sundry bugfixes

Author: erickpeirson

setup.py

@@ -7,7 +7,7 @@
 DESCRIPTION = ('Bibliographic network and corpus analysis for historians')
 LICENSE = 'GNU GPL 3' 
 URL = 'http://diging.github.io/tethne/'
-VERSION = '0.7.0-beta'
+VERSION = '0.6.4-beta'
 
 PACKAGES = [ 'tethne',
              'tethne.analyze',

tethne/analyze/collection.py

@@ -73,13 +73,22 @@ def algorithm(G, method, **kwargs):
             raise(ValueError("No such method in networkx."))
         else:
             for k, g in G.graphs.iteritems():
-                r = networkx.__dict__[method](g, **kwargs)
-                for elem, value in r.iteritems():
-                    try:
-                        results[elem][k] = value
-                    except KeyError:
-                        results[elem] = { k: value }
-                networkx.set_node_attributes(g, method, r)    # [#61510128]
+                try:
+                    r = networkx.__dict__[method](g, **kwargs)
+                except:
+                    r = 0.
+                # Some methods return a value for each node.
+                if type(r) is dict and len(r) == len(g.nodes()):
+                    for elem, value in r.iteritems():
+                        try:
+                            results[elem][k] = value
+                        except KeyError:
+                            results[elem] = { k: value }
+                    # Update the nodes in the graph with the results.
+                    networkx.set_node_attributes(g, method, r)    # [#61510128]
+                # Other methods return other kinds of values.
+                else:
+                    results[k] = r
     return results
 
 def delta(G, attribute):

tethne/analyze/corpus.py

@@ -15,7 +15,6 @@
 import networkx
 import numpy
 from ..networks.helpers import top_cited
-from ..classes import GraphCollection
 
 import matplotlib.pyplot as plt
 import matplotlib.patches as mpatches
@@ -93,7 +92,7 @@ def _top_features(corpus, feature, topn=20, perslice=False, axis='date'):
         top = [ counts.keys()[c] for c in cvalues.argsort()[-topn:][::-1] ]
     return top
 
-def plot_burstness(corpus, feature, k=5, topn=20, perslice=False,
+def plot_burstness(corpus, feature=None, k=5, topn=20, perslice=False,
                                 flist=None, normalize=True, fig=None, **kwargs):
     """
     Generate a figure depicting burstness profiles for ``feature``.
@@ -145,6 +144,9 @@ def plot_burstness(corpus, feature, k=5, topn=20, perslice=False,
        :align: center
 
     """
+    if feature is None:
+        raise ValueError('No feature specified.')
+        
     B = burstness(corpus, feature, k=k, topn=topn, perslice=perslice,
                                 flist=flist, normalize=normalize, **kwargs)
 
@@ -201,7 +203,7 @@ def plot_burstness(corpus, feature, k=5, topn=20, perslice=False,
 
     return fig
 
-def burstness(corpus, feature, k=5, topn=20, perslice=False,
+def burstness(corpus, feature=None, k=5, topn=20, perslice=False,
                       flist=None, normalize=True, **kwargs):
     """
     Estimate burstness profile for the ``topn`` features (or ``flist``) in 
@@ -248,6 +250,9 @@ def burstness(corpus, feature, k=5, topn=20, perslice=False,
        ([1990, 1991, 1992, 1993], [0., 0.4, 0.6, 0.])
 
     """
+    
+    if feature is None:
+        raise ValueError('No feature specified.')
 
     if flist is None:
         top = _top_features(corpus, feature, topn=topn, perslice=perslice)
@@ -267,7 +272,7 @@ def burstness(corpus, feature, k=5, topn=20, perslice=False,
                                             normalize=normalize, **kwargs)
     return B
 
-def feature_burstness(corpus, feature, findex, k=5, normalize=True, **kwargs):
+def feature_burstness(corpus, feature=None, findex=None, k=5, normalize=True, **kwargs):
     """
     Estimate burstness profile for a feature over the ``'date'`` axis.
     
