diff --git a/.gitignore b/.gitignore
index 07f665c6..8161060c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -42,3 +42,6 @@ output/*/index.html
 
 # Sphinx
 docs/_build
+
+# Pytest
+/.cache
diff --git a/parsel/xpathfuncs.py b/parsel/xpathfuncs.py
index 777bd32f..3c664eb7 100644
--- a/parsel/xpathfuncs.py
+++ b/parsel/xpathfuncs.py
@@ -26,6 +26,7 @@ def set_xpathfunc(fname, func):
 
 def setup():
     set_xpathfunc('has-class', has_class)
+    set_xpathfunc('rel-id', rel_id)
 
 
 def has_class(context, *classes):
@@ -52,3 +53,54 @@ def has_class(context, *classes):
         if ' ' + cls + ' ' not in node_cls:
             return False
     return True
+
+
+_id_xpath = etree.XPath('id($node_id)')
+
+
+def rel_id(context, node_id, nodeset=None):
+    """Relative lookup by ID (rel-id function).
+
+    Same as ``id`` function, but relative to some nodeset (current node by
+    default).
+
+    For example, the following XPath expressions will return the same result
+    (however, with different performance)::
+
+        document.xpath("id('foo')")        # fastest
+        document.xpath("rel-id('foo')")    # fast
+        document.xpath("//*[@id='foo']")   # slow, has to iterate
+
+    This function is useful in relative lookups, for example::
+
+        document.xpath("rel-id('bar', id('foo'))")  # fast
+        document.xpath("id('foo')//*[@id='bar']")   # slow, has to iterate
+
+    The above can also be done with::
+
+        document.xpath("id('foo')").xpath("rel-id('bar')")  # fast
+
+    which showcases the fact that the current node is the default nodeset.
+
+    """
+    if not context.eval_context.get('args_checked'):
+        if not isinstance(node_id, string_types):
+            raise ValueError(
+                'XPath error: rel-id: first argument must be a string')
+        if nodeset is not None and not isinstance(nodeset, list):
+            raise ValueError(
+                'XPath error: rel-id: second argument must be a nodeset')
+        context.eval_context['args_checked'] = True
+    if nodeset is None:
+        nodeset = {context.context_node}
+    else:
+        nodeset = set(nodeset)
+
+    result = _id_xpath(context.context_node, node_id=node_id)
+    should_return_result = (
+        not result or
+        nodeset.intersection(result) or
+        nodeset.intersection(result[0].iterancestors()))
+    if should_return_result:
+        return result
+    return []
diff --git a/tests/test_xpathfuncs.py b/tests/test_xpathfuncs.py
index 29b1e5f2..ac497749 100644
--- a/tests/test_xpathfuncs.py
+++ b/tests/test_xpathfuncs.py
@@ -95,3 +95,79 @@ def myfunc(ctx):
         self.assertRaisesRegexp(
             ValueError, 'Unregistered function in myfunc',
             sel.xpath, 'myfunc()')
+
+    def test_rel_id_basic(self):
+        body = u"""
+        <foo><p id="foop">Foo</p></foo>
+        <bar><p id="barp">Bar</p></p>
+        """
+        sel = Selector(text=body)
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("foop")/text()')],
+            [u'Foo'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("foop", .)/text()')],
+            [u'Foo'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("foop", //foo)/text()')],
+            [u'Foo'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("foop", //p)/text()')],
+            [u'Foo'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("foop", //bar)/text()')],
+            [],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('//foo').xpath('rel-id("foop")/text()')],
+            [u'Foo'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('//bar').xpath('rel-id("foop")/text()')],
+            [],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("barp", //bar)/text()')],
+            [u'Bar'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('rel-id("foop", //zzz)/text()')],
+            [],
+        )
+
+    def test_rel_id_in_conditional(self):
+        body = u"""
+        <p><p id="foop">Foo</p></foo>
+        <p><p id="barp">Bar</p></p>
+        """
+        sel = Selector(text=body)
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('//p[rel-id("foop")]//text()')],
+            [u'Foo'],
+        )
+        self.assertEqual(
+            [x.extract() for x in sel.xpath('//p[rel-id("barp")]//text()')],
+            [u'Bar'],
+        )
+
+    def test_rel_id_error_invalid_id(self):
+        body = u"""
+        <p CLASS="foo">First</p>
+        """
+        sel = Selector(text=body)
+        self.assertRaisesRegexp(
+            ValueError, 'rel-id: first argument must be a string',
+            sel.xpath, u'rel-id(123)')
+
+    def test_rel_id_error_invalid_nodeset(self):
+        body = u"""
+        <p CLASS="foo">First</p>
+        """
+        sel = Selector(text=body)
+        self.assertRaisesRegexp(
+            ValueError, 'rel-id: second argument must be a nodeset',
+            sel.xpath, u'rel-id("123", true())')