Fix regression in String.from_raw and improve documentation

tomlkit/api.py

@@ -170,10 +170,12 @@ def string(
 
     For more information, please check the spec: `<https://toml.io/en/v1.0.0#string>`__.
 
-    Common escaping rules will be applied for basic strings.
-    This can be controlled by explicitly setting ``escape=False``.
-    Please note that, if you disable escaping, you will have to make sure that
-    the given strings don't contain any forbidden character or sequence.
+    Dumping produces a TOML string that faithfully represents the given string.
+    This includes character escaping for basic strings.
+    This behaviour can be disabled by setting ``escape=False``.
+    In that case, the given string is used verbatim for the content of the TOML string.
+    Please note that this makes you responsible for ensuring the given string does not
+    contain any forbidden character or sequence.
     """
     type_ = _StringType.select(literal, multiline)
     return String.from_raw(raw, type_, escape)

tomlkit/items.py

@@ -2305,20 +2305,19 @@ def from_raw(
         if any(c in checked for c in invalid):
             raise InvalidStringError(value, invalid, type_.value)
 
-        escaped = type_.escaped_sequences
-        string_value = escape_string(value, escaped) if escape and escaped else value
-
-        if type_ is StringType.MLB and escape:
-            string_value = _BARE_CR.sub(r"\\r", string_value)
-
-        if type_.is_multiline() and string_value[:1] in ("\n", "\r"):
-            # A newline immediately following the opening delimiter of a
-            # multiline string is trimmed by the parser, so a value that
-            # starts with a newline would otherwise lose it on round-trip.
-            # Emit an extra leading newline (the trimmed one) to preserve it.
-            string_value = "\n" + string_value
-
-        return cls(type_, decode(value), string_value, Trivia())
+        string_value = value
+        if escape:
+            string_value = escape_string(string_value, type_.escaped_sequences)
+            if type_ is StringType.MLB and escape:
+                string_value = _BARE_CR.sub(r"\\r", string_value)
+            if type_.is_multiline() and string_value[:1] in ("\n", "\r"):
+                # A newline immediately following the opening delimiter of a
+                # multiline string is trimmed by the parser, so a value that
+                # starts with a newline would otherwise lose it on round-trip.
+                # Emit an extra leading newline (the trimmed one) to preserve it.
+                string_value = "\n" + string_value
+
+        return cls(type_, value, string_value, Trivia())
 
 
 class AoT(Item, _CustomList):  # type: ignore[type-arg]