@@ -287,6 +292,9 @@ def feature_burstness(corpus, feature, findex, k=5, normalize=True, **kwargs):
         Parameters for burstness automaton HMM.
     """
 
+    if feature is None:
+        raise ValueError('No feature specified.')
+    
     # Get time-intervals between occurrences.
     last = min(corpus.axes['date'].keys())-1
     dates = [last]    # Pad start.
@@ -339,7 +347,7 @@ def feature_burstness(corpus, feature, findex, k=5, normalize=True, **kwargs):
     return D, [ A_[d] for d in D ]
 
 
-def plot_sigma(G, corpus, feature, topn=20, sort_by='max', perslice=False,
+def plot_sigma(G=None, corpus=None, feature=None, topn=20, sort_by='max', perslice=False,
                                                 flist=None, fig=None, **kwargs):
     """
     Plot sigma values for the ``topn`` most influential nodes.
@@ -393,6 +401,16 @@ def plot_sigma(G, corpus, feature, topn=20, sort_by='max', perslice=False,
        :width: 600
        :align: center
     """
+    if G is None:
+        raise ValueError('No GraphCollection specified.')
+    
+    
+    if feature is None:
+        raise ValueError('No feature specified.')
+        
+    if corpus is None:
+        raise ValueError('No corpus specified.')                
+    
     G = sigma(G, corpus, feature)
     nodes = G.nodes()
 
@@ -491,7 +509,7 @@ def plot_sigma(G, corpus, feature, topn=20, sort_by='max', perslice=False,
             rect = mpatches.Rectangle(  xy, width, height, fill=True,
                                                            linewidth=0.0    )
             rect.set_facecolor(color)
-            rect.set_alpha(state + 0.1)
+            rect.set_alpha(min(state + 0.1, 1.0))
             ax.add_patch(rect)
 
         ax.set_ylabel(  G.node_index[node], rotation=0,
@@ -500,9 +518,9 @@ def plot_sigma(G, corpus, feature, topn=20, sort_by='max', perslice=False,
 
     plt.subplots_adjust(left=0.5)
     fig.tight_layout(h_pad=0.25)
-    return fig, G
+    return fig
 
-def sigma(G, corpus, feature, **kwargs):
+def sigma(G, corpus=None, feature=None, **kwargs):
     """
     Calculate sigma (from `Chen 2009 <http://arxiv.org/pdf/0904.1439.pdf>`_) for
     all of the nodes in a :class:`.GraphCollection`\.
@@ -564,6 +582,12 @@ def sigma(G, corpus, feature, **kwargs):
 
     """
 
+    if feature is None:
+        raise ValueError('No feature specified.')
+        
+    if corpus is None:
+        raise ValueError('No corpus specified.')      
+    
     nodes = G.node_lookup.keys()
 
     B = burstness(corpus, feature, flist=nodes, **kwargs)

tethne/classes/corpus.py

@@ -12,6 +12,7 @@
 import matplotlib
 from paper import Paper
 from collections import Counter
+from ..analyze import corpus as az_corpus
 from nltk.corpus import stopwords
 import scipy
 
@@ -154,6 +155,15 @@ def __init__(self, papers, features=None, index_by='ayjid',
                         index_citation_by=index_citation_by,
                         exclude=exclude, filt=filt  )
 
+    def analyze(self, method, **kwargs):
+        try:
+            method_callable = az_corpus.__dict__[method]
+        except KeyError:
+            raise ValueError('No such method.')
+        
+        return method_callable(corpus=self, **kwargs)
+        
+
     def index(  self,   papers, features=None, index_by='ayjid',
                         index_citation_by='ayjid', exclude=set([]),
                         filt=None, stem=False   ):
@@ -1204,6 +1214,8 @@ def distribution(self, x_axis, y_axis=None):
             y_size = 1
         shape = (x_size, y_size)
         logger.debug('distribution shape: {0}'.format(shape))
+        
+        # Construct the sparse matrix.
         I = []
         J = []
         K = []
@@ -1223,8 +1235,10 @@ def distribution(self, x_axis, y_axis=None):
                         K.append(k)
 
