Ottaway, Jim writes:
 > > So it takes Emacs approx. half a second to decide whether or not a  
 > > file needs to be published? I didn't look at the source code yet, but  
 > > this looks a bit suspicious to me.
 > 
 > Me too, now you mention it. 
 > 

Ok, here goes. These are the results on a 1.7GHz Centrino Notebook
with 1GB of memory running FreeBSD 8.0-STABLE. The test data were a
backup of the files I use at work, so its like 10 files less but
otherwise the same.

The first run took approx. 4 min to complete, as all files needed
publishing. The ELP output of the second run with just one modified
file looks like this, with only those functions that take longer than
a second shown here:

muse-publish-markup                              11          18.625703     
1.6932457272
muse-publish-markup-region                       3           18.551402     
6.1838006666
muse-project-publish                             1           12.532707     
12.532707
muse-project-publish-default                     1           12.532433     
12.532433
muse-project-publish-file                        966         12.406346999  
0.0128430093
muse-project-publish-file-default                966         11.158744999  
0.0115514958
muse-publish-file                                966         11.097747999  
0.0114883519
muse-publish-markup-buffer                       2           9.623259      
4.8116295
muse-publish-markup-tag                          82          9.1936850000  
0.1121181097
muse-publish-mark-up-tag                         1           9.017514      
9.017514
muse-publish-markup-link                         589         7.8443390000  
0.0133180628
muse-style-element                               67687       7.7524080000  
0.0001145331
muse-publish-insert-url                          589         7.4865610000  
0.0127106298
muse-publish-url                                 589         7.4511330000  
0.0126504804
muse-publish-classify-url                        589         5.6892740000  
0.0096592088
muse-publish-link-page                           589         5.6759699999  
0.0096366213
muse-project-link-page                           589         5.6693740000  
0.0096254227
muse-find-markup-element                         7959        3.1845389999  
0.0004001179
muse-project-page-file                           589         2.8212049999  
0.0047898217
muse-collect-alist                               589         2.5992269999  
0.0044129490
muse-project-resolve-link                        589         1.3908080000  
0.0023613039
muse-style                                       73190       1.2080370000  
1.650...e-05
muse-project-applicable-styles                   1555        1.1961229999  
0.0007692109
muse-project-of-file                             591         1.0742109999  
0.0018176159

So, instead of 10 min it doesn't even take 20 seconds. Is it safe to
conclude that the file-newer-than-file-p implementation on NTEmacs is
somewhat flawed? This is the only platform-dependent function which I
can see in the muse-publish-file body. Here on FreeBSD it takes:

file-newer-than-file-p                           965         0.0586679999  
6.079...e-05

I'll check tomorrow what it says on NTEmacs.

regards,
Markus

-- 
Markus Hoenicka
http://www.mhoenicka.de
AQ score 38

_______________________________________________
Planner-el-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/planner-el-discuss

Reply via email to