From 501fdce29e4d2c46c4708ad7f44056878e0db3fa Mon Sep 17 00:00:00 2001
From: Ruben Beltran del Rio <jj@r.bdr.sh>
Date: Tue, 16 Jun 2026 12:08:21 +0200
Subject: Initial extraction from Norganize

---
 Sources/NorgKit/Helpers/ASCIIByteSet.swift | 30 ++++++++++
 Sources/NorgKit/Helpers/ASCIIHelper.swift  |  5 ++
 Sources/NorgKit/Helpers/TextHelper.swift   | 95 ++++++++++++++++++++++++++++++
 3 files changed, 130 insertions(+)
 create mode 100644 Sources/NorgKit/Helpers/ASCIIByteSet.swift
 create mode 100644 Sources/NorgKit/Helpers/ASCIIHelper.swift
 create mode 100644 Sources/NorgKit/Helpers/TextHelper.swift

(limited to 'Sources/NorgKit/Helpers')

diff --git a/Sources/NorgKit/Helpers/ASCIIByteSet.swift b/Sources/NorgKit/Helpers/ASCIIByteSet.swift
new file mode 100644
index 0000000..59efbe7
--- /dev/null
+++ b/Sources/NorgKit/Helpers/ASCIIByteSet.swift
@@ -0,0 +1,30 @@
+/// A membership test over the ASCII range (bytes `0`–`127`)
+/// We use this for performance, since markers are always ASCII, and this this
+/// faster than using Character.
+struct ASCIIByteSet {
+  private let low: UInt64
+  private let high: UInt64
+
+  /// Builds a set from the ASCII values.
+  init(_ characters: String) {
+    var lo: UInt64 = 0
+    var hi: UInt64 = 0
+    for byte in characters.utf8 {
+      if byte < 64 {
+        lo |= 1 << UInt64(byte)
+      } else if byte < 128 {
+        hi |= 1 << UInt64(byte - 64)
+      }
+    }
+    low = lo
+    high = hi
+  }
+
+  /// Whether `byte` is in the set.
+  @inline(__always)
+  func contains(_ byte: UInt8) -> Bool {
+    if byte < 64 { return low & (1 << UInt64(byte)) != 0 }
+    if byte < 128 { return high & (1 << UInt64(byte &- 64)) != 0 }
+    return false
+  }
+}
diff --git a/Sources/NorgKit/Helpers/ASCIIHelper.swift b/Sources/NorgKit/Helpers/ASCIIHelper.swift
new file mode 100644
index 0000000..98cce42
--- /dev/null
+++ b/Sources/NorgKit/Helpers/ASCIIHelper.swift
@@ -0,0 +1,5 @@
+struct ASCIIHelper {
+  static func isWhitespace(_ b: UInt8) -> Bool {
+    b == 0x20 || (0x09...0x0D).contains(b)
+  }
+}
diff --git a/Sources/NorgKit/Helpers/TextHelper.swift b/Sources/NorgKit/Helpers/TextHelper.swift
new file mode 100644
index 0000000..b0d45fe
--- /dev/null
+++ b/Sources/NorgKit/Helpers/TextHelper.swift
@@ -0,0 +1,95 @@
+import Foundation
+
+/// Text helpers that avoid `CharacterSet` and allocations for performance.
+enum TextHelper {
+
+  /// Splits text into lines as slices.
+  static func lineSlices(_ text: String) -> [Substring] {
+    var result: [Substring] = []
+    enumerateLines(in: text) { line, _ in result.append(line) }
+    return result
+  }
+
+  /// Invokes `body` once per line, passing a `Substring` view and its index.
+  static func enumerateLines(in text: String, _ body: (Substring, Int) -> Void) {
+    let utf8 = text.utf8
+    let end = utf8.endIndex
+    var lineStart = utf8.startIndex
+    var i = utf8.startIndex
+    var index = 0
+    while i < end {
+      if utf8[i] == 0x0A {
+        body(text[lineStart..<strippedLineEnd(utf8, from: lineStart, to: i)], index)
+        index += 1
+        lineStart = utf8.index(after: i)
+      }
+      i = utf8.index(after: i)
+    }
+    body(text[lineStart..<strippedLineEnd(utf8, from: lineStart, to: end)], index)
+  }
+
+  /// The end index of a line `[start, newline)`, backed up by one when the last
+  /// byte is a carriage return. Operating on the UTF-8 view avoids the
+  /// grapheme decode that `Substring.last` would pay on every line.
+  private static func strippedLineEnd(
+    _ utf8: String.UTF8View, from start: String.Index, to newline: String.Index
+  ) -> String.Index {
+    guard newline > start else { return newline }
+    let last = utf8.index(before: newline)
+    return utf8[last] == 0x0D ? last : newline
+  }
+
+  /// Returns the line at `index` (zero-based) as a `Substring`, or `nil` when
+  /// there is no such line. Trailing `\r` is stripped and line counting matches
+  /// ``lines(_:)`` / ``enumerateLines(in:_:)``. Stops as soon as the line is
+  /// found, so locating an early line in a large document is cheap.
+  static func line(in text: String, at index: Int) -> Substring? {
+    guard index >= 0 else { return nil }
+    let utf8 = text.utf8
+    let end = utf8.endIndex
+    var lineStart = utf8.startIndex
+    var i = utf8.startIndex
+    var current = 0
+    while i < end {
+      if utf8[i] == 0x0A {
+        if current == index {
+          return text[lineStart..<strippedLineEnd(utf8, from: lineStart, to: i)]
+        }
+        current += 1
+        lineStart = utf8.index(after: i)
+      }
+      i = utf8.index(after: i)
+    }
+    return current == index
+      ? text[lineStart..<strippedLineEnd(utf8, from: lineStart, to: end)] : nil
+  }
+
+  /// Trims leading and trailing whitespace from a slice using
+  /// `Character.isWhitespace`, avoiding the `CharacterSet` bridging cost that
+  /// `trimmingCharacters(in:)` pays per call. The result is a slice of the
+  /// input, so nothing is copied. This is the single trimming primitive used
+  /// across the parser, scanner, and detached-modifier recogniser.
+  static func whitespaceTrimmed(_ s: Substring) -> Substring {
+    // Trim over the UTF-8 view: Norg indentation and trailing space is always
+    // ASCII whitespace, whose bytes are all `< 0x80` and so never part of a
+    // multi-byte scalar — the trimmed bounds stay on scalar boundaries. This
+    // avoids the grapheme decode `Character.isWhitespace` pays on every line.
+    let utf8 = s.utf8
+    var start = utf8.startIndex
+    var end = utf8.endIndex
+    while start < end, ASCIIHelper.isWhitespace(utf8[start]) { start = utf8.index(after: start) }
+    while start < end {
+      let prev = utf8.index(before: end)
+      guard ASCIIHelper.isWhitespace(utf8[prev]) else { break }
+      end = prev
+    }
+    return s[start..<end]
+  }
+
+  /// Trims leading and trailing whitespace.
+  static func whitespaceTrimmed(_ s: String) -> String {
+    let trimmed = whitespaceTrimmed(s[...])
+    return trimmed.startIndex == s.startIndex && trimmed.endIndex == s.endIndex
+      ? s : String(trimmed)
+  }
+}
-- 
cgit