Allow Search with Pip

[psaavedra]

This patch hacks the code in order to allow pip search PKG. For example:

pip search  --index http://pypi.intenet.com:8080/search/ MyDjango

I wonder if you are interested on this and if this patch is wellcome for your side.

--- /root/_app.py   2015-03-12 11:16:23.000000000 +0000
+++ /usr/local/lib/python2.7/dist-packages/pypiserver/_app.py   2015-03-12 12:26:29.000000000 +0000
@@ -6,6 +6,7 @@
 import mimetypes
 import logging
 import pkg_resources
+import xml.dom.minidom
 
 try:
     from io import BytesIO
@@ -24,6 +25,48 @@
 log = logging.getLogger('pypiserver.http')
 packages = None
 
+XMLRPC_RESPONSE_SKELETON='''
+
+
+
+
+%(objects)s
+
+
+
+'''
+
+def from_obj_to_xmlrpc(obj_):
+    res = ""
+    if isinstance(obj_,int):
+        res += "%s" % obj_
+        return res
+    elif isinstance(obj_,str):
+        res += "%s" % obj_
+        return res
+    elif isinstance(obj_,dict):
+        res += ""
+        for k,v in obj_.iteritems():
+            member = {}
+            member["name"]=k
+            member["value"]=from_obj_to_xmlrpc(v)
+            res_1 = ""
+            res_1 += "%(name)s"
+            res_1 += "%(value)s"
+            res_1 += ""
+            res += res_1 % member
+        res += ""
+        return res
+    elif isinstance(obj_,list):
+        res += ""
+        for i in obj_:
+            res += "%s" % from_obj_to_xmlrpc(i)
+        res += ""
+        return res
+    else:
+       raise Exception("No valid object")
+
+
 
 class Configuration(object):
     def __init__(self):
@@ -267,8 +310,31 @@
     return redirect(request.fullpath + "/")
 
 
-@app.route("/simple/")
+@app.post('/search')
+@app.post('/search/')
 @auth("list")
+def search():
+    value = ""
+    try:
+        parser = xml.dom.minidom.parse(request.body)
+        member = parser.getElementsByTagName("member")[0]
+        value  = parser.getElementsByTagName("string")[0].childNodes[0].wholeText.strip()
+    except Exception, e:
+        value = ""
+
+    response = []
+    ordering = 0
+    for p in packages():
+        if p.pkgname.count(value) > 0:
+            d = {'_pypi_ordering': ordering, 'version': p.version,
+                 'name': p.pkgname, 'summary': p.fn}
+            response.append(d)
+        ordering += 1
+    res = XMLRPC_RESPONSE_SKELETON \
+        % {"objects":from_obj_to_xmlrpc(response)}
+    return res
+
+@app.route("/simple/")
 def simpleindex():
     links = sorted(get_prefixes(packages()))
     tmpl = """\

[ankostis]

Of course we are interested for new features, as long as they do not require http-requests upstream or a db.

Would mind pointing at the points of your patch that you consider "hack"-ish?

[psaavedra]

Yes of course,

• search operation is a XMLRPC request in fact, I just implemented the search request ad-hoc (see XMLRPC_RESPONSE_SKELETON), there is other operations which are in use by pip that they are not implemented. Maybe the best way to do it is using the xmlrpc.server (https://docs.python.org/3/library/xmlrpc.server.html) but this force to use a additional port.
• from_obj_to_xmlrpc is a made by myself translator from built-in data to XML-RPC data model.
• Package summary is replaced by the package filename since the moment I don't know how to recover this info with Anyway, if you just want using the returned object of the function packages()

Anyway, this patch is fullysafe and doesn't modifies any built logic of your code. You can play with it and with the search commnd in our recent installation of PyPIserver : http://pypi.interoud.net/:

pip search --index http://pypi.interoud.com:8080/search/ live

Thanks a lot for your PyPI clone. ;-)

[ankostis]

Would need a test-case here.

[ankostis]

Dear @psaavedra would you mind making a regular pull-request?

[janssen]

I'd kind of like to see this, too.

[janssen]

Why in the world does pypi drag in XML-RPC for this simple thing? You could just send a request the same way (using forms), and send back the response as a pickled Python dict, or as a JSON dict.

[psaavedra]

Pull request submitted: #115

[mplanchard]

Handled by #114, which has been merged.