From dd7cd28097a1794e7e6f23132f5dbafe9c356f0b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?G=C3=BCrer=20=C3=96zen?= <gurer@pardus.org.tr>
Date: Fri, 2 Mar 2007 08:41:42 +0000
Subject: [PATCH] =?UTF-8?q?sha1=20hesaplamas=C4=B1n=C4=B1=20sat=C4=B1r=20y?=
 =?UTF-8?q?erine=20sabit=20bloklar=20=C3=BCzerinden=20yap.?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Bu yamayla hesaplamada %15 civarında bir hızlanma görülüyor.
Heap peak aynı kalırken, heap total'de ciddi düşme var.

Çeşitli dosyalar ve bir iki paketle denedim. Faik, uygunsa
2007 branch'ına da taşıyalım bunu.

İlerde saçma try: except: kısmını da kaldırırsak olacak
bu iş :)
---
 pisi/util.py | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/pisi/util.py b/pisi/util.py
index b134f653..e60dbb78 100644
--- a/pisi/util.py
+++ b/pisi/util.py
@@ -433,8 +433,19 @@ def sha1_file(filename):
     try:
         m = sha.new()
         f = file(filename, 'rb')
-        for line in f:
-            m.update(line)
+        while True:
+            # 256 KB seems ideal for speed/memory tradeoff
+            # It wont get much faster with bigger blocks, but
+            # heap peak grows
+            block = f.read(256 * 1024)
+            if len(block) == 0:
+                # end of file
+                break
+            m.update(block)
+            # Simple trick to keep total heap even lower
+            # Delete the previous block, so while next one is read
+            # we wont have two allocated blocks with same size
+            del block
         return m.hexdigest()
     except IOError:
         raise FileError(_("I/O Error: Cannot calculate SHA1 hash of %s") % filename)