U tdx@sNddlZddlmZmZGdddeZGdddeZd ddZd d ZdS) N)EdgeSeq VertexSeqc@s0eZdZdZddZddZddZdd Zd S) raClass representing a sequence of vertices in the graph. This class is most easily accessed by the C{vs} field of the L{Graph} object, which returns an ordered sequence of all vertices in the graph. The vertex sequence can be refined by invoking the L{VertexSeq.select()} method. L{VertexSeq.select()} can also be accessed by simply calling the L{VertexSeq} object. An alternative way to create a vertex sequence referring to a given graph is to use the constructor directly: >>> g = Graph.Full(3) >>> vs = VertexSeq(g) >>> restricted_vs = VertexSeq(g, [0, 1]) The individual vertices can be accessed by indexing the vertex sequence object. It can be used as an iterable as well, or even in a list comprehension: >>> g=Graph.Full(3) >>> for v in g.vs: ... v["value"] = v.index ** 2 ... >>> [v["value"] ** 0.5 for v in g.vs] [0.0, 1.0, 2.0] The vertex set can also be used as a dictionary where the keys are the attribute names. The values corresponding to the keys are the values of the given attribute for every vertex selected by the sequence. >>> g=Graph.Full(3) >>> for idx, v in enumerate(g.vs): ... v["weight"] = idx*(idx+1) ... >>> g.vs["weight"] [0, 2, 6] >>> g.vs.select(1,2)["weight"] = [10, 20] >>> g.vs["weight"] [0, 10, 20] If you specify a sequence that is shorter than the number of vertices in the VertexSeq, the sequence is reused: >>> g = Graph.Tree(7, 2) >>> g.vs["color"] = ["red", "green"] >>> g.vs["color"] ['red', 'green', 'red', 'green', 'red', 'green', 'red'] You can even pass a single string or integer, it will be considered as a sequence of length 1: >>> g.vs["color"] = "red" >>> g.vs["color"] ['red', 'red', 'red', 'red', 'red', 'red', 'red'] Some methods of the vertex sequences are simply proxy methods to the corresponding methods in the L{Graph} object. One such example is C{VertexSeq.degree()}: >>> g=Graph.Tree(7, 2) >>> g.vs.degree() [2, 3, 3, 1, 1, 1, 1] >>> g.vs.degree() == g.degree() True cCs |jS)zfReturns the list of all the vertex attributes in the graph associated to this vertex sequence.)graphZvertex_attributesselfrC/home/sam/Atlas/atlas_env/lib/python3.8/site-packages/igraph/seq.py attributesLszVertexSeq.attributescOs|sRd|kr|d}nd|kr,|d}nd}|dk rRt|trJ|g}n||d<|rtj|f|}|sn|S|jj|j}n|}|jf|}|r|dSt ddS)aReturns the first vertex of the vertex sequence that matches some criteria. The selection criteria are equal to the ones allowed by L{VertexSeq.select}. See L{VertexSeq.select} for more details. For instance, to find the first vertex with name C{foo} in graph C{g}: >>> g.vs.find(name="foo") #doctest:+SKIP To find an arbitrary isolated vertex: >>> g.vs.find(_degree=0) #doctest:+SKIP nameZname_eqNrzno such vertex) pop isinstancestr _VertexSeqfindrvsselectindex ValueError)rargskwdsr ZvertexrrrrrQs(    zVertexSeq.findc stj|f|}tjtjtjtjtjtjddddd}| D]\}d|ks`| ddkrh|d7}| d\}}}z ||Wn(t k r|d|d}}YnX|ddkrt |j|d d |} n||} fd d t| D} || }qB|S) aSelects a subset of the vertex sequence based on some criteria The selection criteria can be specified by the positional and the keyword arguments. Positional arguments are always processed before keyword arguments. - If the first positional argument is C{None}, an empty sequence is returned. - If the first positional argument is a callable object, the object will be called for every vertex in the sequence. If it returns C{True}, the vertex will be included, otherwise it will be excluded. - If the first positional argument is an iterable, it must return integers and they will be considered as