iplot performance improvement by 4x. Yikes.

Every _repr_* method that exists is called on instantiation of a
IPython.core.display.HTML subclass. For Plotly,
_repr_png, _repr_svg, _repr_jpeg require a call to the image server
to generate the image. Combined, the 4 requests can make iplot commands
take 4 times as long as usual.

We mistakenly thought that the _repr_ commands were only called on demand,
e.g. on a call to `nbconvert`.

Unfortunately, this removes our support for `nbconvert`. An upcoming PR with
cached images should bring this support back.

Author: chriddyp

plotly/tools.py

@@ -1230,34 +1230,3 @@ def __init__(self, url, width, height):
 
         def _repr_html_(self):
             return self.embed_code
-
-        def _repr_svg_(self):
-            url = self.resource + ".svg"
-            res = requests.get(url)
-            if six.PY3:
-                cont = res.content.decode('utf-8', 'replace')
-            else:
-                cont = res.content
-            return cont
-
-        def _repr_png_(self):
-            url = self.resource + ".png"
-            res = requests.get(url)
-            cont = res.content
-            return cont
-
-        def _repr_pdf_(self):
-            url = self.resource + ".pdf"
-            res = requests.get(url)
-            cont = res.content
-            if six.PY3:
-                cont = res.content.decode('utf-8', 'replace')
-            else:
-                cont = res.content
-            return cont
-
-        def _repr_jpeg_(self):
-            url = self.resource + ".jpeg"
-            res = requests.get(url)
-            cont = res.content
-            return cont