From a1210a1414e1c6474b1d34ab4ff50e66263a42d5 Mon Sep 17 00:00:00 2001
From: "C. Titus Brown" <titus@idyll.org>
Date: Sat, 2 Jun 2018 08:39:18 -0700
Subject: [PATCH] change max_n_below -> min_n_below

---
 sourmash/sbt.py   |  8 +++++---
 sourmash/sbtmh.py | 19 +++++++++++++------
 2 files changed, 18 insertions(+), 9 deletions(-)

diff --git a/sourmash/sbt.py b/sourmash/sbt.py
index 729b1e2fb1..d5eea18aba 100644
--- a/sourmash/sbt.py
+++ b/sourmash/sbt.py
@@ -533,15 +533,17 @@ def _fill_min_n_below(self):
             if isinstance(n, Leaf):
                 parent = self.parent(i)
                 if parent.pos not in self.missing_nodes:
-                    min_n_below = parent.node.metadata.get('min_n_below', 0)
+                    min_n_below = parent.node.metadata.get('min_n_below', 1)
                     min_n_below = min(len(n.data.minhash.get_mins()),
                                       min_n_below)
+                    if min_n_below == 0:
+                        min_n_below = 1
                     parent.node.metadata['min_n_below'] = min_n_below
 
                     current = parent
                     parent = self.parent(parent.pos)
                     while parent and parent.pos not in self.missing_nodes:
-                        min_n_below = parent.node.metadata.get('min_n_below', 0)
+                        min_n_below = parent.node.metadata.get('min_n_below', 1)
                         min_n_below = min(current.node.metadata['min_n_below'],
                                           min_n_below)
                         parent.node.metadata['min_n_below'] = min_n_below
@@ -699,7 +701,7 @@ def load(info, storage=None):
 
     def update(self, parent):
         parent.data.update(self.data)
-        min_n_below = min(parent.metadata.get('min_n_below', 0),
+        min_n_below = min(parent.metadata.get('min_n_below', 1),
                           self.metadata.get('min_n_below'))
         parent.metadata['min_n_below'] = min_n_below
 
diff --git a/sourmash/sbtmh.py b/sourmash/sbtmh.py
index 805834caee..205b20e1aa 100644
--- a/sourmash/sbtmh.py
+++ b/sourmash/sbtmh.py
@@ -57,6 +57,7 @@ def update(self, parent):
         min_n_below = parent.metadata.get('min_n_below', 1)
         min_n_below = min(len(self.data.minhash.get_mins()),
                           min_n_below)
+
         parent.metadata['min_n_below'] = min_n_below
 
     @property
@@ -94,10 +95,10 @@ def search_minhashes(node, sig, threshold, results=None, downsample=True):
     else:  # Node or Leaf, Nodegraph by minhash comparison
         if len(mins):
             matches = sum(1 for value in mins if node.data.get(value))
-            max_mins = node.metadata.get('min_n_below', -1)
-            if max_mins == -1:
+            min_n_below = node.metadata.get('min_n_below', -1)
+            if min_n_below == -1:
                 raise Exception('cannot do similarity search on this SBT; need to rebuild.')
-            score = float(matches) / max_mins
+            score = float(matches) / min_n_below
 
     if results is not None:
         results[node.name] = score
@@ -130,11 +131,17 @@ def search(self, node, sig, threshold, results=None):
                     raise
         else:  # internal object, not leaf.
             if len(mins):
+
+                # calculate the maximum possibility similarity score below
+                # this node, based on the number of matches at this node,
+                # divided by the smallest minhash size below this node
+                # (which should be an upper bound on the Jaccard similarity
+                # of any signature below this point)
                 matches = sum(1 for value in mins if node.data.get(value))
-                max_mins = node.metadata.get('min_n_below', -1)
-                if max_mins == -1:
+                min_n_below = node.metadata.get('min_n_below', -1)
+                if min_n_below == -1:
                     raise Exception('cannot do similarity search on this SBT; need to rebuild.')
-                score = float(matches) / max_mins
+                score = float(matches) / min_n_below
 
         if results is not None:
             results[node.name] = score