From cb32001998c8097dd04b97e11f2ef7ab6dc4e0e4 Mon Sep 17 00:00:00 2001
From: Jarl Holta <jarlholta@gmail.com>
Date: Sat, 28 Dec 2024 05:15:18 +0100
Subject: [PATCH] Tweak docgen - support nested comments

---
 DocGen/docgen.py                  | 36 +++++++++++++++++++++++++++----
 Source/simba.container_kdtree.pas |  6 +++++-
 2 files changed, 37 insertions(+), 5 deletions(-)

diff --git a/DocGen/docgen.py b/DocGen/docgen.py
index 2a7f1546a..7b8eac12a 100644
--- a/DocGen/docgen.py
+++ b/DocGen/docgen.py
@@ -9,6 +9,30 @@
 SRC_DIR = '../Source/script/imports/'
 BUILD_DIR = 'source/api/' 
 
+# between that handles nested comments, regex did not handle this!
+def between(str, start, stop):
+  result = []
+  init = 0
+  while True:
+    first = str.find(start, init)
+    if first == -1: break
+
+    first += len(start)
+    last = str.find(stop, first)
+
+    if last == -1: break
+
+    nested_start = str.find(start, first)
+    while nested_start != -1 and nested_start < last:
+      last = str.find(stop, last + 1)
+      if last == -1: break
+      nested_start = str.find(start, nested_start + len(start))
+    
+    result.append(str[first:last])
+    init = last + len(stop)
+  return result
+
+
 # remove old build files, all but index
 for file in glob.glob(f"{BUILD_DIR}*.md"):
   os.remove(file)
@@ -17,20 +41,24 @@
 comments = []
 for source in glob.glob(f"{SRC_DIR}*.pas"):
   with open(source, 'r') as f:
-    comments.extend(re.findall(r'\(\*(.*?)\*\)', f.read(), re.DOTALL))
-   
+    comments.extend(between(f.read(), '(*', '*)'))
+
+
 # write comments, changing the file when a header is found
 # header being a === line rather than ---
 currentfile = None
 for comment in comments:
   comment = comment.strip()
-  if (comment.splitlines()[1][0] == '='):
+  lines   = comment.splitlines();
+  
+  if (len(lines[1]) > 0) and (lines[1][0] == '='):
     if currentfile:
       currentfile.close()
-    currentfile = open(f"{BUILD_DIR}/{comment.splitlines()[0]}.md", 'a')
+    currentfile = open(f"{BUILD_DIR}/{lines[0]}.md", 'a')
     currentfile.write(comment)  
   else:
     currentfile.write('\n\n' + '-----' + '\n\n' + comment) 
+
 if currentfile:
   currentfile.close()
 
diff --git a/Source/simba.container_kdtree.pas b/Source/simba.container_kdtree.pas
index 9d5d0e971..a333901e7 100644
--- a/Source/simba.container_kdtree.pas
+++ b/Source/simba.container_kdtree.pas
@@ -663,8 +663,12 @@ function TKDTree.WeightedKNearestClassify(Vector: TSingleArray; K: Int32): Int32
   Implements an efficient n-dimensional spatial clustering algorithm using a KD-Tree.
   It supports label-based filtering to cluster points within specific categories.
 
+  Best case:                  O(n)
   Average Time Complexity:    O(n log n)
-  Worst Case Time Complexity: O(n log n)
+  Worst Case Time Complexity: O(n^2) in theory .. Maybe still just O(n log n) or thereabout.
+
+  Best case of O(n) can happen when we prune all branches in the first pass,
+  this requires a radii that covers all points from anywhere or luck point selection.
 
   TODO: Add a version that returns T2DIntegerArray where each version represents
         an index in KDTree.Data.