WIP

1 files changed, 68 insertions, 0 deletions

diff --git a/utf/src/utf8.cc b/utf/src/utf8.cc
new file mode 100644
index 0000000..54b0296
--- /dev/null
+++ b/utf/src/utf8.cc
@@ -0,0 +1,68 @@
+#include "utf8.hh"
+
+#include "utf_error.hh"
+
+namespace utf {
+
+namespace {
+
+inline bool valid_codepoint(uint32_t c) {
+  return (c < 0xd800) || (c > 0xdfff && c <= 0x10ffff);
+}
+
+}  // namespace
+
+uint32_t read8(std::string_view data, std::size_t& offset) {
+  if (offset >= data.size())
+    return NEED_MORE;
+  uint32_t ret;
+  uint8_t size;
+  switch (static_cast<uint8_t>(data[offset]) >> 4) {
+    case 15:
+      if (data[offset] & 0x08)
+        return INVALID;
+      ret = static_cast<uint32_t>(data[offset] & 0x07) << 18;
+      size = 4;
+      break;
+    case 14:
+      ret = static_cast<uint32_t>(data[offset] & 0x0f) << 12;
+      size = 3;
+      break;
+    case 13:
+    case 12:
+      ret = static_cast<uint32_t>(data[offset] & 0x1f) << 6;
+      size = 2;
+      break;
+    default:
+      if (data[offset] & 0x80)
+        return INVALID;
+      return data[offset++];
+  }
+  if (data.size() - offset < size)
+    return NEED_MORE;
+  for (uint8_t i = 1; i < size; ++i) {
+    if ((data[offset + i] & 0xc0) != 0x80)
+      return INVALID;
+    ret |= static_cast<uint32_t>(data[offset + i] & 0x3f) << (size - i - 1) * 6;
+  }
+  if (!valid_codepoint(ret))
+    return INVALID;
+  switch (size) {
+    case 4:
+      if (ret < 0x10000)
+        return INVALID;
+      break;
+    case 3:
+      if (ret < 0x800)
+        return INVALID;
+      break;
+    case 2:
+      if (ret < 0x80)
+        return INVALID;
+      break;
+  }
+  offset += size;
+  return ret;
+}
+
+}  // namespace utf