From 5c9e4685ab37e43c384786245a677d6c08d76bfe Mon Sep 17 00:00:00 2001
From: Laouen <laouen.belloli@gmail.com>
Date: Mon, 6 Jan 2025 19:34:21 -0300
Subject: [PATCH] bugfix in collector batch to tensor when top_k > number of
 elements

---
 thoi/collectors.py |  7 ++--
 thoi/graph.py      | 79 ----------------------------------------------
 2 files changed, 5 insertions(+), 81 deletions(-)
 delete mode 100644 thoi/graph.py

diff --git a/thoi/collectors.py b/thoi/collectors.py
index 7fdac12..7fc4f6a 100644
--- a/thoi/collectors.py
+++ b/thoi/collectors.py
@@ -376,9 +376,12 @@ def batch_to_tensor(nplets_idxs: torch.Tensor,
                             metric,
                             largest)
 
+    metric_func = partial(_get_string_metric, metric=metric) if isinstance(metric, str) else metric
+    
     # If not top_k or len(nplets_measuresa) > top_k  return the original values
-    # |k x D x 4|, |k x N|
-    return (nplets_measures, nplets_idxs, None)
+    # |k x D x 4|, |k x N|, |k|
+    values = metric_func(nplets_measures).to(nplets_measures.device)
+    return (nplets_measures, nplets_idxs, values)
 
 
 def concat_batched_tensors(batched_tensors: List[Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]],
diff --git a/thoi/graph.py b/thoi/graph.py
deleted file mode 100644
index 9b2c72a..0000000
--- a/thoi/graph.py
+++ /dev/null
@@ -1,79 +0,0 @@
-from typing import List
-import networkx as nx
-import numpy as np
-from tqdm import tqdm
-import pandas as pd
-
-
-def synergestic_graph(df: pd.DataFrame, marker_cols:List[int], weight_method: str = 'unweighted'):
-    """
-    @bierf: Compute the sinergestic graph to later compute other graph theory metrics
-    
-    @param df: dataframe with the synergestic nplets. The columns are i ... N and psize. where each row is a syneregestic nplet and the column i is True if the nplet contains the node i
-    @param N: number of nodes in the graph
-
-    @return: G: graph of connections where the weight of each edge is the synergestic connection of two nodes.
-
-    if weight_method is 'unweighted' the weight of the edge is the number of synergestic nplets between the two nodes
-    if weight_method is 'weighted' the weight of the edge is the number of synergestic nplets between the two nodes weighted by the psize column as 2*(1/psize)
-    """
-
-    N = len(marker_cols)
-
-    # Create the graph of conections were the weight of the node is the number of synergestic nplets
-    print('Creating the graph of connections...')
-    G = nx.Graph()
-    pbar_i = tqdm(range(N))
-    for i in pbar_i:
-        pbar_i.set_description(f'Node: {i}')
-        G.add_node(i, variable=marker_cols[i])
-        #pbar_j = tqdm(range(i+1, N+1), leave=False)
-        for j in range(i+1, N):
-            #pbar_j.set_description(f'Node: {j}')
-
-            both_in_nplet = (df[marker_cols[i]] & df[marker_cols[j]]).astype(int).values
-
-            # Count the number of nplets in common between the two nodes    
-            if weight_method == 'unweighted':
-
-                # count n rows where cols i and j are both True
-                weight = np.sum(both_in_nplet)
-
-            elif weight_method == 'weighted':
-                
-                # count n rows where cols i and j are both True weighted by the psize column as 2*(1/psize)
-                order = (2/df['order']).values
-                weight = np.sum(both_in_nplet * order)
-
-            if weight > 0:
-                G.add_edge(i, j, weight=weight)
-
-    return G
-
-def synergestic_centrality(G):
-
-    # Compute the centrality of each node
-    print('Computing the centrality of each node...')
-    centrality = nx.eigenvector_centrality(G, weight='weight')
-
-    # Add the centrality information to the graph nodes
-    for i in G.nodes:
-        G.nodes[i]['centrality'] = centrality[i]
-
-    return G
-
-def synergestic_clustering(G):
-    clustering = nx.clustering(G, weight='weight')
-
-    for i in G.nodes:
-        G.nodes[i]['clustering'] = clustering[i]
-    
-    return G
-
-def synergestic_degree(G):
-    degree = nx.degree(G, weight='weight')
-
-    for i in G.nodes:
-        G.nodes[i]['degree'] = degree[i]
-    
-    return G
\ No newline at end of file