indices of the current vertex set (NOT the whole vertex set of the graph -- the difference matters when one filters a vertex set that has already been filtered by a previous invocation of L{VertexSeq.select()}. In this case, the indices do not refer directly to the vertices of the graph but to the elements of the filtered vertex sequence. - If the first positional argument is an integer, all remaining arguments are expected to be integers. They are considered as indices of the current vertex set again. Keyword arguments can be used to filter the vertices based on their attributes. The name of the keyword specifies the name of the attribute and the filtering operator, they should be concatenated by an underscore (C{_}) character. Attribute names can also contain underscores, but operator names don't, so the operator is always the largest trailing substring of the keyword name that does not contain an underscore. Possible operators are: - C{eq}: equal to - C{ne}: not equal to - C{lt}: less than - C{gt}: greater than - C{le}: less than or equal to - C{ge}: greater than or equal to - C{in}: checks if the value of an attribute is in a given list - C{notin}: checks if the value of an attribute is not in a given list For instance, if you want to filter vertices with a numeric C{age} property larger than 200, you have to write: >>> g.vs.select(age_gt=200) #doctest: +SKIP Similarly, to filter vertices whose C{type} is in a list of predefined types: >>> list_of_types = ["HR", "Finance", "Management"] >>> g.vs.select(type_in=list_of_types) #doctest: +SKIP If the operator is omitted, it defaults to C{eq}. For instance, the following selector selects vertices whose C{cluster} property equals to 2: >>> g.vs.select(cluster=2) #doctest: +SKIP In the case of an unknown operator, it is assumed that the recognized operator is part of the attribute name and the actual operator is C{eq}. Attribute names inferred from keyword arguments are treated specially if they start with an underscore (C{_}). These are not real attributes but refer to specific properties of the vertices, e.g., its degree. The rule is as follows: if an attribute name starts with an underscore, the rest of the name is interpreted as a method of the L{Graph} object. This method is called with the vertex sequence as its first argument (all others left at default values) and vertices are filtered according to the value returned by the method. For instance, if you want to exclude isolated vertices: >>> g = Graph.Famous("zachary") >>> non_isolated = g.vs.select(_degree_gt=0) For properties that take a long time to be computed (e.g., betweenness centrality for large graphs), it is advised to calculate the values in advance and store it in a graph attribute. The same applies when you are selecting based on the same property more than once in the same C{select()} call to avoid calculating it twice unnecessarily. For instance, the following would calculate betweenness centralities twice: >>> edges = g.vs.select(_betweenness_gt=10, _betweenness_lt=30) It is advised to use this instead: >>> g.vs["bs"] = g.betweenness() >>> edges = g.vs.select(bs_gt=10, bs_lt=30) @return: the new, filtered vertex sequencecSs||kSNrabrrrz"VertexSeq.select..cSs||kSrrrrrrrrltgtlegeeqneinnotin_r_eqr!Ncsg|]\}}|r|qSrr.0ivfuncvaluerr s z$VertexSeq.select..)rroperatorrrrr r!r"itemsrindex rpartitionKeyErrorgetattrr enumerate) rrrr operatorskeywordattrr%opvalues filtered_idxsrr,rrs0g    zVertexSeq.selectcOs |j||S)zvShorthand notation to select() This method simply passes all its arguments to L{VertexSeq.select()}. rrrrrrr__call__ szVertexSeq.