Remove leading null from blobs and EMPTY_REF from getters

- Blobs now start directly with first string (no wasted \0 prefix)
- Lookup functions handle index 0 by pointing to trailing null of blob
- get_*_ref() simplified: just calls lookup, no separate empty check
- Removes EMPTY_REF constant and per-getter branch

esphome/core/entity_base.cpp

@@ -50,23 +50,16 @@ __attribute__((weak)) const char *entity_device_class_lookup(uint16_t) { return
 __attribute__((weak)) const char *entity_uom_lookup(uint16_t) { return ""; }
 __attribute__((weak)) const char *entity_icon_lookup(uint16_t) { return ""; }
 
-static constexpr auto EMPTY_REF = StringRef::from_lit("");
-
 // Entity device class (from packed index)
 StringRef EntityBase::get_device_class_ref() const {
-  uint16_t idx = (this->entity_string_packed_ >> ENTITY_STR_DC_SHIFT) & ENTITY_STR_DC_MASK;
-  if (idx == 0)
-    return EMPTY_REF;
-  return StringRef(entity_device_class_lookup(idx));
+  return StringRef(
+      entity_device_class_lookup((this->entity_string_packed_ >> ENTITY_STR_DC_SHIFT) & ENTITY_STR_DC_MASK));
 }
 std::string EntityBase::get_device_class() const { return std::string(this->get_device_class_ref().c_str()); }
 
 // Entity unit of measurement (from packed index)
 StringRef EntityBase::get_unit_of_measurement_ref() const {
-  uint16_t idx = (this->entity_string_packed_ >> ENTITY_STR_UOM_SHIFT) & ENTITY_STR_UOM_MASK;
-  if (idx == 0)
-    return EMPTY_REF;
-  return StringRef(entity_uom_lookup(idx));
+  return StringRef(entity_uom_lookup((this->entity_string_packed_ >> ENTITY_STR_UOM_SHIFT) & ENTITY_STR_UOM_MASK));
 }
 std::string EntityBase::get_unit_of_measurement() const {
   return std::string(this->get_unit_of_measurement_ref().c_str());
@@ -75,12 +68,9 @@ std::string EntityBase::get_unit_of_measurement() const {
 // Entity icon (from packed index)
 StringRef EntityBase::get_icon_ref() const {
 #ifdef USE_ENTITY_ICON
-  uint16_t idx = (this->entity_string_packed_ >> ENTITY_STR_ICON_SHIFT) & ENTITY_STR_ICON_MASK;
-  if (idx == 0)
-    return EMPTY_REF;
-  return StringRef(entity_icon_lookup(idx));
+  return StringRef(entity_icon_lookup((this->entity_string_packed_ >> ENTITY_STR_ICON_SHIFT) & ENTITY_STR_ICON_MASK));
 #else
-  return EMPTY_REF;
+  return StringRef(entity_icon_lookup(0));
 #endif
 }
 std::string EntityBase::get_icon() const { return std::string(this->get_icon_ref().c_str()); }

esphome/core/entity_helpers.py

@@ -84,13 +84,14 @@ def _generate_blob_and_offsets(
 ) -> tuple[bytes, list[int]]:
     """Build a packed blob and offset list from a string dict.
 
-    Returns (blob_bytes, offsets) where offsets[0] points to "" (empty)
-    and offsets[i] points to the i-th string.
+    Returns (blob_bytes, offsets) where offsets are 1-based (matching the
+    1-based indices returned by _register_string). Index 0 means "not set"
+    and is handled by the caller before reaching the lookup function.
     """
     # Sort by assigned index to ensure deterministic output
     sorted_strings = sorted(strings.items(), key=lambda x: x[1])
-    blob = bytearray(b"\x00")  # Index 0 = offset 0 = empty string
-    offsets = [0]  # offset for index 0 (empty string)
+    blob = bytearray()
+    offsets = []
     for s, _idx in sorted_strings:
         offsets.append(len(blob))
         blob.extend(s.encode("utf-8"))
@@ -104,7 +105,11 @@ def _generate_category_code(
     lookup_fn: str,
     strings: dict[str, int],
 ) -> str:
-    """Generate C++ code for one string category (blob + offsets + lookup)."""
+    """Generate C++ code for one string category (blob + offsets + lookup).
+
+    Index 0 means "not set" and is handled by get_*_ref() before calling
+    the lookup function. The lookup uses 1-based indexing directly.
+    """
     if not strings:
         return ""
 
@@ -116,12 +121,16 @@ def _generate_category_code(
     blob_escaped = cpp_string_escape(blob_bytes)
     offsets_str = ", ".join(str(o) for o in offsets)
 
+    # The blob's last byte is always '\0' (null terminator of last string).
+    # Point there for index 0 ("not set") and out-of-range to return "".
+    empty_offset = len(blob_bytes) - 1
+
     return (
         f"static const char {blob_var}[] = {blob_escaped};\n"
         f"static const {offset_type} {offsets_var}[] PROGMEM = {{{offsets_str}}};\n"
         f"const char *{lookup_fn}(uint16_t index) {{\n"
-        f"  if (index >= {count}) return {blob_var};\n"
-        f"  return &{blob_var}[{read_fn}(&{offsets_var}[index])];\n"
+        f"  if (index == 0 || index > {count}) return &{blob_var}[{empty_offset}];\n"
+        f"  return &{blob_var}[{read_fn}(&{offsets_var}[index - 1])];\n"
         f"}}\n"
     )