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.cxx | 165 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 165 insertions(+)
 create mode 100644 libbutl/unicode.cxx

(limited to 'libbutl/unicode.cxx')

diff --git a/libbutl/unicode.cxx b/libbutl/unicode.cxx
new file mode 100644
index 0000000..4219846
--- /dev/null
+++ b/libbutl/unicode.cxx
@@ -0,0 +1,165 @@
+// file      : libbutl/unicode.cxx -*- C++ -*-
+// license   : MIT; see accompanying LICENSE file
+
+#ifndef __cpp_modules_ts
+#include <libbutl/unicode.mxx>
+#endif
+
+#ifndef __cpp_lib_modules_ts
+#include <string>
+#include <ostream>
+#include <cstdint>
+
+#include <cstddef>   // size_t
+#include <utility>   // pair
+#include <algorithm> // lower_bound()
+#endif
+
+#ifdef __cpp_modules_ts
+module butl.unicode;
+
+// Only imports additional to interface.
+#ifdef __clang__
+#ifdef __cpp_lib_modules_ts
+import std.core;
+import std.io;
+#endif
+#endif
+
+#endif
+
+using namespace std;
+
+namespace butl
+{
+  // Sorted arrays of the Unicode codepoint ranges corresponding to the
+  // codepoint types (see the Types of Code Points table in the Unicode 12.0
+  // Standard for details). Note that code type range lists (but not ranges
+  // themselves) may overlap.
+  //
+  // Also note that the graphic type codepoints are numerous and scattered.
+  // Thus, we will consider a codepoint to be of the graphic type if it is not
+  // of any other type.
+  //
+  using codepoint_range = pair<char32_t, char32_t>;
+
+  static const codepoint_range cn_rs[] = // Control.
+  {
+    {0x00, 0x1F},
+    {0x7F, 0x9F}
+  };
+
+  static const codepoint_range fr_rs[] = // Format.
+  {
+    {0x000AD, 0x000AD},
+    {0x00600, 0x00605},
+    {0x0061C, 0x0061C},
+    {0x006DD, 0x006DD},
+    {0x0070F, 0x0070F},
+    {0x008E2, 0x008E2},
+    {0x0180E, 0x0180E},
+    {0x0200B, 0x0200F},
+    {0x0202A, 0x0202E},
+    {0x02060, 0x02064},
+    {0x02066, 0x0206F},
+    {0x0FEFF, 0x0FEFF},
+    {0x0FFF9, 0x0FFFB},
+    {0x110BD, 0x110BD},
+    {0x110CD, 0x110CD},
+    {0x13430, 0x13438},
+    {0x1BCA0, 0x1BCA3},
+    {0x1D173, 0x1D17A},
+    {0xE0001, 0xE0001},
+    {0xE0020, 0xE007F}
+  };
+
+  static const codepoint_range pr_rs[] = // Private-use.
+  {
+    {0x00E000, 0x00F8FF},
+    {0x0F0000, 0x10FFFF}
+  };
+
+  static const codepoint_range nc_rs[] = // Non-character.
+  {
+    {0xFDD0, 0xFDEF}
+  };
+
+  static const codepoint_range rs_rs[] = // Reserved.
+  {
+    {0x30000, 0xE0000},
+    {0xE0002, 0xE001F},
+    {0xE0080, 0xE00FF},
+    {0xE01F0, 0xEFFFF}
+  };
+
+  struct codepoint_type_ranges
+  {
+    codepoint_types type;
+    const codepoint_range* begin;
+    const codepoint_range* end;
+  };
+
+  static const codepoint_type_ranges ct_ranges[] =
+  {
+    {
+      codepoint_types::control,
+      cn_rs,
+      cn_rs + sizeof (cn_rs) / sizeof (*cn_rs)
+    },
+    {
+      codepoint_types::format,
+      fr_rs,
+      fr_rs + sizeof (fr_rs) / sizeof (*fr_rs)
+    },
+    {
+      codepoint_types::private_use,
+      pr_rs,
+      pr_rs + sizeof (pr_rs) / sizeof (*pr_rs)
+    },
+    {
+      codepoint_types::non_character,
+      nc_rs,
+      nc_rs + sizeof (nc_rs) / sizeof (*nc_rs)
+    },
+    {
+      codepoint_types::reserved,
+      rs_rs,
+      rs_rs + sizeof (rs_rs) / sizeof (*rs_rs)
+    }
+  };
+
+  // Return the codepoint type of a range if the codepoint value falls into
+  // one and the graphic type otherwise.
+  //
+  // Note that this is a type detection fallback (see codepoint_type() for
+  // details).
+  //
+  codepoint_types
+  codepoint_type_lookup (char32_t c)
+  {
+    // Note that the codepoint type range lists may overlap. Thus, we iterate
+    // over all of them until there is a match.
+    //
+    for (size_t i (0); i != sizeof (ct_ranges) / sizeof (*ct_ranges); ++i)
+    {
+      const codepoint_type_ranges& rs (ct_ranges[i]);
+
+      // Find the range that either contains the codepoint or lays to the
+      // right of it. Note that here we assume a range to be less than a
+      // codepoint value if it lays to the left of the codepoint.
+      //
+      const codepoint_range* r (
+        lower_bound (rs.begin, rs.end,
+                     c,
+                     [] (const codepoint_range& r, char32_t c)
+                     {
+                       return r.second < c;
+                     }));
+
+      if (r != rs.end && r->first <= c) // Contains the codepoint?
+        return rs.type;
+    }
+
+    return codepoint_types::graphic;
+  }
+}
-- 
cgit v1.1