__call__N__name__ __module__ __qualname____doc__r rrr?rrrrr s B3rc@s0eZdZdZddZddZddZdd Zd S) raClass representing a sequence of edges in the graph. This class is most easily accessed by the C{es} field of the L{Graph} object, which returns an ordered sequence of all edges in the graph. The edge sequence can be refined by invoking the L{EdgeSeq.select()} method. L{EdgeSeq.select()} can also be accessed by simply calling the L{EdgeSeq} object. An alternative way to create an edge sequence referring to a given graph is to use the constructor directly: >>> g = Graph.Full(3) >>> es = EdgeSeq(g) >>> restricted_es = EdgeSeq(g, [0, 1]) The individual edges can be accessed by indexing the edge sequence object. It can be used as an iterable as well, or even in a list comprehension: >>> g=Graph.Full(3) >>> for e in g.es: ... print(e.tuple) ... (0, 1) (0, 2) (1, 2) >>> [max(e.tuple) for e in g.es] [1, 2, 2] The edge sequence can also be used as a dictionary where the keys are the attribute names. The values corresponding to the keys are the values of the given attribute of every edge in the graph: >>> g=Graph.Full(3) >>> for idx, e in enumerate(g.es): ... e["weight"] = idx*(idx+1) ... >>> g.es["weight"] [0, 2, 6] >>> g.es["weight"] = range(3) >>> g.es["weight"] [0, 1, 2] If you specify a sequence that is shorter than the number of edges in the EdgeSeq, the sequence is reused: >>> g = Graph.Tree(7, 2) >>> g.es["color"] = ["red", "green"] >>> g.es["color"] ['red', 'green', 'red', 'green', 'red', 'green'] You can even pass a single string or integer, it will be considered as a sequence of length 1: >>> g.es["color"] = "red" >>> g.es["color"] ['red', 'red', 'red', 'red', 'red', 'red'] Some methods of the edge sequences are simply proxy methods to the corresponding methods in the L{Graph} object. One such example is C{EdgeSeq.is_multiple()}: >>> g=Graph(3, [(0,1), (1,0), (1,2)]) >>> g.es.is_multiple() [False, True, False] >>> g.es.is_multiple() == g.is_multiple() True cCs |jS)zbReturns the list of all the edge attributes in the graph associated to this edge sequence.)rZedge_attributesrrrrr YszEdgeSeq.attributescOsV|r.tj|f|}|s|S|jj|j}n|}|jf|}|rJ|dStddS)axReturns the first edge of the edge sequence that matches some criteria. The selection criteria are equal to the ones allowed by L{VertexSeq.select}. See L{VertexSeq.select} for more details. For instance, to find the first edge with weight larger than 5 in graph C{g}: >>> g.es.find(weight_gt=5) #doctest:+SKIP rz no such edgeN)_EdgeSeqrresrrr)rrredgerFrrrr^s  z EdgeSeq.findc stj|f||j}dd}tjtjtjtjtj tj ddddd}| D]\}d|kst| ddkr||d 7}| d}|d|||d d }} z || Wn(t k r|d |d }} YnX|ddkr|d krH|sH| dkr tdd}|d krHtdr>tts>tn tg|dkrr| d krtjjdd} d n*ddD} | dks| dkr|q|dkrr| d krtjjdd} d n*ddD} | dks| dkr|q|dkr|d |tD]} j| q4spfddtD} nt} q|dkrd |tD]} j| qsڇfddtD} nfddD} n|dkrtd krtd!d |d|d tD]} j| q6D]} j| qTsfd"dtD} nfd#dD} ntj|d d } n|} d k rfd$dt| D} | qTS)%a;Selects a subset of the edge sequence based on some criteria The selection criteria can be specified by the positional and the keyword arguments. Positional arguments are always processed before keyword arguments. - If the first positional argument is C{None}, an empty sequence is returned. - If the first positional argument is a callable object, the object will be called for every edge in the sequence. If it returns C{True}, the edge will be included, otherwise it will be excluded. - If the first positional argument is an iterable, it must return integers and they will be considered as indices of the current edge set (NOT