diff --git a/src/filters/rclmpdf.py b/src/filters/rclmpdf.py index 66beec4a..91514a49 100755 --- a/src/filters/rclmpdf.py +++ b/src/filters/rclmpdf.py @@ -233,11 +233,8 @@ class PDFExtractor: inbody = False didcs = False output = b'' - cont = b'' isempty = True for line in input.split(b'\n'): - line = cont + line - cont = b'' if re.search(b'', line): inheader = False if re.search(b'', line): @@ -264,17 +261,11 @@ class PDFExtractor: s = line[0:1] if s != "\x0c" and s != "<": isempty = False - - # Remove end-of-line hyphenation. It's not clear that - # we should do this as pdftotext without the -layout - # option does it ? - #if re.search(r'[-]$', line): - #m = re.search(r'(.*)[ \t]([^ \t]+)$', line) - #if m: - #line = m.group(1) - #cont = m.group(2).rstrip('-') + # We used to remove end-of-line hyphenation (and join + # lines), but but it's not clear that we should do + # this as pdftotext without the -layout option does it ? line = self.em.htmlescape(line) - + if re.search(b'', line): inheader = True if re.search(b'
', line):
diff --git a/src/sampleconf/mimemap b/src/sampleconf/mimemap
index d52c11c2..9c8648f8 100644
--- a/src/sampleconf/mimemap
+++ b/src/sampleconf/mimemap
@@ -150,6 +150,9 @@
 
 .mkv = video/x-matroska
 .ogv = video/ogg
+.flv = video/x-flv
+.mp4 = video/mp4
+.ts = video/MP2T
 
 .png = image/png
 .jp2 = image/jp2
diff --git a/website/perfs.html b/website/perfs.html
index ad7312c0..b821cc96 100644
--- a/website/perfs.html
+++ b/website/perfs.html
@@ -43,9 +43,10 @@
 	store an index. Obviously, your data set will never fit one of
 	the samples, so the results cannot be exactly predicted.

 
-      
The following data was obtained on a machine with a 1800 Mhz
+      
The following very old data was obtained on a machine with a
+        1800 Mhz
 	AMD Duron CPU, 768Mb of Ram, and a 7200 RPM 160 GBytes IDE
-	disk, running Suse 10.1.

+	disk, running Suse 10.1. More recent data follows.

 
       
recollindex (version 1.8.2 with xapian 1.0.0) is
 	executed with the default flush threshold value. 
@@ -106,8 +107,74 @@
 	performance degradation. The resulting index is bigger though,
 	the exact reason is not known to me, possibly because of
 	additional fragmentation 

+
+      
There is more recent performance data (2012) at the end of
+        the article about
+          converting Recoll indexing to multithreading

+
+      
Update, March 2016: I took another sample of PDF performance
+        data on a more modern machine, with Recoll multithreading turned
+        on. The machine has an Intel Core I7-4770T Cpu, which has 4
+        physical cores, and supports hyper-threading for a total of 8
+        threads, 8 GBytes of RAM, and SSD storage (incidentally the PC is
+        fanless, this is not a "beast" computer).

+        
+      
+	
+	  
+	    
+	    
+	    
+	    
+	    
+	  
+	
+	  
+	    
+	    
+	    
+	    
+	    
+	  
+	
+      
DataData sizeIndexing timeIndex sizePeak process memory usage
Random pdfs harvested on Google

+	    Recoll 1.21.5, idxflushmb set to 200, thread
+	    parameters 6/4/1		11 GB, 5320 files3 mn 15 S400 MB545 MB

+        
+      
The indexing process used 21 mn of CPU during these 3mn15 of
+        real time, we are not letting these cores stay idle
+        much... The improvement compared to the numbers above is quite
+        spectacular (a factor of 11, approximately), mostly due to the
+        multiprocessing, but also to the faster CPU and the SSD
+        storage. Note that the peak memory value is for the
+        recollindex process, and does not take into account the
+        multiple Python and pdftotext instances (which are relatively
+        small but things add up...).

+      
+      
Improving indexing performance with hardware:

+      
I think
+      that the following multi-step approach has a good chance to
+        improve performance:
+        

       

 
+      
At some point, the index writing may become the
+        bottleneck. As far as I can think, the only possible approach
+        then is to partition the index.

+