perf: incremental std.join + asciiSafe coverage improvements

sjsonnet/src-js/sjsonnet/CharSWAR.scala

@@ -23,10 +23,11 @@ object CharSWAR {
     false
   }
 
-  def isAsciiJsonSafe(s: String): Boolean = {
-    var i = 0
-    val len = s.length
-    while (i < len) {
+  def isAsciiJsonSafe(s: String): Boolean = isAsciiJsonSafe(s, 0, s.length)
+
+  def isAsciiJsonSafe(s: String, from: Int, to: Int): Boolean = {
+    var i = from
+    while (i < to) {
       val c = s.charAt(i)
       if (c < 32 || c == '"' || c == '\\' || c >= 128) return false
       i += 1

sjsonnet/src-js/sjsonnet/Platform.scala

@@ -38,6 +38,9 @@ object Platform {
   def isAsciiJsonSafe(s: String): Boolean =
     CharSWAR.isAsciiJsonSafe(s)
 
+  def isAsciiJsonSafe(s: String, from: Int, to: Int): Boolean =
+    CharSWAR.isAsciiJsonSafe(s, from, to)
+
   private def nodeToJson(node: Node): ujson.Value = node match {
     case _: Node.ScalarNode =>
       YamlDecoder.forAny.construct(node).getOrElse("") match {

sjsonnet/src-jvm/sjsonnet/CharSWAR.java

@@ -91,12 +91,20 @@ static boolean hasEscapeChar(byte[] arr, int from, int to) {
      * encoding step: all chars must be printable ASCII excluding {@code '"'} and {@code '\\'}.
      */
     static boolean isAsciiJsonSafe(String str) {
-        int len = str.length();
+        return isAsciiJsonSafe(str, 0, str.length());
+    }
+
+    /**
+     * Range variant of {@link #isAsciiJsonSafe(String)}. Used by the format parser to scan literal
+     * windows of a format string without allocating substrings.
+     */
+    static boolean isAsciiJsonSafe(String str, int from, int to) {
+        int len = to - from;
         if (len < 8) {
-            return isAsciiJsonSafeScalar(str, len);
+            return isAsciiJsonSafeScalar(str, from, to);
         }
-        int i = 0;
-        int limit = len - 3; // 4 UTF-16 chars per word
+        int i = from;
+        int limit = to - 3; // 4 UTF-16 chars per word
         while (i < limit) {
             long word =
                     ((long) str.charAt(i)) |
@@ -106,7 +114,7 @@ static boolean isAsciiJsonSafe(String str) {
             if (swarHasUnsafeAsciiChar(word)) return false;
             i += 4;
         }
-        while (i < len) {
+        while (i < to) {
             char c = str.charAt(i);
             if (c < 32 || c == '"' || c == '\\' || c >= 128) return false;
             i++;
@@ -216,8 +224,8 @@ private static boolean hasEscapeCharScalar(String s, int len) {
         return false;
     }
 
-    private static boolean isAsciiJsonSafeScalar(String s, int len) {
-        for (int i = 0; i < len; i++) {
+    private static boolean isAsciiJsonSafeScalar(String s, int from, int to) {
+        for (int i = from; i < to; i++) {
             char c = s.charAt(i);
             if (c < 32 || c == '"' || c == '\\' || c >= 128) return false;
         }

sjsonnet/src-jvm/sjsonnet/Platform.scala

@@ -33,6 +33,9 @@ object Platform {
   def isAsciiJsonSafe(s: String): Boolean =
     CharSWAR.isAsciiJsonSafe(s)
 
+  def isAsciiJsonSafe(s: String, from: Int, to: Int): Boolean =
+    CharSWAR.isAsciiJsonSafe(s, from, to)
+
   def gzipBytes(b: Array[Byte]): String = {
     val outputStream: ByteArrayOutputStream = new ByteArrayOutputStream(b.length)
     val gzip: GZIPOutputStream = new GZIPOutputStream(outputStream)

sjsonnet/src-native/sjsonnet/CharSWAR.scala

@@ -73,12 +73,14 @@ object CharSWAR {
     false
   }
 
-  def isAsciiJsonSafe(s: String): Boolean = {
-    val len = s.length
-    if (len < 8) return isAsciiJsonSafeScalar(s, len)
+  def isAsciiJsonSafe(s: String): Boolean = isAsciiJsonSafe(s, 0, s.length)
 
-    var i = 0
-    val limit = len - 3
+  def isAsciiJsonSafe(s: String, from: Int, to: Int): Boolean = {
+    val len = to - from
+    if (len < 8) return isAsciiJsonSafeScalar(s, from, to)
+
+    var i = from
+    val limit = to - 3
     while (i < limit) {
       val word =
         (s.charAt(i).toLong) |
@@ -88,7 +90,7 @@ object CharSWAR {
       if (swarHasUnsafeAsciiChar(word)) return false
       i += 4
     }
-    while (i < len) {
+    while (i < to) {
       val c = s.charAt(i)
       if (c < 32 || c == '"' || c == '\\' || c >= 128) return false
       i += 1
@@ -166,9 +168,9 @@ object CharSWAR {
   @inline private def zero16(word: Long): Long =
     ~((word & U16_HOLE) + U16_HOLE | word | U16_HOLE)
 
-  @inline private def isAsciiJsonSafeScalar(s: String, len: Int): Boolean = {
-    var i = 0
-    while (i < len) {
+  @inline private def isAsciiJsonSafeScalar(s: String, from: Int, to: Int): Boolean = {
+    var i = from
+    while (i < to) {
       val c = s.charAt(i)
       if (c < 32 || c == '"' || c == '\\' || c >= 128) return false
       i += 1

sjsonnet/src-native/sjsonnet/Platform.scala

@@ -38,6 +38,9 @@ object Platform {
   def isAsciiJsonSafe(s: String): Boolean =
     CharSWAR.isAsciiJsonSafe(s)
 
+  def isAsciiJsonSafe(s: String, from: Int, to: Int): Boolean =
+    CharSWAR.isAsciiJsonSafe(s, from, to)
+
   def gzipBytes(b: Array[Byte]): String = {
     val outputStream: ByteArrayOutputStream = new ByteArrayOutputStream(b.length)
     val gzip: GZIPOutputStream = new GZIPOutputStream(outputStream)

sjsonnet/src/sjsonnet/Format.scala

@@ -294,20 +294,13 @@ object Format {
   }
 
   /**
-   * Scalar ASCII-JSON-safe check over a substring window of `s`. Matches the predicate used by
+   * ASCII-JSON-safe check over a substring window of `s`. Matches the predicate used by
    * [[Platform.isAsciiJsonSafe]] (printable ASCII, no `"` or `\`). Used at format-parse time so the
    * result can be cached on [[RuntimeFormat]] and combined with per-value ASCII-safety at format
    * time.
    */
-  private def isAsciiJsonSafeRange(s: String, from: Int, to: Int): Boolean = {
-    var i = from
-    while (i < to) {
-      val c = s.charAt(i)
-      if (c < 32 || c == '"' || c == '\\' || c >= 128) return false
-      i += 1
-    }
-    true
-  }
+  private def isAsciiJsonSafeRange(s: String, from: Int, to: Int): Boolean =
+    Platform.isAsciiJsonSafe(s, from, to)
 
   /**
    * Hand-written format string scanner. Replaces the fastparse-based parser with direct