the whole edge set of the graph -- the difference matters when one filters an edge set that has already been filtered by a previous invocation of L{EdgeSeq.select()}. In this case, the indices do not refer directly to the edges of the graph but to the elements of the filtered edge sequence. - If the first positional argument is an integer, all remaining arguments are expected to be integers. They are considered as indices of the current edge set again. Keyword arguments can be used to filter the edges based on their attributes and properties. The name of the keyword specifies the name of the attribute and the filtering operator, they should be concatenated by an underscore (C{_}) character. Attribute names can also contain underscores, but operator names don't, so the operator is always the largest trailing substring of the keyword name that does not contain an underscore. Possible operators are: - C{eq}: equal to - C{ne}: not equal to - C{lt}: less than - C{gt}: greater than - C{le}: less than or equal to - C{ge}: greater than or equal to - C{in}: checks if the value of an attribute is in a given list - C{notin}: checks if the value of an attribute is not in a given list For instance, if you want to filter edges with a numeric C{weight} property larger than 50, you have to write: >>> g.es.select(weight_gt=50) #doctest: +SKIP Similarly, to filter edges whose C{type} is in a list of predefined types: >>> list_of_types = ["inhibitory", "excitatory"] >>> g.es.select(type_in=list_of_types) #doctest: +SKIP If the operator is omitted, it defaults to C{eq}. For instance, the following selector selects edges whose C{type} property is C{intracluster}: >>> g.es.select(type="intracluster") #doctest: +SKIP In the case of an unknown operator, it is assumed that the recognized operator is part of the attribute name and the actual operator is C{eq}. Keyword arguments are treated specially if they start with an underscore (C{_}). These are not real attributes but refer to specific properties of the edges, e.g., their centrality. The rules are as follows: 1. C{_source} or {_from} means the source vertex of an edge. For undirected graphs, only the C{eq} operator is supported and it is treated as {_incident} (since undirected graphs have no notion of edge directionality). 2. C{_target} or {_to} means the target vertex of an edge. For undirected graphs, only the C{eq} operator is supported and it is treated as {_incident} (since undirected graphs have no notion of edge directionality). 3. C{_within} ignores the operator and checks whether both endpoints of the edge lie within a specified set. 4. C{_between} ignores the operator and checks whether I{one} endpoint of the edge lies within a specified set and the I{other} endpoint lies within another specified set. The two sets must be given as a tuple. 5. C{_incident} ignores the operator and checks whether the edge is incident on a specific vertex or a set of vertices. 6. Otherwise, the rest of the name is interpreted as a method of the L{Graph} object. This method is called with the edge sequence as its first argument (all others left at default values) and edges are filtered according to the value returned by the method. For instance, if you want to exclude edges with a betweenness centrality less than 2: >>> g = Graph.Famous("zachary") >>> excl = g.es.select(_edge_betweenness_ge = 2) To select edges originating from vertices 2 and 4: >>> edges = g.es.select(_source_in = [2, 4]) To select edges lying entirely within the subgraph spanned by vertices 2, 3, 4 and 7: >>> edges = g.es.select(_within = [2, 3, 4, 7]) To select edges with one endpoint in the vertex set containing vertices 2, 3, 4 and 7 and the other endpoint in the vertex set containing vertices 8 and 9: >>> edges = g.es.select(_between = ([2, 3, 4, 7], [8, 9])) For properties that take a long time to be computed (e.g., betweenness centrality for large graphs), it is advised to calculate the values in advance and store it in a graph attribute. The same applies when you are selecting based on the same property more than once in the same C{select()} call to avoid calculating it twice unnecessarily. For instance, the following would calculate betweenness centralities twice: >>> edges = g.es.select(_edge_betweenness_gt=10, # doctest:+SKIP ... _edge_betweenness_lt=30) It is advised to use this instead: >>> g.es["bs"] = g.edge_betweenness() >>> edges = g.es.select(bs_gt=10, bs_lt=30) @return: the new, filtered edge sequence cSs8t|trtdd|D}nt|ttfs4t|}|S)Ncss|] }|jVqdSrr)r)r+rrr sz6EdgeSeq.select.._ensure_set..)r rset frozenset)r.rrr _ensure_set s  z#EdgeSeq.select.._ensure_setcSs||kSrrrrrrrrz EdgeSeq.select..cSs||kSrrrrrrrrrr%rr&r'Nr!)_source_from_target_to)r!r#z!unsupported for undirected graphsZ _incident__iter__)rMrNout)modecSsg|] }|jqSr)sourcer)errrr/Dsz"EdgeSeq.select..r#r$)rOrPcSsg|] }|jqSr)targetrUrrrr/Pscsg|]\}}|jkr|qSrrHr)r*rV) candidatesrrr/`s Z_withincs2g|]*\}}|jkr|jkr|jkr|qSr)rrTrWrX)rYr.rrr/ss    cs,g|]$}|jkr|jkr|qSrrTrWr)r*)rFr.rrr/}sZ_betweenz._between selector requires two vertex ID listscs<g|]4\}}|jkr |jks4|jkr|jkr|qSrrZrX)set1set2rrr/s   csHg|]@}|jkr$|jks@|jkr|jkr|qSrrZr[)rFr]r^rrr/s csg|]\}}|r|qSrrr(r,rrr/s )rErr is_directedr0rrrr r!r"r1r2r4 RuntimeErrorhasattrr r rJZis_allsortedZincidentupdater6lenrr5) rrrr_rLr7r8posr9r:r<r;r+r)rYrFr-r]r^r.rrys                           zEdgeSeq.selectcOs |j||S)ztShorthand notation to select() This method simply passes all its arguments to L{EdgeSeq.select()}. r=r>rrrr?szEdgeSeq.__call__Nr@rrrrrsE>rcsbddlm}|dkrj}t||tdr>fdd}n fdd}||_dd |i|_|S) aAuxiliary decorator This decorator allows some methods of L{VertexSeq} and L{EdgeSeq} to call their respective counterparts in L{Graph} to avoid code duplication. @param func: the function being decorated. This function will be called on the results of the original L{Graph} method. If C{None}, defaults to the identity function. @param name: the name of the corresponding method in L{Graph}. If C{None}, it defaults to the name of the decorated function. @return: the decorated function r)GraphNr?cs$|dj}|d|f||SNrrrrrr-methodrr decorateds z_graphmethod..decoratedcs|dj}|f||Srgrhri)rkrrrls zProxy method to L{Graph.%(name)s()} This method calls the C{%(name)s()} method of the L{Graph} class restricted to this sequence, and returns the result. @see: Graph.%(name)s() for details. r )ZigraphrfrAr5rarD)r-r rfrlrrjr _graphmethods     rmcCsi}ddddddddd d d d d ddddddddddg|t<ddddddg|t<i}ddi|t<ddd|t<|D]4\}}|D]&}||||}t||td|qqtttd td!d"d dS)#NZdegreeZ betweennessZ bibcouplingZ closenessZ cocitation constraintZ distancesZ diversityZ eccentricityZget_shortest_pathsZ maxdegreeZpagerankZpersonalized_pagerankZshortest_pathsZsimilarity_diceZsimilarity_jaccardZsubgraphZindegreeZ outdegreeZisoclassZdelete_verticesZ is_separatorZis_minimal_separatorZcount_multiple delete_edgesZis_loopZ is_multipleZ is_mutualsubgraph_edgesdelete)rorpZedge_betweennesscsfdd|jDS)Ncsg|] }|qSrrr[resultrrr/"sz8_add_proxy_methods....)indices)rrsrrrrr"rz$_add_proxy_methods..)rrr1getsetattrrm)Zdecorated_methodsZrename_methodsclsmethodsrkZnew_method_namerrr_add_proxy_methodss^  ry)NN)r0Zigraph._igraphrrErrrmryrrrrs - .