Allow empty attribute values. Fixes #165.

Author: rushter

CHANGES.rst

@@ -5,7 +5,7 @@ selectolax Changelog
 - Improve type hints, add docstrings to type hints
 - Prevent decomposing of the root node
 - Unpin Cython version and make it Optional
-
+- Allow empty attribute values. Fixes #165.
 
 Version 0.3.30
 -------------

selectolax/lexbor.pxd

@@ -60,21 +60,21 @@ cdef extern from "lexbor/html/html.h" nogil:
 
     ctypedef struct lxb_dom_node_t:
         lxb_dom_event_target_t event_target
-    
-    
+
+
         uintptr_t              local_name
         uintptr_t              prefix
         uintptr_t              ns
-    
+
         lxb_dom_document_t     *owner_document
-    
+
         lxb_dom_node_t         *next
         lxb_dom_node_t         *prev
         lxb_dom_node_t         *parent
         lxb_dom_node_t         *first_child
         lxb_dom_node_t         *last_child
         void                   *user
-    
+
         lxb_dom_node_type_t    type
 
 
@@ -318,7 +318,6 @@ cdef extern from "lexbor/dom/dom.h" nogil:
     void lxb_dom_node_insert_child(lxb_dom_node_t *to, lxb_dom_node_t *node)
     void lxb_dom_node_insert_before(lxb_dom_node_t *to, lxb_dom_node_t *node)
     void lxb_dom_node_insert_after(lxb_dom_node_t *to, lxb_dom_node_t *node)
-
     lxb_dom_text_t * lxb_dom_document_create_text_node(lxb_dom_document_t *document, const lxb_char_t *data, size_t len)
     void lxb_dom_node_simple_walk(lxb_dom_node_t *root, lxb_dom_node_simple_walker_f walker_cb, void *ctx)
 

selectolax/lexbor.pyi

@@ -1,4 +1,4 @@
-from typing import Any, Iterator, Literal, TypeVar, NoReturn, overload
+from typing import Any, Iterator, Literal, TypeVar, NoReturn, overload, Optional
 
 DefaultT = TypeVar("DefaultT")
 
@@ -13,7 +13,7 @@ class LexborAttributes:
     def __iter__(self) -> Iterator[str]: ...
     def __len__(self) -> int: ...
     def __getitem__(self, key: str) -> str | None: ...
-    def __setitem__(self, key: str, value: str) -> None: ...
+    def __setitem__(self, key: str, value: Optional[str]) -> None: ...
     def __delitem__(self, key: str) -> None: ...
     def __contains__(self, key: str) -> bool: ...
     def __repr__(self) -> str: ...

selectolax/lexbor/attrs.pxi

@@ -1,5 +1,7 @@
 cimport cython
 
+from typing import Optional
+
 @cython.final
 cdef class LexborAttributes:
     """A dict-like object that represents attributes."""
@@ -23,16 +25,32 @@ cdef class LexborAttributes:
                 yield key.decode(_ENCODING)
             attr = attr.next
 
-    def __setitem__(self, str key, value):
-        value = str(value)
+    def __setitem__(self, str key, object value):
+        value = value
         bytes_key = key.encode(_ENCODING)
-        bytes_value = value.encode(_ENCODING)
-
-        lxb_dom_element_set_attribute(
-            <lxb_dom_element_t *> self.node,
-            <lxb_char_t *> bytes_key, len(bytes_key),
-            <lxb_char_t *> bytes_value, len(bytes_value),
-        )
+        bytes_value = value.encode(_ENCODING) if value else b""
+        cdef lxb_dom_attr_t *attr
+        cdef lxb_dom_document_t *doc
+
+        if value is None:
+            # N.B. This is suboptimal, but there is not API to set empty attributes
+            attr = lxb_dom_element_set_attribute(
+                <lxb_dom_element_t *> self.node,
+                <lxb_char_t *> bytes_key, len(bytes_key),
+                NULL, 0
+            )
+            doc = (<lxb_dom_node_t*>attr).owner_document
+            lexbor_str_destroy(attr.value, doc.text, 0)
+            attr.value = NULL
+
+        elif isinstance(value, str) or isinstance(value, unicode) :
+            lxb_dom_element_set_attribute(
+                <lxb_dom_element_t *> self.node,
+                <lxb_char_t *> bytes_key, len(bytes_key),
+                <lxb_char_t *> bytes_value, len(bytes_value),
+            )
+        else:
+            raise TypeError("Expected str or unicode, got %s" % type(value))
 
     def __delitem__(self, key):
         try:

tests/test_parser.py

@@ -4,7 +4,7 @@
 import pytest
 from selectolax.parser import HTMLParser, Node
 
-from selectolax.lexbor import LexborHTMLParser, LexborNode, SelectolaxError
+from selectolax.lexbor import LexborHTMLParser, LexborNode, SelectolaxError, create_tag
 
 """
 We'are testing only our own code.
@@ -303,3 +303,8 @@ def test_decompose_root_node():
     html_parser = LexborHTMLParser("<div><p>test</p></div>")
     with pytest.raises(SelectolaxError):
         html_parser.root.decompose()
+
+def test_empty_attribute_lexbor():
+    div = create_tag("div")
+    div.attrs["hidden"] = None
+    assert div.html == "<div hidden></div>"