Improve text extraction

selectolax/lexbor/node.pxi

@@ -772,20 +772,25 @@ cdef class LexborNode:
             return container.text
 @cython.final
 cdef class TextContainer:
-    cdef public str text
+    cdef str _text
     cdef public str separator
     cdef public bool strip
 
     def __init__(self, str separator = '', bool strip = False):
-        self.text = ""
+        self._text = ""
         self.separator = separator
         self.strip = strip
 
     def append(self, node_text):
         if self.strip:
-            self.text += node_text.strip() + self.separator
+            self._text += node_text.strip() + self.separator
         else:
-            self.text += node_text + self.separator
+            self._text += node_text + self.separator
+    @property
+    def text(self):
+        if self.separator and self._text and self._text.endswith(self.separator):
+            self._text = self._text[:-len(self.separator)]
+        return self._text
 
 
 cdef lexbor_action_t text_callback(lxb_dom_node_t *node, void *ctx):

selectolax/modest/node.pxi

@@ -265,7 +265,9 @@ cdef class Node:
                         text = append_text(text, node_text, separator, strip)
                 node = node.next
         else:
-            return self._text_deep(self.node, separator=separator, strip=strip)
+            text = self._text_deep(self.node, separator=separator, strip=strip)
+        if separator and text and text.endswith(separator):
+            text = text[:-len(separator)]
         return text
 
     cdef inline _text_deep(self, myhtml_tree_node_t *node, separator='', strip=False):

tests/test_nodes.py

@@ -542,4 +542,4 @@ def test_merge_text_nodes(parser):
     node.merge_text_nodes()
     assert node.html == "<div><p>John</p><p>Doe</p></div>"
     text = tree.text(deep=True, separator=" ", strip=True)
-    assert text == "John Doe "
+    assert text == "John Doe"