! as a lading payload type. It accepts EBNF, PEG, or ANTLR v4 grammar files
 //! and generates conforming structured output.
+//!
+//! Unlike other lading payload types, the grammar generator does **not** inject
+//! delimiters (e.g. newlines) between samples. If you need newline-delimited
+//! output, include the newline in your grammar's start production.
 
 use std::{io::Write, path::PathBuf};
 
@@ -161,13 +165,11 @@ impl crate::Serialize for Grammar {
                         continue;
                     }
                     consecutive_failures = 0;
-                    // +1 for the trailing newline
-                    let needed = sample.len() + 1;
+                    let needed = sample.len();
                     let Some(remainder) = bytes_remaining.checked_sub(needed) else {
                         break;
                     };
                     writer.write_all(&sample)?;
-                    writer.write_all(b"\n")?;
                     bytes_remaining = remainder;
                 }
                 Err(GenerateError::BudgetExhausted { .. }) => {
@@ -232,7 +234,8 @@ mod test {
 
     #[test]
     fn ebnf_generates_expected_output() {
-        let source = "greeting = \"hello\" | \"world\" ;";
+        // Grammar includes a trailing newline so each sample is its own line.
+        let source = "greeting = ( \"hello\" | \"world\" ) , \"\\n\" ;";
         let (_dir, config) = config_from_source(source, GrammarFormat::Ebnf, "ebnf");
 
         let mut grammar = Grammar::new(&config).unwrap();
@@ -256,7 +259,7 @@ mod test {
 
     #[test]
     fn peg_generates_expected_output() {
-        let source = "greeting <- \"hello\" / \"world\"";
+        let source = "greeting <- (\"hello\" / \"world\") \"\\n\"";
         let (_dir, config) = config_from_source(source, GrammarFormat::Peg, "peg");
 
         let mut grammar = Grammar::new(&config).unwrap();
@@ -278,7 +281,7 @@ mod test {
 
     #[test]
     fn antlr_generates_expected_output() {
-        let source = "grammar Test;\ngreeting : 'hello' | 'world' ;";
+        let source = "grammar Test;\ngreeting : ('hello' | 'world') '\\n' ;";
         let (_dir, config) = config_from_source(source, GrammarFormat::Antlr, "g4");
 
         let mut grammar = Grammar::new(&config).unwrap();