From 5ae9686adac1508873f2d980e84becd3496244c2 Mon Sep 17 00:00:00 2001
From: Karen Arutyunov <karen@codesynthesis.com>
Date: Wed, 26 Feb 2020 17:16:45 +0300
Subject: Add notion of validator to char_scanner and make sure manifest is
 UTF-8

This involves implementing utf8_validator and UTF-8 utility functions and
using them during the manifest parsing, serialization, and rewriting.
---
 libbutl/unicode.mxx | 82 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 82 insertions(+)
 create mode 100644 libbutl/unicode.mxx

(limited to 'libbutl/unicode.mxx')

diff --git a/libbutl/unicode.mxx b/libbutl/unicode.mxx
new file mode 100644
index 0000000..b846476
--- /dev/null
+++ b/libbutl/unicode.mxx
@@ -0,0 +1,82 @@
+// file      : libbutl/unicode.mxx -*- C++ -*-
+// license   : MIT; see accompanying LICENSE file
+
+#ifndef __cpp_modules_ts
+#pragma once
+#endif
+
+// C includes.
+
+#ifndef __cpp_lib_modules_ts
+#include <string>
+#include <ostream>
+#include <cstdint> // uint16_t
+#endif
+
+// Other includes.
+
+#ifdef __cpp_modules_ts
+export module butl.unicode;
+#ifdef __cpp_lib_modules_ts
+import std.core;
+import std.io;
+#endif
+#endif
+
+#include <libbutl/export.hxx>
+
+LIBBUTL_MODEXPORT namespace butl
+{
+  // Note that the Unicode Standard requires the surrogates ([D800 DFFF]) to
+  // only be used in the context of the UTF-16 character encoding form. Thus,
+  // we omit the surrogate codepoint type and assume surrogates as invalid
+  // codepoints.
+  //
+  enum class codepoint_types: std::uint16_t
+  {
+    // Useful to denote invalid codepoints or when building the type set
+    // incrementally.
+    //
+    none          = 0x00,
+
+    graphic       = 0x01, // L(etter), M(ark), N(number), P(uncturation),
+                          // S(symbol), Zs(separator, space)
+    format        = 0x02,
+    control       = 0x04,
+    private_use   = 0x08,
+    non_character = 0x10,
+    reserved      = 0x20,
+
+    any           = 0x3f
+  };
+
+  codepoint_types operator&  (codepoint_types,  codepoint_types);
+  codepoint_types operator|  (codepoint_types,  codepoint_types);
+  codepoint_types operator&= (codepoint_types&, codepoint_types);
+  codepoint_types operator|= (codepoint_types&, codepoint_types);
+
+  // Return the codepoint type for a valid codepoint value and none otherwise.
+  //
+  // Note that the valid codepoint ranges are [0 D800) and (DFFF 10FFFF].
+  //
+  codepoint_types
+  codepoint_type (char32_t);
+
+  // Return the type name for a single codepoint type and empty string for
+  // `none` and `any`.
+  //
+  // Potential future improvements:
+  //  - add the none value name parameter ("invalid" by default)
+  //  - produce names for type masks ("graphic, format", "any", etc)
+  //
+  std::string
+  to_string (codepoint_types);
+
+  inline std::ostream&
+  operator<< (std::ostream& os, codepoint_types ts)
+  {
+    return os << to_string (ts);
+  }
+}
+
+#include <libbutl/unicode.ixx>
-- 
cgit v1.1