[embedhtml] only import from lxml if we're actually about to use it

This allows nbconvert_support to import the module, even if lxml is missing
or has missing dependency libs, etc.
In this way, problems with lxml should only bother people who are
actually tryying to use this.

Author: jcb91

src/jupyter_contrib_nbextensions/nbconvert_support/embedhtml.py

@@ -3,7 +3,6 @@
 import base64
 import os
 
-import lxml.etree as et
 from ipython_genutils.ipstruct import Struct
 from nbconvert.exporters.html import HTMLExporter
 
@@ -66,6 +65,14 @@ def replfunc(self, node):
         node.attrib["src"] = prefix + b64_data
 
     def from_notebook_node(self, nb, resources=None, **kw):
+        # The parent nbconvert_support module imports this module, and
+        # nbconvert_support is imported as part of our install scripts, and
+        # other fairly basic stuff.
+        # By keeping lxml import in this method, we can still import this
+        # module even if lxml isn't available, or is missing dependencies, etc.
+        # In this way, problems with lxml should only bother people who are
+        # actually trying to *use* this.
+        import lxml.etree as et
         output, resources = super(
             EmbedHTMLExporter, self).from_notebook_node(nb, resources)