         # TODO: Move away from SciPy, to facilitate PyPy compatibility?
-        dist = np.array(scipy.sparse.coo_matrix((K, (I,J)), shape=shape).todense())
-
+        csr = scipy.sparse.coo_matrix((K, (I,J)), shape=shape).tocsr()
+        nonzero = csr.nonzero()
+        dist = np.array(csr.todense())
+        
         return dist
 
     # TODO: Merge this with :func:`.distribution`
@@ -1320,7 +1334,12 @@ def feature_distribution(self, featureset, feature, x_axis, y_axis=None,
         fvalues = self.features[featureset]['features']
 
         def _get_value(papers):
-            vtuples = [ fv for p in papers for fv in fvalues[p] ]
+            vtuples = []
+            for p in papers:
+                if p in fvalues:
+                    vtuples += fvalues[p]
+
+#            vtuples = [ fv for p in papers for fv in fvalues[p] ]
             values = [ v for f,v in vtuples if f == findex ]
 
             if mode == 'counts':

tethne/classes/graphcollection.py

@@ -11,6 +11,7 @@
 import networkx
 import numpy as np
 import matplotlib.pyplot as plt
+from ..analyze.collection import algorithm
 
 import warnings
 
@@ -121,7 +122,57 @@ def __delitem__(self, key):
 
     def __len__(self):
         return len(self.graphs)
-    
+        
+    def analyze(self, method, **kwargs):
+        """
+        Apply a ``method`` from NetworkX to all ``networkx.Graph`` objects in the
+        :class:`.GraphCollection` ``G``.
+        
+        For options, see the `list of algorithms
+        <http://networkx.github.io/documentation/networkx-1.9/reference/algorithms.html>`_
+        in the NetworkX documentation. Not all of these have been tested.
+
+        Parameters
+        ----------
+        G : :class:`.GraphCollection`
+            The :class:`.GraphCollection` to analyze. The specified method will be
+            applied to each graph in ``G``.
+        method : string
+            Name of a method in NetworkX to execute on graph collection.
+        **kwargs
+            A list of keyword arguments that should correspond to the parameters
+            of the specified method.
+
+        Returns
+        -------
+        results : dict
+            Indexed by element (node or edge) and graph index (e.g. ``date``).
+
+        Raises
+        ------
+        ValueError
+            If no such method exists.
+
+        Examples
+        --------
+
+        *Betweenness centrality:* (``G`` is a :class:`.GraphCollection`\)
+
+        .. code-block:: python
+
+           >>> from tethne.analyze import collection
+           >>> BC = collection.algorithm(G, 'betweenness_centrality')
+           >>> print BC[0]
+           {1999: 0.010101651117889644,
+           2000: 0.0008689093723107329,
+           2001: 0.010504898852426189,
+           2002: 0.009338654511194512,
+           2003: 0.007519105636349891}
+
+        """    
+        results = algorithm(self, method, **kwargs)
+        return results
+        
     def build(self, corpus, axis, node_type, graph_type, method_kwargs={},
                                                                       **kwargs):
         """
@@ -471,7 +522,10 @@ def attr_distribution(self, attr='weight', etype='edge', stat=np.mean):
             # Ignore warnings; will handle NaNs below.
             with warnings.catch_warnings():
                 warnings.simplefilter('ignore')
-                v = stat(A)
+                try:
+                    v = stat(A)
+                except ValueError:  # Raised by max with empty sequence.
+                    v = 0.
 
             if np.isnan(v):
                 v = 0.

tethne/networks/authors.py

@@ -83,7 +83,7 @@ def author_papers(papers, node_id='ayjid', paper_attribs=[], **kwargs):
 
     return author_papers_graph
 
-def institutions(papers, threshold=1, edge_attrbs=['ayjid'], 
+def institutions(papers, threshold=1, edge_attrbs=[], 
                  node_attribs=['authors'], geocode=False, **kwargs):
     """
     Generates an institutional network based on coauthorship.